diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,315021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 52500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.409973142857143e-05,
+      "loss": 2.4704,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4099462857142857e-05,
+      "loss": 2.1719,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4099194285714287e-05,
+      "loss": 2.4029,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4098925714285715e-05,
+      "loss": 2.6226,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4098657142857143e-05,
+      "loss": 2.5564,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4098388571428571e-05,
+      "loss": 2.5126,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4098120000000001e-05,
+      "loss": 2.556,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.409785142857143e-05,
+      "loss": 3.0154,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4097582857142857e-05,
+      "loss": 1.9816,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4097314285714286e-05,
+      "loss": 2.3849,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4097045714285716e-05,
+      "loss": 2.2227,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4096777142857142e-05,
+      "loss": 2.4171,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4096508571428572e-05,
+      "loss": 2.449,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4096240000000002e-05,
+      "loss": 2.7367,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4095971428571428e-05,
+      "loss": 2.1412,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4095702857142858e-05,
+      "loss": 2.4095,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4095434285714286e-05,
+      "loss": 2.5251,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4095165714285716e-05,
+      "loss": 2.233,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4094897142857143e-05,
+      "loss": 2.5313,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4094628571428573e-05,
+      "loss": 2.2079,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.409436e-05,
+      "loss": 2.3759,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4094091428571429e-05,
+      "loss": 2.4212,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4093822857142857e-05,
+      "loss": 2.2587,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4093554285714287e-05,
+      "loss": 2.3453,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4093285714285713e-05,
+      "loss": 2.5103,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4093017142857143e-05,
+      "loss": 2.7174,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4092748571428573e-05,
+      "loss": 2.2869,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.409248e-05,
+      "loss": 2.3436,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.409221142857143e-05,
+      "loss": 2.0717,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4091942857142858e-05,
+      "loss": 2.1793,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4091674285714288e-05,
+      "loss": 2.4808,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4091405714285714e-05,
+      "loss": 2.3491,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4091137142857144e-05,
+      "loss": 2.6348,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4090868571428572e-05,
+      "loss": 2.4105,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.40906e-05,
+      "loss": 2.0603,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4090331428571428e-05,
+      "loss": 2.7343,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4090062857142858e-05,
+      "loss": 2.5107,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4089794285714285e-05,
+      "loss": 2.526,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4089525714285715e-05,
+      "loss": 2.127,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4089257142857145e-05,
+      "loss": 2.656,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4088988571428573e-05,
+      "loss": 2.369,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4088720000000001e-05,
+      "loss": 2.3302,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4088451428571429e-05,
+      "loss": 2.4607,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4088182857142859e-05,
+      "loss": 2.1877,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4087914285714285e-05,
+      "loss": 2.1618,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4087645714285715e-05,
+      "loss": 2.3737,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4087377142857143e-05,
+      "loss": 2.4672,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4087108571428572e-05,
+      "loss": 2.5254,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.408684e-05,
+      "loss": 2.2584,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.408657142857143e-05,
+      "loss": 2.6952,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4086302857142856e-05,
+      "loss": 2.4489,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4086034285714286e-05,
+      "loss": 2.2922,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4085765714285716e-05,
+      "loss": 2.0905,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4085497142857144e-05,
+      "loss": 2.3487,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4085228571428572e-05,
+      "loss": 2.1037,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.408496e-05,
+      "loss": 2.4169,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.408469142857143e-05,
+      "loss": 2.5066,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4084422857142857e-05,
+      "loss": 2.3986,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4084154285714287e-05,
+      "loss": 2.4803,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4083885714285715e-05,
+      "loss": 2.2268,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4083617142857143e-05,
+      "loss": 2.4407,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4083348571428571e-05,
+      "loss": 2.4679,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4083080000000001e-05,
+      "loss": 2.5867,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.408281142857143e-05,
+      "loss": 2.5165,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4082542857142857e-05,
+      "loss": 2.1087,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4082274285714286e-05,
+      "loss": 2.5684,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4082005714285715e-05,
+      "loss": 2.255,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4081737142857144e-05,
+      "loss": 2.3354,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4081468571428572e-05,
+      "loss": 2.2312,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4081200000000002e-05,
+      "loss": 2.2069,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4080931428571428e-05,
+      "loss": 2.7145,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4080662857142858e-05,
+      "loss": 1.9873,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4080394285714286e-05,
+      "loss": 2.5003,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4080125714285714e-05,
+      "loss": 2.2474,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4079857142857143e-05,
+      "loss": 2.4758,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4079588571428572e-05,
+      "loss": 2.2544,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.407932e-05,
+      "loss": 2.23,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4079051428571429e-05,
+      "loss": 2.3485,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4078782857142857e-05,
+      "loss": 2.3794,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4078514285714287e-05,
+      "loss": 2.359,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4078245714285715e-05,
+      "loss": 2.2533,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4077977142857143e-05,
+      "loss": 2.6117,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4077708571428573e-05,
+      "loss": 2.21,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.407744e-05,
+      "loss": 2.0136,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.407717142857143e-05,
+      "loss": 2.4167,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4076902857142858e-05,
+      "loss": 2.0057,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4076634285714286e-05,
+      "loss": 2.5468,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4076365714285714e-05,
+      "loss": 2.498,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4076097142857144e-05,
+      "loss": 2.4728,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4075828571428572e-05,
+      "loss": 2.1734,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.407556e-05,
+      "loss": 2.3084,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4075291428571428e-05,
+      "loss": 2.3424,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4075022857142858e-05,
+      "loss": 2.5474,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4074754285714286e-05,
+      "loss": 1.8952,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4074485714285715e-05,
+      "loss": 2.3007,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4074217142857144e-05,
+      "loss": 2.2707,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4073948571428571e-05,
+      "loss": 2.0865,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.407368e-05,
+      "loss": 2.2448,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4073411428571429e-05,
+      "loss": 2.1617,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4073142857142859e-05,
+      "loss": 2.2575,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4072874285714285e-05,
+      "loss": 2.2756,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4072605714285715e-05,
+      "loss": 2.2002,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4072337142857143e-05,
+      "loss": 2.3351,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4072068571428572e-05,
+      "loss": 2.0545,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.40718e-05,
+      "loss": 2.2565,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.407153142857143e-05,
+      "loss": 2.0558,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4071262857142858e-05,
+      "loss": 1.9841,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4070994285714286e-05,
+      "loss": 2.3578,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4070725714285716e-05,
+      "loss": 2.3806,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4070457142857142e-05,
+      "loss": 2.222,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4070188571428572e-05,
+      "loss": 2.4775,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.406992e-05,
+      "loss": 2.4887,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.406965142857143e-05,
+      "loss": 2.2061,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4069382857142857e-05,
+      "loss": 2.195,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4069114285714287e-05,
+      "loss": 2.2708,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4068845714285715e-05,
+      "loss": 2.3408,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4068577142857143e-05,
+      "loss": 2.374,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4068308571428571e-05,
+      "loss": 2.2338,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4068040000000001e-05,
+      "loss": 2.1466,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4067771428571429e-05,
+      "loss": 2.1818,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4067502857142857e-05,
+      "loss": 2.2634,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4067234285714287e-05,
+      "loss": 2.2528,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4066965714285715e-05,
+      "loss": 2.2107,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4066697142857143e-05,
+      "loss": 2.2617,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4066428571428572e-05,
+      "loss": 2.076,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4066160000000002e-05,
+      "loss": 2.0209,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4065891428571428e-05,
+      "loss": 2.1236,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4065622857142858e-05,
+      "loss": 2.0479,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4065354285714286e-05,
+      "loss": 1.9634,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4065085714285714e-05,
+      "loss": 2.2957,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4064817142857142e-05,
+      "loss": 2.182,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4064548571428572e-05,
+      "loss": 2.1956,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.406428e-05,
+      "loss": 2.1467,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4064011428571429e-05,
+      "loss": 2.1747,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4063742857142859e-05,
+      "loss": 2.1132,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4063474285714287e-05,
+      "loss": 2.0595,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4063205714285715e-05,
+      "loss": 2.0583,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4062937142857143e-05,
+      "loss": 2.1697,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4062668571428573e-05,
+      "loss": 2.7655,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.40624e-05,
+      "loss": 2.0558,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.406213142857143e-05,
+      "loss": 2.6233,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4061862857142857e-05,
+      "loss": 2.2175,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4061594285714286e-05,
+      "loss": 2.2232,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4061325714285714e-05,
+      "loss": 1.9576,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4061057142857144e-05,
+      "loss": 1.9168,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4060788571428574e-05,
+      "loss": 2.3891,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.406052e-05,
+      "loss": 2.1752,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.406025142857143e-05,
+      "loss": 2.2035,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4059982857142858e-05,
+      "loss": 2.0642,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4059714285714286e-05,
+      "loss": 2.2762,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4059445714285714e-05,
+      "loss": 2.0637,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4059177142857144e-05,
+      "loss": 2.1591,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.405890857142857e-05,
+      "loss": 2.5024,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.405864e-05,
+      "loss": 2.2015,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4058371428571429e-05,
+      "loss": 2.3999,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4058102857142857e-05,
+      "loss": 2.0475,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4057834285714285e-05,
+      "loss": 2.1259,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4057565714285715e-05,
+      "loss": 2.0036,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4057297142857145e-05,
+      "loss": 2.0724,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4057028571428571e-05,
+      "loss": 2.1608,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4056760000000001e-05,
+      "loss": 2.1209,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.405649142857143e-05,
+      "loss": 2.1369,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4056222857142858e-05,
+      "loss": 2.2366,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4055954285714286e-05,
+      "loss": 2.4772,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4055685714285716e-05,
+      "loss": 1.9412,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4055417142857142e-05,
+      "loss": 1.8935,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4055148571428572e-05,
+      "loss": 2.3573,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.405488e-05,
+      "loss": 2.1973,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.405461142857143e-05,
+      "loss": 2.2914,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4054342857142857e-05,
+      "loss": 2.3378,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4054074285714286e-05,
+      "loss": 2.2694,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4053805714285715e-05,
+      "loss": 2.0797,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4053537142857143e-05,
+      "loss": 2.4416,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4053268571428573e-05,
+      "loss": 2.2632,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4053e-05,
+      "loss": 2.1921,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4052731428571429e-05,
+      "loss": 2.2821,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4052462857142857e-05,
+      "loss": 1.9742,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4052194285714287e-05,
+      "loss": 2.2223,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4051925714285714e-05,
+      "loss": 1.8724,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4051657142857143e-05,
+      "loss": 2.3624,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4051388571428572e-05,
+      "loss": 2.1866,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4051120000000001e-05,
+      "loss": 2.4879,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4050851428571428e-05,
+      "loss": 2.2866,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4050582857142858e-05,
+      "loss": 2.2845,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4050314285714286e-05,
+      "loss": 1.9321,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4050045714285714e-05,
+      "loss": 2.1716,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4049777142857144e-05,
+      "loss": 1.9788,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4049508571428572e-05,
+      "loss": 2.1002,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.404924e-05,
+      "loss": 1.9806,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4048971428571429e-05,
+      "loss": 2.0065,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4048702857142858e-05,
+      "loss": 2.306,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4048434285714287e-05,
+      "loss": 2.0682,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4048165714285715e-05,
+      "loss": 2.0873,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4047897142857143e-05,
+      "loss": 1.9564,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4047628571428573e-05,
+      "loss": 2.2787,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.404736e-05,
+      "loss": 1.9624,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4047091428571429e-05,
+      "loss": 1.9133,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4046822857142857e-05,
+      "loss": 1.9734,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4046554285714286e-05,
+      "loss": 2.0805,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4046285714285715e-05,
+      "loss": 2.1329,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4046017142857144e-05,
+      "loss": 2.2763,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4045748571428572e-05,
+      "loss": 2.2048,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.404548e-05,
+      "loss": 2.1463,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.404521142857143e-05,
+      "loss": 2.2368,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4044942857142858e-05,
+      "loss": 2.0212,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4044674285714286e-05,
+      "loss": 2.3456,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4044405714285714e-05,
+      "loss": 2.558,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4044137142857144e-05,
+      "loss": 2.1227,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.404386857142857e-05,
+      "loss": 2.4837,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.40436e-05,
+      "loss": 2.3579,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4043331428571429e-05,
+      "loss": 2.1119,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4043062857142857e-05,
+      "loss": 1.9488,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4042794285714287e-05,
+      "loss": 2.2437,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4042525714285715e-05,
+      "loss": 2.1967,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4042257142857145e-05,
+      "loss": 2.1971,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4041988571428571e-05,
+      "loss": 2.1658,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4041720000000001e-05,
+      "loss": 1.9174,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.404145142857143e-05,
+      "loss": 2.3429,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4041182857142857e-05,
+      "loss": 2.1856,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4040914285714286e-05,
+      "loss": 2.2029,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4040645714285716e-05,
+      "loss": 2.1768,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4040377142857142e-05,
+      "loss": 2.2991,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4040108571428572e-05,
+      "loss": 1.8148,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.403984e-05,
+      "loss": 2.2307,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4039571428571428e-05,
+      "loss": 2.047,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4039302857142858e-05,
+      "loss": 2.3129,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4039034285714286e-05,
+      "loss": 2.1007,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4038765714285716e-05,
+      "loss": 2.5192,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4038497142857143e-05,
+      "loss": 2.2632,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4038228571428573e-05,
+      "loss": 1.9443,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.403796e-05,
+      "loss": 2.4049,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4037691428571429e-05,
+      "loss": 2.2578,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4037422857142857e-05,
+      "loss": 2.2356,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4037154285714287e-05,
+      "loss": 2.0465,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4036885714285713e-05,
+      "loss": 2.0505,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4036617142857143e-05,
+      "loss": 2.2809,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4036348571428571e-05,
+      "loss": 1.8666,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4036080000000001e-05,
+      "loss": 2.2967,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.403581142857143e-05,
+      "loss": 2.1487,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4035542857142858e-05,
+      "loss": 2.1722,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4035274285714288e-05,
+      "loss": 2.0628,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4035005714285714e-05,
+      "loss": 2.0917,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4034737142857144e-05,
+      "loss": 2.2436,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4034468571428572e-05,
+      "loss": 2.0785,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.40342e-05,
+      "loss": 1.8802,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4033931428571428e-05,
+      "loss": 2.2039,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4033662857142858e-05,
+      "loss": 2.1812,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4033394285714285e-05,
+      "loss": 2.3214,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4033125714285715e-05,
+      "loss": 2.243,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4032857142857143e-05,
+      "loss": 2.1313,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4032588571428573e-05,
+      "loss": 2.1437,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4032320000000001e-05,
+      "loss": 2.3323,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4032051428571429e-05,
+      "loss": 2.3721,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4031782857142859e-05,
+      "loss": 2.2878,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4031514285714285e-05,
+      "loss": 2.5014,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4031245714285715e-05,
+      "loss": 2.2944,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4030977142857143e-05,
+      "loss": 2.2923,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4030708571428572e-05,
+      "loss": 2.5002,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.403044e-05,
+      "loss": 2.3946,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.403017142857143e-05,
+      "loss": 2.4195,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4029902857142856e-05,
+      "loss": 1.8768,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4029634285714286e-05,
+      "loss": 1.9211,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4029365714285714e-05,
+      "loss": 2.2478,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4029097142857144e-05,
+      "loss": 2.0345,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4028828571428572e-05,
+      "loss": 1.8171,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.402856e-05,
+      "loss": 2.1928,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.402829142857143e-05,
+      "loss": 2.2263,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4028022857142857e-05,
+      "loss": 2.1083,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4027754285714287e-05,
+      "loss": 2.4028,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4027485714285715e-05,
+      "loss": 2.2319,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4027217142857143e-05,
+      "loss": 2.176,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4026948571428571e-05,
+      "loss": 1.9382,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4026680000000001e-05,
+      "loss": 2.18,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.402641142857143e-05,
+      "loss": 1.997,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4026142857142857e-05,
+      "loss": 1.9802,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4025874285714286e-05,
+      "loss": 1.9645,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4025605714285715e-05,
+      "loss": 2.1028,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4025337142857144e-05,
+      "loss": 2.0328,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4025068571428572e-05,
+      "loss": 2.1083,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4024800000000002e-05,
+      "loss": 1.8856,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4024531428571428e-05,
+      "loss": 2.1594,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4024262857142858e-05,
+      "loss": 2.035,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4023994285714286e-05,
+      "loss": 1.8584,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4023725714285714e-05,
+      "loss": 2.1632,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4023457142857143e-05,
+      "loss": 1.974,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4023188571428572e-05,
+      "loss": 2.4741,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.402292e-05,
+      "loss": 2.3177,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4022651428571429e-05,
+      "loss": 1.9625,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4022382857142857e-05,
+      "loss": 2.0282,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4022114285714287e-05,
+      "loss": 2.2494,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4021845714285715e-05,
+      "loss": 2.1761,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4021577142857143e-05,
+      "loss": 2.1101,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4021308571428573e-05,
+      "loss": 2.1156,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.402104e-05,
+      "loss": 2.0997,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.402077142857143e-05,
+      "loss": 2.3998,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4020502857142858e-05,
+      "loss": 1.9402,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4020234285714287e-05,
+      "loss": 1.767,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4019965714285714e-05,
+      "loss": 2.2199,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4019697142857144e-05,
+      "loss": 2.1948,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4019428571428572e-05,
+      "loss": 2.0103,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.401916e-05,
+      "loss": 2.2076,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4018891428571428e-05,
+      "loss": 2.0006,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4018622857142858e-05,
+      "loss": 2.201,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4018354285714286e-05,
+      "loss": 1.989,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4018085714285715e-05,
+      "loss": 2.3027,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4017817142857144e-05,
+      "loss": 2.1946,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4017548571428571e-05,
+      "loss": 2.1091,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.401728e-05,
+      "loss": 1.9183,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4017011428571429e-05,
+      "loss": 2.1565,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4016742857142859e-05,
+      "loss": 1.9146,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4016474285714285e-05,
+      "loss": 2.1331,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4016205714285715e-05,
+      "loss": 2.0675,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4015937142857143e-05,
+      "loss": 2.0308,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4015668571428572e-05,
+      "loss": 2.0529,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.40154e-05,
+      "loss": 2.1318,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.401513142857143e-05,
+      "loss": 1.7694,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4014862857142858e-05,
+      "loss": 2.2692,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4014594285714286e-05,
+      "loss": 2.0856,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4014325714285716e-05,
+      "loss": 2.0765,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4014057142857144e-05,
+      "loss": 1.8858,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4013788571428572e-05,
+      "loss": 1.9112,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.401352e-05,
+      "loss": 2.169,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.401325142857143e-05,
+      "loss": 2.1483,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4012982857142857e-05,
+      "loss": 1.9247,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4012714285714287e-05,
+      "loss": 2.3503,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4012445714285715e-05,
+      "loss": 2.2712,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4012177142857143e-05,
+      "loss": 1.9283,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4011908571428571e-05,
+      "loss": 2.0429,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4011640000000001e-05,
+      "loss": 2.0685,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4011371428571429e-05,
+      "loss": 2.2147,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4011102857142857e-05,
+      "loss": 2.41,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4010834285714287e-05,
+      "loss": 2.0999,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4010565714285715e-05,
+      "loss": 2.0259,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4010297142857143e-05,
+      "loss": 1.9982,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4010028571428572e-05,
+      "loss": 1.7386,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4009760000000002e-05,
+      "loss": 2.0731,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4009491428571428e-05,
+      "loss": 2.0856,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4009222857142858e-05,
+      "loss": 1.961,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4008954285714286e-05,
+      "loss": 2.1344,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4008685714285714e-05,
+      "loss": 1.9244,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4008417142857142e-05,
+      "loss": 2.3038,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4008148571428572e-05,
+      "loss": 2.204,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400788e-05,
+      "loss": 2.02,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4007611428571429e-05,
+      "loss": 1.9181,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4007342857142859e-05,
+      "loss": 2.0829,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4007074285714287e-05,
+      "loss": 2.1456,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4006805714285715e-05,
+      "loss": 2.2623,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4006537142857143e-05,
+      "loss": 2.187,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4006268571428573e-05,
+      "loss": 1.9866,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4006e-05,
+      "loss": 1.9356,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400573142857143e-05,
+      "loss": 2.4457,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4005462857142857e-05,
+      "loss": 2.0941,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4005194285714286e-05,
+      "loss": 2.2812,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4004925714285714e-05,
+      "loss": 2.0746,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4004657142857144e-05,
+      "loss": 2.101,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4004388571428572e-05,
+      "loss": 2.3593,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400412e-05,
+      "loss": 1.8508,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400385142857143e-05,
+      "loss": 2.0159,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4003582857142858e-05,
+      "loss": 2.0302,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4003314285714286e-05,
+      "loss": 2.0289,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4003045714285714e-05,
+      "loss": 1.6124,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4002777142857144e-05,
+      "loss": 2.0513,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400250857142857e-05,
+      "loss": 2.0163,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400224e-05,
+      "loss": 1.8413,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4001971428571429e-05,
+      "loss": 2.1579,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4001702857142859e-05,
+      "loss": 2.0632,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4001434285714285e-05,
+      "loss": 1.9919,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4001165714285715e-05,
+      "loss": 2.1156,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4000897142857143e-05,
+      "loss": 1.7572,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4000628571428571e-05,
+      "loss": 2.0903,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4000360000000001e-05,
+      "loss": 1.948,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400009142857143e-05,
+      "loss": 2.2282,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3999822857142858e-05,
+      "loss": 1.9008,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3999554285714286e-05,
+      "loss": 1.899,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3999285714285716e-05,
+      "loss": 1.9418,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3999017142857142e-05,
+      "loss": 1.6516,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3998748571428572e-05,
+      "loss": 1.9032,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.399848e-05,
+      "loss": 2.0788,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.399821142857143e-05,
+      "loss": 1.9618,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3997942857142857e-05,
+      "loss": 2.2681,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3997674285714286e-05,
+      "loss": 1.8962,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3997405714285715e-05,
+      "loss": 2.1241,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3997137142857143e-05,
+      "loss": 1.8675,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3996868571428573e-05,
+      "loss": 2.3138,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.39966e-05,
+      "loss": 2.048,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3996331428571429e-05,
+      "loss": 2.0244,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3996062857142857e-05,
+      "loss": 1.8232,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3995794285714287e-05,
+      "loss": 2.22,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3995525714285715e-05,
+      "loss": 1.9704,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3995257142857143e-05,
+      "loss": 1.9608,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3994988571428572e-05,
+      "loss": 2.2766,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3994720000000001e-05,
+      "loss": 1.9085,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3994451428571428e-05,
+      "loss": 2.0045,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3994182857142858e-05,
+      "loss": 1.6102,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3993914285714286e-05,
+      "loss": 2.1883,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3993645714285714e-05,
+      "loss": 2.0365,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3993377142857144e-05,
+      "loss": 2.033,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3993108571428572e-05,
+      "loss": 2.0205,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.399284e-05,
+      "loss": 1.993,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3992571428571429e-05,
+      "loss": 2.1209,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3992302857142858e-05,
+      "loss": 1.858,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3992034285714287e-05,
+      "loss": 1.9865,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3991765714285715e-05,
+      "loss": 2.0166,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3991497142857143e-05,
+      "loss": 2.0315,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3991228571428573e-05,
+      "loss": 2.0684,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.399096e-05,
+      "loss": 1.974,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3990691428571429e-05,
+      "loss": 1.8667,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3990422857142857e-05,
+      "loss": 2.3502,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3990154285714286e-05,
+      "loss": 1.9249,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3989885714285715e-05,
+      "loss": 2.1226,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3989617142857144e-05,
+      "loss": 1.9637,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3989348571428573e-05,
+      "loss": 2.0607,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.398908e-05,
+      "loss": 2.0634,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.398881142857143e-05,
+      "loss": 2.421,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3988542857142858e-05,
+      "loss": 2.1423,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3988274285714286e-05,
+      "loss": 2.0502,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3988005714285714e-05,
+      "loss": 2.2872,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3987737142857144e-05,
+      "loss": 2.2124,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.398746857142857e-05,
+      "loss": 1.8576,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.39872e-05,
+      "loss": 1.935,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3986931428571429e-05,
+      "loss": 1.8961,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3986662857142857e-05,
+      "loss": 1.7612,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3986394285714287e-05,
+      "loss": 2.0684,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3986125714285715e-05,
+      "loss": 2.1569,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3985857142857145e-05,
+      "loss": 1.9355,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3985588571428571e-05,
+      "loss": 1.9545,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3985320000000001e-05,
+      "loss": 1.8821,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.398505142857143e-05,
+      "loss": 2.2343,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3984782857142857e-05,
+      "loss": 2.1229,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3984514285714286e-05,
+      "loss": 2.0165,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3984245714285716e-05,
+      "loss": 2.2319,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3983977142857142e-05,
+      "loss": 2.2475,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3983708571428572e-05,
+      "loss": 2.0774,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.398344e-05,
+      "loss": 2.0234,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3983171428571428e-05,
+      "loss": 2.067,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3982902857142858e-05,
+      "loss": 2.2176,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3982634285714286e-05,
+      "loss": 1.9642,
+      "step": 437
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3982365714285716e-05,
+      "loss": 2.0677,
+      "step": 438
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3982097142857143e-05,
+      "loss": 2.0972,
+      "step": 439
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3981828571428573e-05,
+      "loss": 1.9747,
+      "step": 440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.398156e-05,
+      "loss": 1.9903,
+      "step": 441
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3981291428571429e-05,
+      "loss": 2.1029,
+      "step": 442
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3981022857142857e-05,
+      "loss": 2.149,
+      "step": 443
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3980754285714287e-05,
+      "loss": 1.9419,
+      "step": 444
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3980485714285713e-05,
+      "loss": 2.1757,
+      "step": 445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3980217142857143e-05,
+      "loss": 2.1239,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3979948571428571e-05,
+      "loss": 1.9009,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3979680000000001e-05,
+      "loss": 1.8881,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.397941142857143e-05,
+      "loss": 1.8345,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3979142857142858e-05,
+      "loss": 1.9666,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3978874285714288e-05,
+      "loss": 1.895,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3978605714285714e-05,
+      "loss": 2.3325,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3978337142857144e-05,
+      "loss": 1.998,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3978068571428572e-05,
+      "loss": 2.1138,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.39778e-05,
+      "loss": 1.9084,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3977531428571428e-05,
+      "loss": 1.7856,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3977262857142858e-05,
+      "loss": 2.0749,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3976994285714285e-05,
+      "loss": 2.107,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3976725714285715e-05,
+      "loss": 2.3436,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3976457142857143e-05,
+      "loss": 1.9772,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3976188571428573e-05,
+      "loss": 2.2179,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3975920000000001e-05,
+      "loss": 1.7511,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3975651428571429e-05,
+      "loss": 2.1492,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3975382857142859e-05,
+      "loss": 1.8804,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3975114285714285e-05,
+      "loss": 2.2666,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3974845714285715e-05,
+      "loss": 1.9866,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3974577142857143e-05,
+      "loss": 1.9873,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3974308571428572e-05,
+      "loss": 1.8686,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.397404e-05,
+      "loss": 1.987,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.397377142857143e-05,
+      "loss": 1.736,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3973502857142858e-05,
+      "loss": 2.0589,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3973234285714286e-05,
+      "loss": 1.9688,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3972965714285714e-05,
+      "loss": 1.8936,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3972697142857144e-05,
+      "loss": 2.1826,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3972428571428572e-05,
+      "loss": 2.1911,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.397216e-05,
+      "loss": 1.9877,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.397189142857143e-05,
+      "loss": 2.1421,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3971622857142857e-05,
+      "loss": 1.9906,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3971354285714287e-05,
+      "loss": 1.8778,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3971085714285715e-05,
+      "loss": 2.0446,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3970817142857143e-05,
+      "loss": 1.9057,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3970548571428571e-05,
+      "loss": 1.7625,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3970280000000001e-05,
+      "loss": 2.3107,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.397001142857143e-05,
+      "loss": 2.1659,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3969742857142857e-05,
+      "loss": 2.0275,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3969474285714286e-05,
+      "loss": 2.1009,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3969205714285715e-05,
+      "loss": 1.8244,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3968937142857144e-05,
+      "loss": 2.1911,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3968668571428572e-05,
+      "loss": 2.1924,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3968400000000002e-05,
+      "loss": 2.0863,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3968131428571428e-05,
+      "loss": 2.0048,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3967862857142858e-05,
+      "loss": 2.0046,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3967594285714286e-05,
+      "loss": 1.8893,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3967325714285716e-05,
+      "loss": 1.9584,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3967057142857143e-05,
+      "loss": 1.7593,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3966788571428572e-05,
+      "loss": 2.1624,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.396652e-05,
+      "loss": 1.8516,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3966251428571429e-05,
+      "loss": 2.082,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3965982857142857e-05,
+      "loss": 1.8199,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3965714285714287e-05,
+      "loss": 2.0895,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3965445714285715e-05,
+      "loss": 1.9236,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3965177142857143e-05,
+      "loss": 1.9638,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3964908571428573e-05,
+      "loss": 2.1058,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.396464e-05,
+      "loss": 1.7087,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.396437142857143e-05,
+      "loss": 1.8773,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3964102857142858e-05,
+      "loss": 2.0194,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3963834285714287e-05,
+      "loss": 1.9797,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3963565714285714e-05,
+      "loss": 1.7745,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3963297142857144e-05,
+      "loss": 1.907,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3963028571428572e-05,
+      "loss": 2.0947,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.396276e-05,
+      "loss": 1.8496,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3962491428571428e-05,
+      "loss": 1.9484,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3962222857142858e-05,
+      "loss": 1.787,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3961954285714286e-05,
+      "loss": 1.8771,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3961685714285715e-05,
+      "loss": 2.0291,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3961417142857144e-05,
+      "loss": 2.173,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3961148571428573e-05,
+      "loss": 2.4175,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.396088e-05,
+      "loss": 1.8573,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3960611428571429e-05,
+      "loss": 1.6708,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3960342857142859e-05,
+      "loss": 2.0539,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3960074285714285e-05,
+      "loss": 2.0835,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3959805714285715e-05,
+      "loss": 2.1886,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3959537142857143e-05,
+      "loss": 1.8448,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3959268571428571e-05,
+      "loss": 2.1843,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3959e-05,
+      "loss": 2.3223,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.395873142857143e-05,
+      "loss": 1.7144,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3958462857142858e-05,
+      "loss": 2.0091,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3958194285714286e-05,
+      "loss": 2.0293,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3957925714285716e-05,
+      "loss": 2.0418,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3957657142857144e-05,
+      "loss": 2.0483,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3957388571428572e-05,
+      "loss": 2.259,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.395712e-05,
+      "loss": 1.8377,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.395685142857143e-05,
+      "loss": 1.8235,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3956582857142857e-05,
+      "loss": 2.2228,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3956314285714287e-05,
+      "loss": 1.854,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3956045714285715e-05,
+      "loss": 1.8891,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3955777142857143e-05,
+      "loss": 2.1827,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3955508571428571e-05,
+      "loss": 1.6326,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3955240000000001e-05,
+      "loss": 1.8553,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3954971428571429e-05,
+      "loss": 1.8786,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3954702857142857e-05,
+      "loss": 1.8672,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3954434285714287e-05,
+      "loss": 1.6969,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3954165714285715e-05,
+      "loss": 1.9552,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3953897142857143e-05,
+      "loss": 1.9913,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3953628571428572e-05,
+      "loss": 2.1777,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3953360000000002e-05,
+      "loss": 2.1291,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3953091428571428e-05,
+      "loss": 1.971,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3952822857142858e-05,
+      "loss": 1.8759,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3952554285714286e-05,
+      "loss": 2.1907,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3952285714285714e-05,
+      "loss": 1.9291,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3952017142857142e-05,
+      "loss": 2.1397,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3951748571428572e-05,
+      "loss": 1.8438,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.395148e-05,
+      "loss": 1.8316,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3951211428571429e-05,
+      "loss": 1.8142,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3950942857142859e-05,
+      "loss": 2.0906,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3950674285714287e-05,
+      "loss": 1.9185,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3950405714285715e-05,
+      "loss": 2.1122,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3950137142857143e-05,
+      "loss": 1.9335,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3949868571428573e-05,
+      "loss": 2.4092,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.39496e-05,
+      "loss": 2.2183,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394933142857143e-05,
+      "loss": 2.0952,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3949062857142857e-05,
+      "loss": 2.0305,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3948794285714287e-05,
+      "loss": 1.7693,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3948525714285714e-05,
+      "loss": 1.9388,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3948257142857144e-05,
+      "loss": 2.0108,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3947988571428572e-05,
+      "loss": 1.9088,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394772e-05,
+      "loss": 1.9387,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394745142857143e-05,
+      "loss": 2.0387,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3947182857142858e-05,
+      "loss": 1.8318,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3946914285714286e-05,
+      "loss": 1.6023,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3946645714285714e-05,
+      "loss": 1.9861,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3946377142857144e-05,
+      "loss": 2.1646,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394610857142857e-05,
+      "loss": 2.1462,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394584e-05,
+      "loss": 2.0914,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3945571428571429e-05,
+      "loss": 2.1116,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3945302857142859e-05,
+      "loss": 1.9791,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3945034285714285e-05,
+      "loss": 1.9422,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3944765714285715e-05,
+      "loss": 2.1879,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3944497142857143e-05,
+      "loss": 1.7946,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3944228571428571e-05,
+      "loss": 1.9753,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3943960000000001e-05,
+      "loss": 2.1392,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394369142857143e-05,
+      "loss": 2.0445,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3943422857142858e-05,
+      "loss": 1.6616,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3943154285714286e-05,
+      "loss": 1.9845,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3942885714285716e-05,
+      "loss": 1.757,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3942617142857142e-05,
+      "loss": 1.9844,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3942348571428572e-05,
+      "loss": 2.0477,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394208e-05,
+      "loss": 2.1583,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.394181142857143e-05,
+      "loss": 1.7525,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3941542857142857e-05,
+      "loss": 1.7769,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3941274285714286e-05,
+      "loss": 2.0017,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3941005714285715e-05,
+      "loss": 1.8492,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3940737142857143e-05,
+      "loss": 1.8165,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3940468571428573e-05,
+      "loss": 2.0202,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.39402e-05,
+      "loss": 1.7458,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3939931428571429e-05,
+      "loss": 1.9208,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3939662857142857e-05,
+      "loss": 1.7242,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3939394285714287e-05,
+      "loss": 2.1561,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3939125714285715e-05,
+      "loss": 1.8999,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3938857142857143e-05,
+      "loss": 1.7788,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3938588571428572e-05,
+      "loss": 1.9655,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3938320000000001e-05,
+      "loss": 2.3672,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3938051428571428e-05,
+      "loss": 1.9459,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3937782857142858e-05,
+      "loss": 1.7808,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3937514285714286e-05,
+      "loss": 1.8844,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3937245714285714e-05,
+      "loss": 2.1531,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3936977142857144e-05,
+      "loss": 1.9201,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3936708571428572e-05,
+      "loss": 1.8035,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.393644e-05,
+      "loss": 1.8384,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3936171428571429e-05,
+      "loss": 2.0931,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3935902857142858e-05,
+      "loss": 1.9004,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3935634285714287e-05,
+      "loss": 2.0839,
+      "step": 612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3935365714285715e-05,
+      "loss": 1.809,
+      "step": 613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3935097142857143e-05,
+      "loss": 1.9751,
+      "step": 614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3934828571428573e-05,
+      "loss": 1.9756,
+      "step": 615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.393456e-05,
+      "loss": 1.9682,
+      "step": 616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3934291428571429e-05,
+      "loss": 2.2276,
+      "step": 617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3934022857142857e-05,
+      "loss": 1.9492,
+      "step": 618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3933754285714286e-05,
+      "loss": 2.2277,
+      "step": 619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3933485714285715e-05,
+      "loss": 2.0832,
+      "step": 620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3933217142857144e-05,
+      "loss": 2.108,
+      "step": 621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3932948571428573e-05,
+      "loss": 1.9705,
+      "step": 622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.393268e-05,
+      "loss": 1.7998,
+      "step": 623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.393241142857143e-05,
+      "loss": 1.6927,
+      "step": 624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3932142857142858e-05,
+      "loss": 2.11,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3931874285714286e-05,
+      "loss": 2.0711,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3931605714285714e-05,
+      "loss": 1.8901,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3931337142857144e-05,
+      "loss": 1.9968,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.393106857142857e-05,
+      "loss": 1.652,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.39308e-05,
+      "loss": 2.152,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3930531428571429e-05,
+      "loss": 2.0659,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3930262857142857e-05,
+      "loss": 2.2895,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3929994285714287e-05,
+      "loss": 2.1172,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3929725714285715e-05,
+      "loss": 2.2986,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3929457142857145e-05,
+      "loss": 1.8745,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3929188571428571e-05,
+      "loss": 1.8258,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3928920000000001e-05,
+      "loss": 1.647,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.392865142857143e-05,
+      "loss": 1.8754,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3928382857142857e-05,
+      "loss": 1.8728,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3928114285714286e-05,
+      "loss": 1.9788,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3927845714285716e-05,
+      "loss": 1.8488,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3927577142857142e-05,
+      "loss": 1.9725,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3927308571428572e-05,
+      "loss": 2.1291,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.392704e-05,
+      "loss": 1.8271,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.392677142857143e-05,
+      "loss": 2.1179,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3926502857142858e-05,
+      "loss": 1.6368,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3926234285714286e-05,
+      "loss": 2.0355,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3925965714285716e-05,
+      "loss": 1.7604,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3925697142857143e-05,
+      "loss": 1.9396,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3925428571428573e-05,
+      "loss": 1.7701,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.392516e-05,
+      "loss": 1.9472,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3924891428571429e-05,
+      "loss": 1.6567,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3924622857142857e-05,
+      "loss": 2.0684,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3924354285714287e-05,
+      "loss": 1.8844,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3924085714285713e-05,
+      "loss": 1.8916,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3923817142857143e-05,
+      "loss": 2.1616,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3923548571428571e-05,
+      "loss": 1.9869,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3923280000000001e-05,
+      "loss": 2.0383,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.392301142857143e-05,
+      "loss": 1.9009,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3922742857142858e-05,
+      "loss": 1.8381,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3922474285714288e-05,
+      "loss": 2.0626,
+      "step": 661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3922205714285714e-05,
+      "loss": 1.9158,
+      "step": 662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3921937142857144e-05,
+      "loss": 1.9425,
+      "step": 663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3921668571428572e-05,
+      "loss": 2.1881,
+      "step": 664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.39214e-05,
+      "loss": 2.0291,
+      "step": 665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3921131428571428e-05,
+      "loss": 1.9894,
+      "step": 666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3920862857142858e-05,
+      "loss": 1.9193,
+      "step": 667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3920594285714286e-05,
+      "loss": 1.8414,
+      "step": 668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3920325714285715e-05,
+      "loss": 2.0694,
+      "step": 669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3920057142857143e-05,
+      "loss": 1.9669,
+      "step": 670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3919788571428573e-05,
+      "loss": 1.7748,
+      "step": 671
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3919520000000001e-05,
+      "loss": 1.8088,
+      "step": 672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3919251428571429e-05,
+      "loss": 1.8411,
+      "step": 673
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3918982857142859e-05,
+      "loss": 2.1309,
+      "step": 674
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3918714285714285e-05,
+      "loss": 1.9768,
+      "step": 675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3918445714285715e-05,
+      "loss": 2.0242,
+      "step": 676
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3918177142857143e-05,
+      "loss": 2.1636,
+      "step": 677
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3917908571428572e-05,
+      "loss": 1.9223,
+      "step": 678
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.391764e-05,
+      "loss": 1.8586,
+      "step": 679
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.391737142857143e-05,
+      "loss": 1.9656,
+      "step": 680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3917102857142858e-05,
+      "loss": 2.2567,
+      "step": 681
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3916834285714286e-05,
+      "loss": 1.9428,
+      "step": 682
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3916565714285714e-05,
+      "loss": 2.1767,
+      "step": 683
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3916297142857144e-05,
+      "loss": 2.1774,
+      "step": 684
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3916028571428572e-05,
+      "loss": 1.8307,
+      "step": 685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.391576e-05,
+      "loss": 1.9812,
+      "step": 686
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.391549142857143e-05,
+      "loss": 2.1726,
+      "step": 687
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3915222857142857e-05,
+      "loss": 1.8376,
+      "step": 688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3914954285714287e-05,
+      "loss": 1.7729,
+      "step": 689
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3914685714285715e-05,
+      "loss": 1.9563,
+      "step": 690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3914417142857145e-05,
+      "loss": 1.75,
+      "step": 691
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3914148571428571e-05,
+      "loss": 1.9423,
+      "step": 692
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3913880000000001e-05,
+      "loss": 1.8998,
+      "step": 693
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.391361142857143e-05,
+      "loss": 2.1492,
+      "step": 694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3913342857142857e-05,
+      "loss": 1.7381,
+      "step": 695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3913074285714286e-05,
+      "loss": 2.1013,
+      "step": 696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3912805714285715e-05,
+      "loss": 1.8736,
+      "step": 697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3912537142857144e-05,
+      "loss": 2.1172,
+      "step": 698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3912268571428572e-05,
+      "loss": 2.12,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3912000000000002e-05,
+      "loss": 1.6337,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3911731428571428e-05,
+      "loss": 1.8776,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3911462857142858e-05,
+      "loss": 1.8234,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3911194285714286e-05,
+      "loss": 2.087,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3910925714285716e-05,
+      "loss": 1.853,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3910657142857143e-05,
+      "loss": 2.095,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3910388571428572e-05,
+      "loss": 2.1468,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.391012e-05,
+      "loss": 1.8569,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3909851428571429e-05,
+      "loss": 2.067,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3909582857142857e-05,
+      "loss": 1.6271,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3909314285714287e-05,
+      "loss": 2.298,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3909045714285715e-05,
+      "loss": 1.9229,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3908777142857143e-05,
+      "loss": 2.1842,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3908508571428573e-05,
+      "loss": 2.0065,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3908240000000001e-05,
+      "loss": 2.084,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.390797142857143e-05,
+      "loss": 1.9182,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3907702857142858e-05,
+      "loss": 1.7741,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3907434285714287e-05,
+      "loss": 2.0103,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3907165714285714e-05,
+      "loss": 1.9574,
+      "step": 718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3906897142857144e-05,
+      "loss": 1.8351,
+      "step": 719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3906628571428572e-05,
+      "loss": 2.0117,
+      "step": 720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.390636e-05,
+      "loss": 1.9781,
+      "step": 721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3906091428571428e-05,
+      "loss": 2.0344,
+      "step": 722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3905822857142858e-05,
+      "loss": 1.4859,
+      "step": 723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3905554285714286e-05,
+      "loss": 2.248,
+      "step": 724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3905285714285715e-05,
+      "loss": 1.9697,
+      "step": 725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3905017142857144e-05,
+      "loss": 1.6722,
+      "step": 726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3904748571428573e-05,
+      "loss": 1.8812,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.390448e-05,
+      "loss": 1.8067,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3904211428571429e-05,
+      "loss": 1.8367,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3903942857142859e-05,
+      "loss": 1.7424,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3903674285714285e-05,
+      "loss": 1.921,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3903405714285715e-05,
+      "loss": 2.0529,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3903137142857143e-05,
+      "loss": 1.7939,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3902868571428571e-05,
+      "loss": 1.825,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.39026e-05,
+      "loss": 1.9872,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.390233142857143e-05,
+      "loss": 1.7326,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3902062857142858e-05,
+      "loss": 2.0646,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3901794285714286e-05,
+      "loss": 1.8911,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3901525714285716e-05,
+      "loss": 2.0604,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3901257142857144e-05,
+      "loss": 2.0509,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3900988571428572e-05,
+      "loss": 1.8363,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.390072e-05,
+      "loss": 1.9085,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.390045142857143e-05,
+      "loss": 1.6426,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3900182857142857e-05,
+      "loss": 1.8473,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3899914285714287e-05,
+      "loss": 1.9003,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3899645714285715e-05,
+      "loss": 1.5581,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3899377142857143e-05,
+      "loss": 1.713,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3899108571428571e-05,
+      "loss": 1.7131,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3898840000000001e-05,
+      "loss": 1.859,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3898571428571429e-05,
+      "loss": 1.9208,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3898302857142857e-05,
+      "loss": 1.6221,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3898034285714287e-05,
+      "loss": 2.1268,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3897765714285715e-05,
+      "loss": 1.8533,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3897497142857143e-05,
+      "loss": 1.6631,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3897228571428572e-05,
+      "loss": 2.0805,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3896960000000002e-05,
+      "loss": 1.6775,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3896691428571428e-05,
+      "loss": 2.1374,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3896422857142858e-05,
+      "loss": 1.9421,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3896154285714286e-05,
+      "loss": 1.9905,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3895885714285714e-05,
+      "loss": 1.8443,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3895617142857142e-05,
+      "loss": 1.6955,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3895348571428572e-05,
+      "loss": 1.672,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.389508e-05,
+      "loss": 1.7026,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3894811428571429e-05,
+      "loss": 1.9158,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3894542857142859e-05,
+      "loss": 1.9159,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3894274285714287e-05,
+      "loss": 1.8951,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3894005714285715e-05,
+      "loss": 1.7842,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3893737142857143e-05,
+      "loss": 2.2338,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3893468571428573e-05,
+      "loss": 1.9526,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.38932e-05,
+      "loss": 1.9796,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.389293142857143e-05,
+      "loss": 1.8245,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3892662857142857e-05,
+      "loss": 1.6772,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3892394285714287e-05,
+      "loss": 1.7122,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3892125714285714e-05,
+      "loss": 1.9945,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3891857142857144e-05,
+      "loss": 1.7209,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3891588571428572e-05,
+      "loss": 1.6879,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.389132e-05,
+      "loss": 2.2233,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.389105142857143e-05,
+      "loss": 1.9286,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3890782857142858e-05,
+      "loss": 1.6273,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3890514285714286e-05,
+      "loss": 2.2121,
+      "step": 780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3890245714285714e-05,
+      "loss": 1.9878,
+      "step": 781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3889977142857144e-05,
+      "loss": 1.9292,
+      "step": 782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.388970857142857e-05,
+      "loss": 2.0023,
+      "step": 783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.388944e-05,
+      "loss": 1.9118,
+      "step": 784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3889171428571429e-05,
+      "loss": 1.6643,
+      "step": 785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3888902857142859e-05,
+      "loss": 1.555,
+      "step": 786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3888634285714285e-05,
+      "loss": 2.0219,
+      "step": 787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3888365714285715e-05,
+      "loss": 1.8741,
+      "step": 788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3888097142857143e-05,
+      "loss": 2.0954,
+      "step": 789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3887828571428571e-05,
+      "loss": 1.9173,
+      "step": 790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3887560000000001e-05,
+      "loss": 1.9222,
+      "step": 791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.388729142857143e-05,
+      "loss": 1.8772,
+      "step": 792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3887022857142858e-05,
+      "loss": 1.9834,
+      "step": 793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3886754285714286e-05,
+      "loss": 1.9458,
+      "step": 794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3886485714285716e-05,
+      "loss": 1.9193,
+      "step": 795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3886217142857144e-05,
+      "loss": 1.8704,
+      "step": 796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3885948571428572e-05,
+      "loss": 2.0655,
+      "step": 797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.388568e-05,
+      "loss": 1.9158,
+      "step": 798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.388541142857143e-05,
+      "loss": 1.9664,
+      "step": 799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3885142857142857e-05,
+      "loss": 2.0401,
+      "step": 800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3884874285714286e-05,
+      "loss": 1.8958,
+      "step": 801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3884605714285715e-05,
+      "loss": 1.8958,
+      "step": 802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3884337142857143e-05,
+      "loss": 1.605,
+      "step": 803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3884068571428573e-05,
+      "loss": 2.0905,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.38838e-05,
+      "loss": 1.863,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3883531428571429e-05,
+      "loss": 1.9463,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3883262857142857e-05,
+      "loss": 2.0525,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3882994285714287e-05,
+      "loss": 2.0352,
+      "step": 808
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3882725714285715e-05,
+      "loss": 1.8277,
+      "step": 809
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3882457142857143e-05,
+      "loss": 2.2084,
+      "step": 810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3882188571428572e-05,
+      "loss": 2.1031,
+      "step": 811
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3881920000000001e-05,
+      "loss": 2.0614,
+      "step": 812
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3881651428571428e-05,
+      "loss": 1.8388,
+      "step": 813
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3881382857142858e-05,
+      "loss": 1.6096,
+      "step": 814
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3881114285714286e-05,
+      "loss": 2.1765,
+      "step": 815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3880845714285714e-05,
+      "loss": 1.9088,
+      "step": 816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3880577142857144e-05,
+      "loss": 2.1287,
+      "step": 817
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3880308571428572e-05,
+      "loss": 1.983,
+      "step": 818
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3880040000000002e-05,
+      "loss": 1.9613,
+      "step": 819
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3879771428571429e-05,
+      "loss": 1.8406,
+      "step": 820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3879502857142858e-05,
+      "loss": 1.6544,
+      "step": 821
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3879234285714287e-05,
+      "loss": 2.0212,
+      "step": 822
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3878965714285715e-05,
+      "loss": 2.0622,
+      "step": 823
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3878697142857143e-05,
+      "loss": 1.9181,
+      "step": 824
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3878428571428573e-05,
+      "loss": 1.8897,
+      "step": 825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.387816e-05,
+      "loss": 1.832,
+      "step": 826
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3877891428571429e-05,
+      "loss": 1.8731,
+      "step": 827
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3877622857142857e-05,
+      "loss": 1.5693,
+      "step": 828
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3877354285714285e-05,
+      "loss": 2.0005,
+      "step": 829
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3877085714285715e-05,
+      "loss": 1.9372,
+      "step": 830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3876817142857144e-05,
+      "loss": 1.8118,
+      "step": 831
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3876548571428573e-05,
+      "loss": 1.9798,
+      "step": 832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.387628e-05,
+      "loss": 1.9241,
+      "step": 833
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.387601142857143e-05,
+      "loss": 1.8513,
+      "step": 834
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3875742857142858e-05,
+      "loss": 2.1211,
+      "step": 835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3875474285714286e-05,
+      "loss": 1.8736,
+      "step": 836
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3875205714285714e-05,
+      "loss": 2.0758,
+      "step": 837
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3874937142857144e-05,
+      "loss": 2.0094,
+      "step": 838
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.387466857142857e-05,
+      "loss": 1.5176,
+      "step": 839
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.38744e-05,
+      "loss": 2.3543,
+      "step": 840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3874131428571429e-05,
+      "loss": 1.7076,
+      "step": 841
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3873862857142859e-05,
+      "loss": 1.788,
+      "step": 842
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3873594285714287e-05,
+      "loss": 1.9199,
+      "step": 843
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3873325714285715e-05,
+      "loss": 1.7847,
+      "step": 844
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3873057142857145e-05,
+      "loss": 1.787,
+      "step": 845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3872788571428571e-05,
+      "loss": 1.9645,
+      "step": 846
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3872520000000001e-05,
+      "loss": 1.9198,
+      "step": 847
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.387225142857143e-05,
+      "loss": 2.0535,
+      "step": 848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3871982857142857e-05,
+      "loss": 1.8737,
+      "step": 849
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3871714285714286e-05,
+      "loss": 2.1553,
+      "step": 850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3871445714285716e-05,
+      "loss": 2.075,
+      "step": 851
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3871177142857142e-05,
+      "loss": 1.7766,
+      "step": 852
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3870908571428572e-05,
+      "loss": 1.8718,
+      "step": 853
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.387064e-05,
+      "loss": 1.8636,
+      "step": 854
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.387037142857143e-05,
+      "loss": 2.0346,
+      "step": 855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3870102857142856e-05,
+      "loss": 2.0717,
+      "step": 856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3869834285714286e-05,
+      "loss": 1.9735,
+      "step": 857
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3869565714285716e-05,
+      "loss": 1.8442,
+      "step": 858
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3869297142857143e-05,
+      "loss": 1.8466,
+      "step": 859
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3869028571428573e-05,
+      "loss": 1.6191,
+      "step": 860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.386876e-05,
+      "loss": 1.8943,
+      "step": 861
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3868491428571429e-05,
+      "loss": 1.8663,
+      "step": 862
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3868222857142857e-05,
+      "loss": 1.8489,
+      "step": 863
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3867954285714287e-05,
+      "loss": 1.8165,
+      "step": 864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3867685714285715e-05,
+      "loss": 1.7075,
+      "step": 865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3867417142857143e-05,
+      "loss": 1.7622,
+      "step": 866
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3867148571428571e-05,
+      "loss": 1.6848,
+      "step": 867
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3866880000000001e-05,
+      "loss": 1.725,
+      "step": 868
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3866611428571428e-05,
+      "loss": 1.8734,
+      "step": 869
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3866342857142858e-05,
+      "loss": 1.9317,
+      "step": 870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3866074285714288e-05,
+      "loss": 1.6048,
+      "step": 871
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3865805714285714e-05,
+      "loss": 1.9844,
+      "step": 872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3865537142857144e-05,
+      "loss": 1.8493,
+      "step": 873
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3865268571428572e-05,
+      "loss": 2.05,
+      "step": 874
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3865e-05,
+      "loss": 1.6645,
+      "step": 875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3864731428571428e-05,
+      "loss": 2.0904,
+      "step": 876
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3864462857142858e-05,
+      "loss": 2.0622,
+      "step": 877
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3864194285714286e-05,
+      "loss": 1.8044,
+      "step": 878
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3863925714285715e-05,
+      "loss": 1.9895,
+      "step": 879
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3863657142857143e-05,
+      "loss": 1.6886,
+      "step": 880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3863388571428573e-05,
+      "loss": 1.9864,
+      "step": 881
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.386312e-05,
+      "loss": 1.7269,
+      "step": 882
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3862851428571429e-05,
+      "loss": 1.8694,
+      "step": 883
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3862582857142859e-05,
+      "loss": 1.8545,
+      "step": 884
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3862314285714285e-05,
+      "loss": 1.7989,
+      "step": 885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3862045714285715e-05,
+      "loss": 2.0829,
+      "step": 886
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3861777142857143e-05,
+      "loss": 1.6233,
+      "step": 887
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3861508571428573e-05,
+      "loss": 1.9262,
+      "step": 888
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.386124e-05,
+      "loss": 1.9713,
+      "step": 889
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.386097142857143e-05,
+      "loss": 1.9598,
+      "step": 890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3860702857142858e-05,
+      "loss": 2.1689,
+      "step": 891
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3860434285714286e-05,
+      "loss": 2.11,
+      "step": 892
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3860165714285714e-05,
+      "loss": 2.0805,
+      "step": 893
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3859897142857144e-05,
+      "loss": 1.7419,
+      "step": 894
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.385962857142857e-05,
+      "loss": 2.0083,
+      "step": 895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.385936e-05,
+      "loss": 1.8565,
+      "step": 896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.385909142857143e-05,
+      "loss": 2.1203,
+      "step": 897
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3858822857142857e-05,
+      "loss": 1.8916,
+      "step": 898
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3858554285714287e-05,
+      "loss": 1.8381,
+      "step": 899
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3858285714285715e-05,
+      "loss": 2.1263,
+      "step": 900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3858017142857145e-05,
+      "loss": 1.999,
+      "step": 901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3857748571428571e-05,
+      "loss": 2.0834,
+      "step": 902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3857480000000001e-05,
+      "loss": 1.8031,
+      "step": 903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.385721142857143e-05,
+      "loss": 2.0087,
+      "step": 904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3856942857142857e-05,
+      "loss": 1.9577,
+      "step": 905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3856674285714286e-05,
+      "loss": 1.9194,
+      "step": 906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3856405714285715e-05,
+      "loss": 2.0452,
+      "step": 907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3856137142857142e-05,
+      "loss": 1.8298,
+      "step": 908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3855868571428572e-05,
+      "loss": 2.0968,
+      "step": 909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3855600000000002e-05,
+      "loss": 2.063,
+      "step": 910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.385533142857143e-05,
+      "loss": 1.6463,
+      "step": 911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3855062857142858e-05,
+      "loss": 1.7591,
+      "step": 912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3854794285714286e-05,
+      "loss": 1.8694,
+      "step": 913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3854525714285716e-05,
+      "loss": 2.0081,
+      "step": 914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3854257142857143e-05,
+      "loss": 1.7464,
+      "step": 915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3853988571428572e-05,
+      "loss": 1.9161,
+      "step": 916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.385372e-05,
+      "loss": 2.2335,
+      "step": 917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3853451428571429e-05,
+      "loss": 1.6547,
+      "step": 918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3853182857142857e-05,
+      "loss": 1.8366,
+      "step": 919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3852914285714287e-05,
+      "loss": 1.4712,
+      "step": 920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3852645714285713e-05,
+      "loss": 1.7597,
+      "step": 921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3852377142857143e-05,
+      "loss": 1.7585,
+      "step": 922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3852108571428573e-05,
+      "loss": 1.8419,
+      "step": 923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3851840000000001e-05,
+      "loss": 1.7032,
+      "step": 924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.385157142857143e-05,
+      "loss": 1.9065,
+      "step": 925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3851302857142858e-05,
+      "loss": 1.8747,
+      "step": 926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3851034285714287e-05,
+      "loss": 1.7035,
+      "step": 927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3850765714285714e-05,
+      "loss": 1.8398,
+      "step": 928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3850497142857144e-05,
+      "loss": 1.7742,
+      "step": 929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3850228571428572e-05,
+      "loss": 2.2222,
+      "step": 930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.384996e-05,
+      "loss": 2.0037,
+      "step": 931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3849691428571428e-05,
+      "loss": 2.1107,
+      "step": 932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3849422857142858e-05,
+      "loss": 1.8639,
+      "step": 933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3849154285714285e-05,
+      "loss": 2.0648,
+      "step": 934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3848885714285715e-05,
+      "loss": 1.8441,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3848617142857144e-05,
+      "loss": 1.9207,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3848348571428573e-05,
+      "loss": 2.1746,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.384808e-05,
+      "loss": 2.0136,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3847811428571429e-05,
+      "loss": 1.7735,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3847542857142859e-05,
+      "loss": 1.8369,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3847274285714285e-05,
+      "loss": 1.8882,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3847005714285715e-05,
+      "loss": 1.9527,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3846737142857143e-05,
+      "loss": 1.8963,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3846468571428571e-05,
+      "loss": 2.1255,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.38462e-05,
+      "loss": 1.8748,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.384593142857143e-05,
+      "loss": 1.9465,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3845662857142858e-05,
+      "loss": 1.8165,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3845394285714286e-05,
+      "loss": 2.0142,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3845125714285716e-05,
+      "loss": 1.8438,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3844857142857144e-05,
+      "loss": 1.7979,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3844588571428572e-05,
+      "loss": 1.641,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.384432e-05,
+      "loss": 1.8224,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.384405142857143e-05,
+      "loss": 2.0426,
+      "step": 953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3843782857142857e-05,
+      "loss": 1.8228,
+      "step": 954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3843514285714287e-05,
+      "loss": 1.8521,
+      "step": 955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3843245714285715e-05,
+      "loss": 1.9675,
+      "step": 956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3842977142857143e-05,
+      "loss": 1.7368,
+      "step": 957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3842708571428571e-05,
+      "loss": 1.5704,
+      "step": 958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3842440000000001e-05,
+      "loss": 1.909,
+      "step": 959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3842171428571429e-05,
+      "loss": 1.8754,
+      "step": 960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3841902857142857e-05,
+      "loss": 1.8478,
+      "step": 961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3841634285714285e-05,
+      "loss": 2.0136,
+      "step": 962
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3841365714285715e-05,
+      "loss": 2.0803,
+      "step": 963
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3841097142857143e-05,
+      "loss": 1.67,
+      "step": 964
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3840828571428572e-05,
+      "loss": 2.0758,
+      "step": 965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3840560000000002e-05,
+      "loss": 1.775,
+      "step": 966
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3840291428571428e-05,
+      "loss": 1.7642,
+      "step": 967
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3840022857142858e-05,
+      "loss": 2.0297,
+      "step": 968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3839754285714286e-05,
+      "loss": 1.6912,
+      "step": 969
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3839485714285716e-05,
+      "loss": 1.9137,
+      "step": 970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3839217142857142e-05,
+      "loss": 1.7814,
+      "step": 971
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3838948571428572e-05,
+      "loss": 1.6611,
+      "step": 972
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.383868e-05,
+      "loss": 1.9094,
+      "step": 973
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3838411428571429e-05,
+      "loss": 1.7419,
+      "step": 974
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3838142857142857e-05,
+      "loss": 1.7932,
+      "step": 975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3837874285714287e-05,
+      "loss": 2.1068,
+      "step": 976
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3837605714285715e-05,
+      "loss": 1.8853,
+      "step": 977
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3837337142857143e-05,
+      "loss": 1.8545,
+      "step": 978
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3837068571428573e-05,
+      "loss": 1.8344,
+      "step": 979
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.38368e-05,
+      "loss": 1.7764,
+      "step": 980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.383653142857143e-05,
+      "loss": 1.7142,
+      "step": 981
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3836262857142857e-05,
+      "loss": 1.9975,
+      "step": 982
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3835994285714287e-05,
+      "loss": 1.7012,
+      "step": 983
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3835725714285714e-05,
+      "loss": 2.0036,
+      "step": 984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3835457142857144e-05,
+      "loss": 1.8932,
+      "step": 985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3835188571428572e-05,
+      "loss": 1.7613,
+      "step": 986
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.383492e-05,
+      "loss": 1.8193,
+      "step": 987
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3834651428571428e-05,
+      "loss": 2.0312,
+      "step": 988
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3834382857142858e-05,
+      "loss": 1.751,
+      "step": 989
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3834114285714286e-05,
+      "loss": 1.7399,
+      "step": 990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3833845714285714e-05,
+      "loss": 1.9774,
+      "step": 991
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3833577142857144e-05,
+      "loss": 1.868,
+      "step": 992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3833308571428572e-05,
+      "loss": 1.7673,
+      "step": 993
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.383304e-05,
+      "loss": 1.9701,
+      "step": 994
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3832771428571429e-05,
+      "loss": 2.0085,
+      "step": 995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3832502857142859e-05,
+      "loss": 1.6656,
+      "step": 996
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3832234285714285e-05,
+      "loss": 1.9488,
+      "step": 997
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3831965714285715e-05,
+      "loss": 1.6869,
+      "step": 998
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3831697142857143e-05,
+      "loss": 1.7061,
+      "step": 999
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3831428571428571e-05,
+      "loss": 1.8224,
+      "step": 1000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.383116e-05,
+      "loss": 1.8077,
+      "step": 1001
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.383089142857143e-05,
+      "loss": 1.8805,
+      "step": 1002
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3830622857142858e-05,
+      "loss": 1.8493,
+      "step": 1003
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3830354285714286e-05,
+      "loss": 1.9838,
+      "step": 1004
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3830085714285716e-05,
+      "loss": 1.7998,
+      "step": 1005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3829817142857144e-05,
+      "loss": 1.9014,
+      "step": 1006
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3829548571428572e-05,
+      "loss": 2.1141,
+      "step": 1007
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.382928e-05,
+      "loss": 2.0962,
+      "step": 1008
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.382901142857143e-05,
+      "loss": 1.9159,
+      "step": 1009
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3828742857142857e-05,
+      "loss": 1.7947,
+      "step": 1010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3828474285714286e-05,
+      "loss": 1.9726,
+      "step": 1011
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3828205714285715e-05,
+      "loss": 1.8745,
+      "step": 1012
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3827937142857143e-05,
+      "loss": 1.854,
+      "step": 1013
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3827668571428571e-05,
+      "loss": 1.6724,
+      "step": 1014
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.38274e-05,
+      "loss": 1.6966,
+      "step": 1015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.382713142857143e-05,
+      "loss": 2.3105,
+      "step": 1016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3826862857142857e-05,
+      "loss": 1.4947,
+      "step": 1017
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3826594285714287e-05,
+      "loss": 1.7281,
+      "step": 1018
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3826325714285715e-05,
+      "loss": 1.6852,
+      "step": 1019
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3826057142857143e-05,
+      "loss": 2.0752,
+      "step": 1020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3825788571428572e-05,
+      "loss": 1.8735,
+      "step": 1021
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3825520000000001e-05,
+      "loss": 1.7395,
+      "step": 1022
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3825251428571428e-05,
+      "loss": 1.8953,
+      "step": 1023
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3824982857142858e-05,
+      "loss": 2.049,
+      "step": 1024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3824714285714286e-05,
+      "loss": 1.931,
+      "step": 1025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3824445714285714e-05,
+      "loss": 1.951,
+      "step": 1026
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3824177142857142e-05,
+      "loss": 2.0206,
+      "step": 1027
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3823908571428572e-05,
+      "loss": 1.726,
+      "step": 1028
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3823640000000002e-05,
+      "loss": 1.7752,
+      "step": 1029
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3823371428571429e-05,
+      "loss": 1.8353,
+      "step": 1030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3823102857142858e-05,
+      "loss": 1.9413,
+      "step": 1031
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3822834285714287e-05,
+      "loss": 1.5369,
+      "step": 1032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3822565714285715e-05,
+      "loss": 1.914,
+      "step": 1033
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3822297142857143e-05,
+      "loss": 1.7954,
+      "step": 1034
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3822028571428573e-05,
+      "loss": 1.7882,
+      "step": 1035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.382176e-05,
+      "loss": 2.0805,
+      "step": 1036
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3821491428571429e-05,
+      "loss": 1.7503,
+      "step": 1037
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3821222857142857e-05,
+      "loss": 1.7966,
+      "step": 1038
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3820954285714287e-05,
+      "loss": 1.5869,
+      "step": 1039
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3820685714285714e-05,
+      "loss": 2.0778,
+      "step": 1040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3820417142857144e-05,
+      "loss": 1.7252,
+      "step": 1041
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3820148571428573e-05,
+      "loss": 1.8926,
+      "step": 1042
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.381988e-05,
+      "loss": 2.0161,
+      "step": 1043
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.381961142857143e-05,
+      "loss": 2.0138,
+      "step": 1044
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3819342857142858e-05,
+      "loss": 2.0324,
+      "step": 1045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3819074285714286e-05,
+      "loss": 1.9668,
+      "step": 1046
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3818805714285714e-05,
+      "loss": 2.1228,
+      "step": 1047
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3818537142857144e-05,
+      "loss": 1.8028,
+      "step": 1048
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.381826857142857e-05,
+      "loss": 1.3291,
+      "step": 1049
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3818e-05,
+      "loss": 1.818,
+      "step": 1050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3817731428571429e-05,
+      "loss": 2.0664,
+      "step": 1051
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3817462857142859e-05,
+      "loss": 1.8046,
+      "step": 1052
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3817194285714285e-05,
+      "loss": 1.8571,
+      "step": 1053
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3816925714285715e-05,
+      "loss": 1.7383,
+      "step": 1054
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3816657142857145e-05,
+      "loss": 1.6691,
+      "step": 1055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3816388571428571e-05,
+      "loss": 1.8232,
+      "step": 1056
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3816120000000001e-05,
+      "loss": 1.8695,
+      "step": 1057
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.381585142857143e-05,
+      "loss": 1.782,
+      "step": 1058
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3815582857142857e-05,
+      "loss": 1.6073,
+      "step": 1059
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3815314285714286e-05,
+      "loss": 2.0035,
+      "step": 1060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3815045714285716e-05,
+      "loss": 1.8561,
+      "step": 1061
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3814777142857144e-05,
+      "loss": 2.0537,
+      "step": 1062
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3814508571428572e-05,
+      "loss": 1.9125,
+      "step": 1063
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.381424e-05,
+      "loss": 1.9935,
+      "step": 1064
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.381397142857143e-05,
+      "loss": 1.8367,
+      "step": 1065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3813702857142856e-05,
+      "loss": 1.9172,
+      "step": 1066
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3813434285714286e-05,
+      "loss": 1.6592,
+      "step": 1067
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3813165714285714e-05,
+      "loss": 1.93,
+      "step": 1068
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3812897142857143e-05,
+      "loss": 2.1151,
+      "step": 1069
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3812628571428573e-05,
+      "loss": 1.7521,
+      "step": 1070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.381236e-05,
+      "loss": 1.6689,
+      "step": 1071
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3812091428571429e-05,
+      "loss": 1.9493,
+      "step": 1072
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3811822857142857e-05,
+      "loss": 1.941,
+      "step": 1073
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3811554285714287e-05,
+      "loss": 1.766,
+      "step": 1074
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3811285714285715e-05,
+      "loss": 1.567,
+      "step": 1075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3811017142857143e-05,
+      "loss": 1.7886,
+      "step": 1076
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3810748571428571e-05,
+      "loss": 1.8427,
+      "step": 1077
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3810480000000001e-05,
+      "loss": 1.6823,
+      "step": 1078
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3810211428571428e-05,
+      "loss": 1.839,
+      "step": 1079
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3809942857142858e-05,
+      "loss": 1.894,
+      "step": 1080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3809674285714286e-05,
+      "loss": 1.7494,
+      "step": 1081
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3809405714285714e-05,
+      "loss": 1.8085,
+      "step": 1082
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3809137142857144e-05,
+      "loss": 1.6975,
+      "step": 1083
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3808868571428572e-05,
+      "loss": 1.8097,
+      "step": 1084
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.38086e-05,
+      "loss": 1.5431,
+      "step": 1085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3808331428571428e-05,
+      "loss": 2.2277,
+      "step": 1086
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3808062857142858e-05,
+      "loss": 1.5046,
+      "step": 1087
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3807794285714286e-05,
+      "loss": 1.8539,
+      "step": 1088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3807525714285715e-05,
+      "loss": 1.7899,
+      "step": 1089
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3807257142857143e-05,
+      "loss": 1.7364,
+      "step": 1090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3806988571428573e-05,
+      "loss": 1.8069,
+      "step": 1091
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.380672e-05,
+      "loss": 1.9662,
+      "step": 1092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3806451428571429e-05,
+      "loss": 1.9083,
+      "step": 1093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3806182857142857e-05,
+      "loss": 2.0446,
+      "step": 1094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3805914285714285e-05,
+      "loss": 2.0306,
+      "step": 1095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3805645714285715e-05,
+      "loss": 1.8584,
+      "step": 1096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3805377142857143e-05,
+      "loss": 1.7681,
+      "step": 1097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3805108571428573e-05,
+      "loss": 2.1962,
+      "step": 1098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.380484e-05,
+      "loss": 1.7053,
+      "step": 1099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.380457142857143e-05,
+      "loss": 1.6763,
+      "step": 1100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3804302857142858e-05,
+      "loss": 1.6934,
+      "step": 1101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3804034285714286e-05,
+      "loss": 1.9147,
+      "step": 1102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3803765714285714e-05,
+      "loss": 1.7846,
+      "step": 1103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3803497142857144e-05,
+      "loss": 1.798,
+      "step": 1104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.380322857142857e-05,
+      "loss": 2.0307,
+      "step": 1105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.380296e-05,
+      "loss": 1.9005,
+      "step": 1106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3802691428571429e-05,
+      "loss": 2.0359,
+      "step": 1107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3802422857142857e-05,
+      "loss": 1.7126,
+      "step": 1108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3802154285714287e-05,
+      "loss": 1.932,
+      "step": 1109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3801885714285715e-05,
+      "loss": 1.8549,
+      "step": 1110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3801617142857145e-05,
+      "loss": 1.4123,
+      "step": 1111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3801348571428571e-05,
+      "loss": 1.9298,
+      "step": 1112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3801080000000001e-05,
+      "loss": 1.8727,
+      "step": 1113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.380081142857143e-05,
+      "loss": 1.9352,
+      "step": 1114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3800542857142857e-05,
+      "loss": 1.6878,
+      "step": 1115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3800274285714286e-05,
+      "loss": 1.857,
+      "step": 1116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3800005714285715e-05,
+      "loss": 1.7999,
+      "step": 1117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3799737142857142e-05,
+      "loss": 2.0103,
+      "step": 1118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3799468571428572e-05,
+      "loss": 1.5861,
+      "step": 1119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.37992e-05,
+      "loss": 1.9793,
+      "step": 1120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.379893142857143e-05,
+      "loss": 1.8524,
+      "step": 1121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3798662857142858e-05,
+      "loss": 1.5875,
+      "step": 1122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3798394285714286e-05,
+      "loss": 1.9046,
+      "step": 1123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3798125714285716e-05,
+      "loss": 1.7087,
+      "step": 1124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3797857142857143e-05,
+      "loss": 1.8093,
+      "step": 1125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3797588571428572e-05,
+      "loss": 2.1192,
+      "step": 1126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.379732e-05,
+      "loss": 1.8204,
+      "step": 1127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3797051428571429e-05,
+      "loss": 1.9127,
+      "step": 1128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3796782857142857e-05,
+      "loss": 1.865,
+      "step": 1129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3796514285714287e-05,
+      "loss": 1.9235,
+      "step": 1130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3796245714285713e-05,
+      "loss": 2.0097,
+      "step": 1131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3795977142857143e-05,
+      "loss": 2.1124,
+      "step": 1132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3795708571428571e-05,
+      "loss": 1.8323,
+      "step": 1133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3795440000000001e-05,
+      "loss": 1.9224,
+      "step": 1134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.379517142857143e-05,
+      "loss": 1.9309,
+      "step": 1135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3794902857142858e-05,
+      "loss": 1.822,
+      "step": 1136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.3794634285714287e-05,
+      "loss": 1.9083,
+      "step": 1137
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3794365714285714e-05,
+      "loss": 1.8793,
+      "step": 1138
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3794097142857144e-05,
+      "loss": 1.7931,
+      "step": 1139
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3793828571428572e-05,
+      "loss": 1.7575,
+      "step": 1140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.379356e-05,
+      "loss": 1.9645,
+      "step": 1141
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3793291428571428e-05,
+      "loss": 1.9775,
+      "step": 1142
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3793022857142858e-05,
+      "loss": 1.909,
+      "step": 1143
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3792754285714286e-05,
+      "loss": 2.0313,
+      "step": 1144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3792485714285715e-05,
+      "loss": 2.1235,
+      "step": 1145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3792217142857143e-05,
+      "loss": 1.6011,
+      "step": 1146
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3791948571428573e-05,
+      "loss": 1.811,
+      "step": 1147
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.379168e-05,
+      "loss": 1.6864,
+      "step": 1148
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3791411428571429e-05,
+      "loss": 1.6096,
+      "step": 1149
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3791142857142859e-05,
+      "loss": 1.856,
+      "step": 1150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3790874285714285e-05,
+      "loss": 1.9935,
+      "step": 1151
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3790605714285715e-05,
+      "loss": 1.792,
+      "step": 1152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3790337142857143e-05,
+      "loss": 1.8164,
+      "step": 1153
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3790068571428571e-05,
+      "loss": 1.9304,
+      "step": 1154
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.37898e-05,
+      "loss": 2.1108,
+      "step": 1155
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.378953142857143e-05,
+      "loss": 1.9556,
+      "step": 1156
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3789262857142858e-05,
+      "loss": 1.8026,
+      "step": 1157
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3788994285714286e-05,
+      "loss": 1.8245,
+      "step": 1158
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3788725714285714e-05,
+      "loss": 1.9801,
+      "step": 1159
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3788457142857144e-05,
+      "loss": 1.9222,
+      "step": 1160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3788188571428572e-05,
+      "loss": 2.0323,
+      "step": 1161
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.378792e-05,
+      "loss": 1.6416,
+      "step": 1162
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.378765142857143e-05,
+      "loss": 1.7883,
+      "step": 1163
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3787382857142857e-05,
+      "loss": 1.9765,
+      "step": 1164
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3787114285714287e-05,
+      "loss": 1.8592,
+      "step": 1165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3786845714285715e-05,
+      "loss": 1.7061,
+      "step": 1166
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3786577142857145e-05,
+      "loss": 1.8298,
+      "step": 1167
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3786308571428571e-05,
+      "loss": 1.8962,
+      "step": 1168
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3786040000000001e-05,
+      "loss": 1.9575,
+      "step": 1169
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3785771428571429e-05,
+      "loss": 1.9457,
+      "step": 1170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3785502857142857e-05,
+      "loss": 1.6877,
+      "step": 1171
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3785234285714285e-05,
+      "loss": 1.9386,
+      "step": 1172
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3784965714285715e-05,
+      "loss": 1.8101,
+      "step": 1173
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3784697142857143e-05,
+      "loss": 1.8382,
+      "step": 1174
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3784428571428572e-05,
+      "loss": 1.6832,
+      "step": 1175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3784160000000002e-05,
+      "loss": 1.5198,
+      "step": 1176
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3783891428571428e-05,
+      "loss": 1.9396,
+      "step": 1177
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3783622857142858e-05,
+      "loss": 1.8228,
+      "step": 1178
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3783354285714286e-05,
+      "loss": 1.8262,
+      "step": 1179
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3783085714285716e-05,
+      "loss": 1.7204,
+      "step": 1180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3782817142857142e-05,
+      "loss": 2.0577,
+      "step": 1181
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3782548571428572e-05,
+      "loss": 1.8665,
+      "step": 1182
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.378228e-05,
+      "loss": 1.8631,
+      "step": 1183
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3782011428571429e-05,
+      "loss": 1.7822,
+      "step": 1184
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3781742857142857e-05,
+      "loss": 1.9977,
+      "step": 1185
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3781474285714287e-05,
+      "loss": 1.7571,
+      "step": 1186
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3781205714285715e-05,
+      "loss": 1.9345,
+      "step": 1187
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3780937142857143e-05,
+      "loss": 1.7735,
+      "step": 1188
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3780668571428573e-05,
+      "loss": 1.8313,
+      "step": 1189
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3780400000000001e-05,
+      "loss": 1.9704,
+      "step": 1190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.378013142857143e-05,
+      "loss": 1.7751,
+      "step": 1191
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3779862857142857e-05,
+      "loss": 1.6898,
+      "step": 1192
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3779594285714287e-05,
+      "loss": 1.8804,
+      "step": 1193
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3779325714285714e-05,
+      "loss": 1.7111,
+      "step": 1194
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3779057142857144e-05,
+      "loss": 2.0523,
+      "step": 1195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3778788571428572e-05,
+      "loss": 1.8596,
+      "step": 1196
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.377852e-05,
+      "loss": 1.5995,
+      "step": 1197
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3778251428571428e-05,
+      "loss": 1.9227,
+      "step": 1198
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3777982857142858e-05,
+      "loss": 1.858,
+      "step": 1199
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3777714285714286e-05,
+      "loss": 1.7013,
+      "step": 1200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3777445714285714e-05,
+      "loss": 2.0002,
+      "step": 1201
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3777177142857144e-05,
+      "loss": 1.9082,
+      "step": 1202
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3776908571428572e-05,
+      "loss": 1.812,
+      "step": 1203
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.377664e-05,
+      "loss": 1.6797,
+      "step": 1204
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3776371428571429e-05,
+      "loss": 2.0985,
+      "step": 1205
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3776102857142859e-05,
+      "loss": 2.0249,
+      "step": 1206
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3775834285714285e-05,
+      "loss": 2.1728,
+      "step": 1207
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3775565714285715e-05,
+      "loss": 1.7825,
+      "step": 1208
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3775297142857143e-05,
+      "loss": 1.6637,
+      "step": 1209
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3775028571428571e-05,
+      "loss": 1.6625,
+      "step": 1210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.377476e-05,
+      "loss": 1.8641,
+      "step": 1211
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.377449142857143e-05,
+      "loss": 1.8921,
+      "step": 1212
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3774222857142858e-05,
+      "loss": 1.8876,
+      "step": 1213
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3773954285714286e-05,
+      "loss": 1.8423,
+      "step": 1214
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3773685714285716e-05,
+      "loss": 1.6718,
+      "step": 1215
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3773417142857144e-05,
+      "loss": 1.8559,
+      "step": 1216
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3773148571428572e-05,
+      "loss": 1.7695,
+      "step": 1217
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.377288e-05,
+      "loss": 1.8202,
+      "step": 1218
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.377261142857143e-05,
+      "loss": 1.9207,
+      "step": 1219
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3772342857142857e-05,
+      "loss": 1.8687,
+      "step": 1220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3772074285714286e-05,
+      "loss": 1.6403,
+      "step": 1221
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3771805714285715e-05,
+      "loss": 1.7338,
+      "step": 1222
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3771537142857143e-05,
+      "loss": 2.0496,
+      "step": 1223
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3771268571428571e-05,
+      "loss": 1.7609,
+      "step": 1224
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3771e-05,
+      "loss": 2.0805,
+      "step": 1225
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3770731428571429e-05,
+      "loss": 2.0277,
+      "step": 1226
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3770462857142857e-05,
+      "loss": 1.5309,
+      "step": 1227
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3770194285714287e-05,
+      "loss": 1.706,
+      "step": 1228
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3769925714285715e-05,
+      "loss": 1.9132,
+      "step": 1229
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3769657142857143e-05,
+      "loss": 2.1275,
+      "step": 1230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3769388571428572e-05,
+      "loss": 1.6862,
+      "step": 1231
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3769120000000001e-05,
+      "loss": 1.8802,
+      "step": 1232
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3768851428571428e-05,
+      "loss": 2.0264,
+      "step": 1233
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3768582857142858e-05,
+      "loss": 1.7446,
+      "step": 1234
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3768314285714286e-05,
+      "loss": 1.98,
+      "step": 1235
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3768045714285716e-05,
+      "loss": 1.7541,
+      "step": 1236
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3767777142857142e-05,
+      "loss": 1.5116,
+      "step": 1237
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3767508571428572e-05,
+      "loss": 1.7616,
+      "step": 1238
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.376724e-05,
+      "loss": 2.1069,
+      "step": 1239
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3766971428571429e-05,
+      "loss": 1.883,
+      "step": 1240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3766702857142858e-05,
+      "loss": 2.0043,
+      "step": 1241
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3766434285714287e-05,
+      "loss": 1.753,
+      "step": 1242
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3766165714285715e-05,
+      "loss": 1.6529,
+      "step": 1243
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3765897142857143e-05,
+      "loss": 1.6246,
+      "step": 1244
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3765628571428573e-05,
+      "loss": 1.8316,
+      "step": 1245
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.376536e-05,
+      "loss": 1.99,
+      "step": 1246
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3765091428571429e-05,
+      "loss": 1.8827,
+      "step": 1247
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3764822857142857e-05,
+      "loss": 1.7054,
+      "step": 1248
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3764554285714287e-05,
+      "loss": 2.0461,
+      "step": 1249
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3764285714285714e-05,
+      "loss": 1.6813,
+      "step": 1250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3764017142857144e-05,
+      "loss": 1.7659,
+      "step": 1251
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3763748571428572e-05,
+      "loss": 1.8794,
+      "step": 1252
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.376348e-05,
+      "loss": 2.117,
+      "step": 1253
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.376321142857143e-05,
+      "loss": 1.6551,
+      "step": 1254
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3762942857142858e-05,
+      "loss": 1.8859,
+      "step": 1255
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3762674285714286e-05,
+      "loss": 1.6937,
+      "step": 1256
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3762405714285714e-05,
+      "loss": 1.9766,
+      "step": 1257
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3762137142857144e-05,
+      "loss": 1.8706,
+      "step": 1258
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.376186857142857e-05,
+      "loss": 1.7404,
+      "step": 1259
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.37616e-05,
+      "loss": 1.7318,
+      "step": 1260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3761331428571429e-05,
+      "loss": 1.7148,
+      "step": 1261
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3761062857142859e-05,
+      "loss": 1.5877,
+      "step": 1262
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3760794285714285e-05,
+      "loss": 1.6143,
+      "step": 1263
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3760525714285715e-05,
+      "loss": 1.9818,
+      "step": 1264
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3760257142857143e-05,
+      "loss": 1.9468,
+      "step": 1265
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3759988571428571e-05,
+      "loss": 1.705,
+      "step": 1266
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3759720000000001e-05,
+      "loss": 1.99,
+      "step": 1267
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.375945142857143e-05,
+      "loss": 1.8092,
+      "step": 1268
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3759182857142857e-05,
+      "loss": 1.8728,
+      "step": 1269
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3758914285714286e-05,
+      "loss": 1.9475,
+      "step": 1270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3758645714285716e-05,
+      "loss": 1.7687,
+      "step": 1271
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3758377142857144e-05,
+      "loss": 1.83,
+      "step": 1272
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3758108571428572e-05,
+      "loss": 1.6303,
+      "step": 1273
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.375784e-05,
+      "loss": 1.7257,
+      "step": 1274
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.375757142857143e-05,
+      "loss": 1.6975,
+      "step": 1275
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3757302857142856e-05,
+      "loss": 1.5962,
+      "step": 1276
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3757034285714286e-05,
+      "loss": 1.986,
+      "step": 1277
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3756765714285714e-05,
+      "loss": 1.7771,
+      "step": 1278
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3756497142857143e-05,
+      "loss": 1.9963,
+      "step": 1279
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3756228571428573e-05,
+      "loss": 1.7435,
+      "step": 1280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.375596e-05,
+      "loss": 1.6746,
+      "step": 1281
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3755691428571429e-05,
+      "loss": 1.7048,
+      "step": 1282
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3755422857142857e-05,
+      "loss": 1.6751,
+      "step": 1283
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3755154285714287e-05,
+      "loss": 1.8427,
+      "step": 1284
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3754885714285715e-05,
+      "loss": 1.6252,
+      "step": 1285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3754617142857143e-05,
+      "loss": 1.8523,
+      "step": 1286
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3754348571428571e-05,
+      "loss": 1.7309,
+      "step": 1287
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3754080000000001e-05,
+      "loss": 1.9167,
+      "step": 1288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3753811428571428e-05,
+      "loss": 1.9956,
+      "step": 1289
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3753542857142858e-05,
+      "loss": 1.7987,
+      "step": 1290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3753274285714286e-05,
+      "loss": 1.5062,
+      "step": 1291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3753005714285714e-05,
+      "loss": 1.6612,
+      "step": 1292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3752737142857144e-05,
+      "loss": 1.831,
+      "step": 1293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3752468571428572e-05,
+      "loss": 1.6575,
+      "step": 1294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3752200000000002e-05,
+      "loss": 1.7274,
+      "step": 1295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3751931428571428e-05,
+      "loss": 1.6289,
+      "step": 1296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3751662857142858e-05,
+      "loss": 1.6894,
+      "step": 1297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3751394285714286e-05,
+      "loss": 1.7426,
+      "step": 1298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3751125714285715e-05,
+      "loss": 1.804,
+      "step": 1299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3750857142857143e-05,
+      "loss": 1.6948,
+      "step": 1300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3750588571428573e-05,
+      "loss": 1.9546,
+      "step": 1301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.375032e-05,
+      "loss": 1.7923,
+      "step": 1302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3750051428571429e-05,
+      "loss": 1.8818,
+      "step": 1303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3749782857142857e-05,
+      "loss": 1.6819,
+      "step": 1304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3749514285714285e-05,
+      "loss": 1.949,
+      "step": 1305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3749245714285715e-05,
+      "loss": 1.472,
+      "step": 1306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3748977142857143e-05,
+      "loss": 1.6227,
+      "step": 1307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3748708571428573e-05,
+      "loss": 1.8807,
+      "step": 1308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.374844e-05,
+      "loss": 1.7583,
+      "step": 1309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.374817142857143e-05,
+      "loss": 1.8692,
+      "step": 1310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3747902857142858e-05,
+      "loss": 1.6156,
+      "step": 1311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3747634285714286e-05,
+      "loss": 1.4971,
+      "step": 1312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3747365714285714e-05,
+      "loss": 1.9164,
+      "step": 1313
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3747097142857144e-05,
+      "loss": 1.8725,
+      "step": 1314
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.374682857142857e-05,
+      "loss": 1.6939,
+      "step": 1315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.374656e-05,
+      "loss": 2.0157,
+      "step": 1316
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3746291428571429e-05,
+      "loss": 1.6325,
+      "step": 1317
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3746022857142858e-05,
+      "loss": 1.6377,
+      "step": 1318
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3745754285714287e-05,
+      "loss": 1.8041,
+      "step": 1319
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3745485714285715e-05,
+      "loss": 1.803,
+      "step": 1320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3745217142857145e-05,
+      "loss": 1.8643,
+      "step": 1321
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3744948571428571e-05,
+      "loss": 1.7831,
+      "step": 1322
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3744680000000001e-05,
+      "loss": 1.9656,
+      "step": 1323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.374441142857143e-05,
+      "loss": 1.976,
+      "step": 1324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3744142857142857e-05,
+      "loss": 1.9453,
+      "step": 1325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3743874285714286e-05,
+      "loss": 1.7973,
+      "step": 1326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3743605714285715e-05,
+      "loss": 1.9361,
+      "step": 1327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3743337142857142e-05,
+      "loss": 1.7485,
+      "step": 1328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3743068571428572e-05,
+      "loss": 1.7835,
+      "step": 1329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.37428e-05,
+      "loss": 1.7151,
+      "step": 1330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.374253142857143e-05,
+      "loss": 1.6416,
+      "step": 1331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3742262857142858e-05,
+      "loss": 1.8416,
+      "step": 1332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3741994285714286e-05,
+      "loss": 1.9134,
+      "step": 1333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3741725714285716e-05,
+      "loss": 1.8486,
+      "step": 1334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3741457142857143e-05,
+      "loss": 1.9457,
+      "step": 1335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3741188571428572e-05,
+      "loss": 1.7294,
+      "step": 1336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.374092e-05,
+      "loss": 1.6833,
+      "step": 1337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3740651428571429e-05,
+      "loss": 1.8719,
+      "step": 1338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3740382857142857e-05,
+      "loss": 1.9124,
+      "step": 1339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3740114285714287e-05,
+      "loss": 1.7526,
+      "step": 1340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3739845714285715e-05,
+      "loss": 1.934,
+      "step": 1341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3739577142857143e-05,
+      "loss": 1.8501,
+      "step": 1342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3739308571428571e-05,
+      "loss": 1.8822,
+      "step": 1343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3739040000000001e-05,
+      "loss": 1.8439,
+      "step": 1344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.373877142857143e-05,
+      "loss": 1.5413,
+      "step": 1345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3738502857142858e-05,
+      "loss": 1.6849,
+      "step": 1346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3738234285714287e-05,
+      "loss": 1.6344,
+      "step": 1347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3737965714285714e-05,
+      "loss": 2.1266,
+      "step": 1348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3737697142857144e-05,
+      "loss": 1.9443,
+      "step": 1349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3737428571428572e-05,
+      "loss": 1.4814,
+      "step": 1350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.373716e-05,
+      "loss": 1.6049,
+      "step": 1351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3736891428571428e-05,
+      "loss": 1.689,
+      "step": 1352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3736622857142858e-05,
+      "loss": 1.9349,
+      "step": 1353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3736354285714286e-05,
+      "loss": 2.0471,
+      "step": 1354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3736085714285715e-05,
+      "loss": 1.4892,
+      "step": 1355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3735817142857143e-05,
+      "loss": 1.6634,
+      "step": 1356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3735548571428573e-05,
+      "loss": 1.8805,
+      "step": 1357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.373528e-05,
+      "loss": 1.6658,
+      "step": 1358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3735011428571429e-05,
+      "loss": 1.9886,
+      "step": 1359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3734742857142859e-05,
+      "loss": 2.0366,
+      "step": 1360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3734474285714285e-05,
+      "loss": 1.8658,
+      "step": 1361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3734205714285715e-05,
+      "loss": 1.7477,
+      "step": 1362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3733937142857143e-05,
+      "loss": 1.8581,
+      "step": 1363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3733668571428573e-05,
+      "loss": 1.7192,
+      "step": 1364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.37334e-05,
+      "loss": 1.8544,
+      "step": 1365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.373313142857143e-05,
+      "loss": 1.7893,
+      "step": 1366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3732862857142858e-05,
+      "loss": 1.985,
+      "step": 1367
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3732594285714286e-05,
+      "loss": 1.6349,
+      "step": 1368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3732325714285714e-05,
+      "loss": 1.9766,
+      "step": 1369
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3732057142857144e-05,
+      "loss": 1.7371,
+      "step": 1370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3731788571428572e-05,
+      "loss": 1.7382,
+      "step": 1371
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.373152e-05,
+      "loss": 1.6328,
+      "step": 1372
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.373125142857143e-05,
+      "loss": 1.4883,
+      "step": 1373
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3730982857142857e-05,
+      "loss": 1.8057,
+      "step": 1374
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3730714285714287e-05,
+      "loss": 2.1826,
+      "step": 1375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3730445714285715e-05,
+      "loss": 1.7747,
+      "step": 1376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3730177142857145e-05,
+      "loss": 1.7193,
+      "step": 1377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3729908571428571e-05,
+      "loss": 1.8808,
+      "step": 1378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3729640000000001e-05,
+      "loss": 1.7424,
+      "step": 1379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3729371428571429e-05,
+      "loss": 1.9128,
+      "step": 1380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3729102857142857e-05,
+      "loss": 1.8905,
+      "step": 1381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3728834285714285e-05,
+      "loss": 1.8263,
+      "step": 1382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3728565714285715e-05,
+      "loss": 2.0555,
+      "step": 1383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3728297142857143e-05,
+      "loss": 1.8006,
+      "step": 1384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3728028571428572e-05,
+      "loss": 1.6676,
+      "step": 1385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3727760000000002e-05,
+      "loss": 1.8639,
+      "step": 1386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.372749142857143e-05,
+      "loss": 1.9433,
+      "step": 1387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3727222857142858e-05,
+      "loss": 1.8881,
+      "step": 1388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3726954285714286e-05,
+      "loss": 1.8204,
+      "step": 1389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3726685714285716e-05,
+      "loss": 1.5288,
+      "step": 1390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3726417142857142e-05,
+      "loss": 1.7843,
+      "step": 1391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3726148571428572e-05,
+      "loss": 1.597,
+      "step": 1392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.372588e-05,
+      "loss": 1.6772,
+      "step": 1393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3725611428571429e-05,
+      "loss": 1.8843,
+      "step": 1394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3725342857142857e-05,
+      "loss": 1.5831,
+      "step": 1395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3725074285714287e-05,
+      "loss": 1.8802,
+      "step": 1396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3724805714285715e-05,
+      "loss": 1.7862,
+      "step": 1397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3724537142857143e-05,
+      "loss": 1.7607,
+      "step": 1398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3724268571428573e-05,
+      "loss": 1.7817,
+      "step": 1399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3724000000000001e-05,
+      "loss": 1.6997,
+      "step": 1400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.372373142857143e-05,
+      "loss": 1.723,
+      "step": 1401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3723462857142857e-05,
+      "loss": 1.4868,
+      "step": 1402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3723194285714287e-05,
+      "loss": 1.7787,
+      "step": 1403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3722925714285714e-05,
+      "loss": 1.7146,
+      "step": 1404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3722657142857144e-05,
+      "loss": 1.8298,
+      "step": 1405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3722388571428572e-05,
+      "loss": 1.9065,
+      "step": 1406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.372212e-05,
+      "loss": 1.8731,
+      "step": 1407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3721851428571428e-05,
+      "loss": 1.9593,
+      "step": 1408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3721582857142858e-05,
+      "loss": 1.9001,
+      "step": 1409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3721314285714286e-05,
+      "loss": 1.5443,
+      "step": 1410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3721045714285714e-05,
+      "loss": 1.6975,
+      "step": 1411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3720777142857144e-05,
+      "loss": 1.614,
+      "step": 1412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3720508571428572e-05,
+      "loss": 2.119,
+      "step": 1413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.372024e-05,
+      "loss": 1.7364,
+      "step": 1414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3719971428571429e-05,
+      "loss": 1.7336,
+      "step": 1415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3719702857142859e-05,
+      "loss": 1.5885,
+      "step": 1416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3719434285714285e-05,
+      "loss": 2.0351,
+      "step": 1417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3719165714285715e-05,
+      "loss": 1.6965,
+      "step": 1418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3718897142857143e-05,
+      "loss": 1.5888,
+      "step": 1419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3718628571428571e-05,
+      "loss": 2.1546,
+      "step": 1420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.371836e-05,
+      "loss": 1.5824,
+      "step": 1421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.371809142857143e-05,
+      "loss": 1.8631,
+      "step": 1422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3717822857142858e-05,
+      "loss": 1.903,
+      "step": 1423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3717554285714286e-05,
+      "loss": 1.8294,
+      "step": 1424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3717285714285716e-05,
+      "loss": 2.0305,
+      "step": 1425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3717017142857144e-05,
+      "loss": 1.5572,
+      "step": 1426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3716748571428572e-05,
+      "loss": 1.7124,
+      "step": 1427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.371648e-05,
+      "loss": 1.5446,
+      "step": 1428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.371621142857143e-05,
+      "loss": 1.6446,
+      "step": 1429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3715942857142857e-05,
+      "loss": 1.6298,
+      "step": 1430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3715674285714286e-05,
+      "loss": 1.9116,
+      "step": 1431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3715405714285715e-05,
+      "loss": 1.5899,
+      "step": 1432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3715137142857143e-05,
+      "loss": 1.7673,
+      "step": 1433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3714868571428571e-05,
+      "loss": 1.7035,
+      "step": 1434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.37146e-05,
+      "loss": 1.8953,
+      "step": 1435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3714331428571429e-05,
+      "loss": 1.7415,
+      "step": 1436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3714062857142857e-05,
+      "loss": 1.6237,
+      "step": 1437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3713794285714287e-05,
+      "loss": 1.9,
+      "step": 1438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3713525714285715e-05,
+      "loss": 1.957,
+      "step": 1439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3713257142857143e-05,
+      "loss": 1.5875,
+      "step": 1440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3712988571428572e-05,
+      "loss": 1.6746,
+      "step": 1441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3712720000000001e-05,
+      "loss": 1.8062,
+      "step": 1442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3712451428571428e-05,
+      "loss": 1.7836,
+      "step": 1443
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3712182857142858e-05,
+      "loss": 1.729,
+      "step": 1444
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3711914285714286e-05,
+      "loss": 1.7606,
+      "step": 1445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3711645714285716e-05,
+      "loss": 1.9973,
+      "step": 1446
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3711377142857142e-05,
+      "loss": 1.9767,
+      "step": 1447
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3711108571428572e-05,
+      "loss": 2.03,
+      "step": 1448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.371084e-05,
+      "loss": 1.7482,
+      "step": 1449
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3710571428571429e-05,
+      "loss": 1.6597,
+      "step": 1450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3710302857142858e-05,
+      "loss": 1.7815,
+      "step": 1451
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3710034285714287e-05,
+      "loss": 1.6393,
+      "step": 1452
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3709765714285715e-05,
+      "loss": 1.821,
+      "step": 1453
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3709497142857143e-05,
+      "loss": 1.6211,
+      "step": 1454
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3709228571428573e-05,
+      "loss": 1.7043,
+      "step": 1455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.370896e-05,
+      "loss": 2.0176,
+      "step": 1456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3708691428571429e-05,
+      "loss": 1.8729,
+      "step": 1457
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3708422857142857e-05,
+      "loss": 1.6837,
+      "step": 1458
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3708154285714287e-05,
+      "loss": 1.7394,
+      "step": 1459
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3707885714285714e-05,
+      "loss": 1.6933,
+      "step": 1460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3707617142857144e-05,
+      "loss": 2.0576,
+      "step": 1461
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3707348571428572e-05,
+      "loss": 2.0646,
+      "step": 1462
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.370708e-05,
+      "loss": 1.6168,
+      "step": 1463
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.370681142857143e-05,
+      "loss": 1.8955,
+      "step": 1464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3706542857142858e-05,
+      "loss": 1.743,
+      "step": 1465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3706274285714286e-05,
+      "loss": 1.7169,
+      "step": 1466
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3706005714285714e-05,
+      "loss": 2.1075,
+      "step": 1467
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3705737142857144e-05,
+      "loss": 1.7184,
+      "step": 1468
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3705468571428572e-05,
+      "loss": 1.7246,
+      "step": 1469
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.37052e-05,
+      "loss": 1.5225,
+      "step": 1470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3704931428571429e-05,
+      "loss": 1.8787,
+      "step": 1471
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3704662857142859e-05,
+      "loss": 1.7245,
+      "step": 1472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3704394285714285e-05,
+      "loss": 1.8388,
+      "step": 1473
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3704125714285715e-05,
+      "loss": 1.9859,
+      "step": 1474
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3703857142857143e-05,
+      "loss": 1.9729,
+      "step": 1475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3703588571428571e-05,
+      "loss": 1.8409,
+      "step": 1476
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3703320000000001e-05,
+      "loss": 1.5953,
+      "step": 1477
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.370305142857143e-05,
+      "loss": 1.8209,
+      "step": 1478
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3702782857142857e-05,
+      "loss": 1.9805,
+      "step": 1479
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3702514285714286e-05,
+      "loss": 1.8738,
+      "step": 1480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3702245714285716e-05,
+      "loss": 1.5078,
+      "step": 1481
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3701977142857144e-05,
+      "loss": 1.7616,
+      "step": 1482
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3701708571428572e-05,
+      "loss": 1.6751,
+      "step": 1483
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.370144e-05,
+      "loss": 1.7036,
+      "step": 1484
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.370117142857143e-05,
+      "loss": 1.9331,
+      "step": 1485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3700902857142856e-05,
+      "loss": 1.8385,
+      "step": 1486
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3700634285714286e-05,
+      "loss": 1.3601,
+      "step": 1487
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3700365714285714e-05,
+      "loss": 1.6674,
+      "step": 1488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3700097142857143e-05,
+      "loss": 1.8192,
+      "step": 1489
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3699828571428573e-05,
+      "loss": 1.727,
+      "step": 1490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.369956e-05,
+      "loss": 2.073,
+      "step": 1491
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.369929142857143e-05,
+      "loss": 2.0731,
+      "step": 1492
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3699022857142857e-05,
+      "loss": 1.7726,
+      "step": 1493
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3698754285714287e-05,
+      "loss": 1.7078,
+      "step": 1494
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3698485714285715e-05,
+      "loss": 1.9427,
+      "step": 1495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3698217142857143e-05,
+      "loss": 1.5777,
+      "step": 1496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3697948571428571e-05,
+      "loss": 1.8046,
+      "step": 1497
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3697680000000001e-05,
+      "loss": 1.5925,
+      "step": 1498
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3697411428571428e-05,
+      "loss": 1.7956,
+      "step": 1499
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3697142857142858e-05,
+      "loss": 1.5516,
+      "step": 1500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3696874285714286e-05,
+      "loss": 1.729,
+      "step": 1501
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3696605714285714e-05,
+      "loss": 1.8423,
+      "step": 1502
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3696337142857144e-05,
+      "loss": 2.0479,
+      "step": 1503
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3696068571428572e-05,
+      "loss": 1.8426,
+      "step": 1504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3695800000000002e-05,
+      "loss": 1.5654,
+      "step": 1505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3695531428571428e-05,
+      "loss": 1.7683,
+      "step": 1506
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3695262857142858e-05,
+      "loss": 1.6219,
+      "step": 1507
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3694994285714286e-05,
+      "loss": 2.033,
+      "step": 1508
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3694725714285715e-05,
+      "loss": 1.7762,
+      "step": 1509
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3694457142857143e-05,
+      "loss": 1.6279,
+      "step": 1510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3694188571428573e-05,
+      "loss": 1.8322,
+      "step": 1511
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.369392e-05,
+      "loss": 1.9166,
+      "step": 1512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3693651428571429e-05,
+      "loss": 1.7412,
+      "step": 1513
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3693382857142857e-05,
+      "loss": 1.9197,
+      "step": 1514
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3693114285714287e-05,
+      "loss": 1.7571,
+      "step": 1515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3692845714285715e-05,
+      "loss": 1.4532,
+      "step": 1516
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3692577142857143e-05,
+      "loss": 1.9974,
+      "step": 1517
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3692308571428573e-05,
+      "loss": 1.7291,
+      "step": 1518
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.369204e-05,
+      "loss": 1.7697,
+      "step": 1519
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.369177142857143e-05,
+      "loss": 1.9899,
+      "step": 1520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3691502857142858e-05,
+      "loss": 1.9458,
+      "step": 1521
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3691234285714286e-05,
+      "loss": 1.9681,
+      "step": 1522
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3690965714285714e-05,
+      "loss": 1.3624,
+      "step": 1523
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3690697142857144e-05,
+      "loss": 1.8836,
+      "step": 1524
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.369042857142857e-05,
+      "loss": 1.5552,
+      "step": 1525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.369016e-05,
+      "loss": 1.7244,
+      "step": 1526
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3689891428571429e-05,
+      "loss": 1.7165,
+      "step": 1527
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3689622857142858e-05,
+      "loss": 1.956,
+      "step": 1528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3689354285714287e-05,
+      "loss": 1.6821,
+      "step": 1529
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3689085714285715e-05,
+      "loss": 1.6652,
+      "step": 1530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3688817142857145e-05,
+      "loss": 1.6029,
+      "step": 1531
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3688548571428571e-05,
+      "loss": 1.5104,
+      "step": 1532
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3688280000000001e-05,
+      "loss": 1.8125,
+      "step": 1533
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.368801142857143e-05,
+      "loss": 1.6792,
+      "step": 1534
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3687742857142857e-05,
+      "loss": 1.7384,
+      "step": 1535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3687474285714286e-05,
+      "loss": 1.8233,
+      "step": 1536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3687205714285715e-05,
+      "loss": 1.7286,
+      "step": 1537
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3686937142857144e-05,
+      "loss": 1.6445,
+      "step": 1538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3686668571428572e-05,
+      "loss": 2.0472,
+      "step": 1539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.36864e-05,
+      "loss": 1.7254,
+      "step": 1540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.368613142857143e-05,
+      "loss": 1.4405,
+      "step": 1541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3685862857142858e-05,
+      "loss": 1.6682,
+      "step": 1542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3685594285714286e-05,
+      "loss": 1.9957,
+      "step": 1543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3685325714285716e-05,
+      "loss": 1.7068,
+      "step": 1544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3685057142857143e-05,
+      "loss": 2.0099,
+      "step": 1545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3684788571428572e-05,
+      "loss": 1.8055,
+      "step": 1546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.368452e-05,
+      "loss": 1.8507,
+      "step": 1547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3684251428571429e-05,
+      "loss": 1.623,
+      "step": 1548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3683982857142857e-05,
+      "loss": 1.7313,
+      "step": 1549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3683714285714287e-05,
+      "loss": 1.5439,
+      "step": 1550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3683445714285715e-05,
+      "loss": 1.4951,
+      "step": 1551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3683177142857143e-05,
+      "loss": 1.7512,
+      "step": 1552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3682908571428571e-05,
+      "loss": 1.6406,
+      "step": 1553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3682640000000001e-05,
+      "loss": 1.5853,
+      "step": 1554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.368237142857143e-05,
+      "loss": 1.5476,
+      "step": 1555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3682102857142858e-05,
+      "loss": 1.9851,
+      "step": 1556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3681834285714287e-05,
+      "loss": 1.7019,
+      "step": 1557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3681565714285714e-05,
+      "loss": 1.8754,
+      "step": 1558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3681297142857144e-05,
+      "loss": 2.0297,
+      "step": 1559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3681028571428572e-05,
+      "loss": 2.0248,
+      "step": 1560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3680760000000002e-05,
+      "loss": 1.8116,
+      "step": 1561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3680491428571428e-05,
+      "loss": 1.8797,
+      "step": 1562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3680222857142858e-05,
+      "loss": 1.9481,
+      "step": 1563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3679954285714286e-05,
+      "loss": 1.5338,
+      "step": 1564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3679685714285715e-05,
+      "loss": 1.8726,
+      "step": 1565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3679417142857143e-05,
+      "loss": 1.7314,
+      "step": 1566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3679148571428573e-05,
+      "loss": 1.6541,
+      "step": 1567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.367888e-05,
+      "loss": 1.7617,
+      "step": 1568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3678611428571429e-05,
+      "loss": 1.6375,
+      "step": 1569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3678342857142859e-05,
+      "loss": 1.7614,
+      "step": 1570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3678074285714285e-05,
+      "loss": 1.7294,
+      "step": 1571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3677805714285715e-05,
+      "loss": 1.9267,
+      "step": 1572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3677537142857143e-05,
+      "loss": 1.7084,
+      "step": 1573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3677268571428573e-05,
+      "loss": 1.8281,
+      "step": 1574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3677e-05,
+      "loss": 1.7739,
+      "step": 1575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.367673142857143e-05,
+      "loss": 1.5583,
+      "step": 1576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3676462857142858e-05,
+      "loss": 1.4733,
+      "step": 1577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3676194285714286e-05,
+      "loss": 1.8282,
+      "step": 1578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3675925714285714e-05,
+      "loss": 1.7543,
+      "step": 1579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3675657142857144e-05,
+      "loss": 1.7654,
+      "step": 1580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3675388571428572e-05,
+      "loss": 1.8721,
+      "step": 1581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.367512e-05,
+      "loss": 1.5303,
+      "step": 1582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.367485142857143e-05,
+      "loss": 1.915,
+      "step": 1583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3674582857142857e-05,
+      "loss": 1.7844,
+      "step": 1584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3674314285714287e-05,
+      "loss": 1.5616,
+      "step": 1585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3674045714285715e-05,
+      "loss": 2.0721,
+      "step": 1586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3673777142857145e-05,
+      "loss": 1.8419,
+      "step": 1587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3673508571428571e-05,
+      "loss": 1.8606,
+      "step": 1588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3673240000000001e-05,
+      "loss": 1.8335,
+      "step": 1589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3672971428571429e-05,
+      "loss": 1.836,
+      "step": 1590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3672702857142857e-05,
+      "loss": 1.7916,
+      "step": 1591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3672434285714285e-05,
+      "loss": 1.8942,
+      "step": 1592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3672165714285715e-05,
+      "loss": 1.8728,
+      "step": 1593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3671897142857143e-05,
+      "loss": 1.5764,
+      "step": 1594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3671628571428572e-05,
+      "loss": 1.7518,
+      "step": 1595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3671360000000002e-05,
+      "loss": 1.683,
+      "step": 1596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.367109142857143e-05,
+      "loss": 1.7509,
+      "step": 1597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3670822857142858e-05,
+      "loss": 1.8934,
+      "step": 1598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3670554285714286e-05,
+      "loss": 1.8528,
+      "step": 1599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3670285714285716e-05,
+      "loss": 1.9548,
+      "step": 1600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3670017142857142e-05,
+      "loss": 1.8245,
+      "step": 1601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3669748571428572e-05,
+      "loss": 2.0698,
+      "step": 1602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.366948e-05,
+      "loss": 1.71,
+      "step": 1603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3669211428571429e-05,
+      "loss": 1.818,
+      "step": 1604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3668942857142857e-05,
+      "loss": 1.7744,
+      "step": 1605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3668674285714287e-05,
+      "loss": 1.723,
+      "step": 1606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3668405714285715e-05,
+      "loss": 1.6499,
+      "step": 1607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3668137142857143e-05,
+      "loss": 1.7855,
+      "step": 1608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3667868571428573e-05,
+      "loss": 1.5629,
+      "step": 1609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3667600000000001e-05,
+      "loss": 1.7274,
+      "step": 1610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.366733142857143e-05,
+      "loss": 1.6678,
+      "step": 1611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3667062857142857e-05,
+      "loss": 1.7418,
+      "step": 1612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3666794285714287e-05,
+      "loss": 1.7242,
+      "step": 1613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3666525714285714e-05,
+      "loss": 1.7937,
+      "step": 1614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3666257142857144e-05,
+      "loss": 1.5306,
+      "step": 1615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3665988571428572e-05,
+      "loss": 1.8034,
+      "step": 1616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.366572e-05,
+      "loss": 1.7901,
+      "step": 1617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3665451428571428e-05,
+      "loss": 1.421,
+      "step": 1618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3665182857142858e-05,
+      "loss": 1.5708,
+      "step": 1619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3664914285714286e-05,
+      "loss": 1.9114,
+      "step": 1620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3664645714285714e-05,
+      "loss": 1.7506,
+      "step": 1621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3664377142857144e-05,
+      "loss": 2.0229,
+      "step": 1622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3664108571428572e-05,
+      "loss": 1.5955,
+      "step": 1623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.366384e-05,
+      "loss": 1.4692,
+      "step": 1624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3663571428571429e-05,
+      "loss": 2.154,
+      "step": 1625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3663302857142859e-05,
+      "loss": 1.7447,
+      "step": 1626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3663034285714285e-05,
+      "loss": 1.9503,
+      "step": 1627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3662765714285715e-05,
+      "loss": 1.4442,
+      "step": 1628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3662497142857143e-05,
+      "loss": 1.5595,
+      "step": 1629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3662228571428571e-05,
+      "loss": 1.6774,
+      "step": 1630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.366196e-05,
+      "loss": 2.2366,
+      "step": 1631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.366169142857143e-05,
+      "loss": 1.8052,
+      "step": 1632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3661422857142858e-05,
+      "loss": 2.1172,
+      "step": 1633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3661154285714286e-05,
+      "loss": 1.7238,
+      "step": 1634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3660885714285716e-05,
+      "loss": 1.8352,
+      "step": 1635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3660617142857144e-05,
+      "loss": 1.9147,
+      "step": 1636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3660348571428572e-05,
+      "loss": 1.6417,
+      "step": 1637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.366008e-05,
+      "loss": 1.7214,
+      "step": 1638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.365981142857143e-05,
+      "loss": 1.6318,
+      "step": 1639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3659542857142857e-05,
+      "loss": 1.6519,
+      "step": 1640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3659274285714286e-05,
+      "loss": 1.66,
+      "step": 1641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3659005714285715e-05,
+      "loss": 1.8048,
+      "step": 1642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3658737142857144e-05,
+      "loss": 1.8259,
+      "step": 1643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3658468571428571e-05,
+      "loss": 1.7424,
+      "step": 1644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.36582e-05,
+      "loss": 1.7653,
+      "step": 1645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3657931428571429e-05,
+      "loss": 1.6959,
+      "step": 1646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3657662857142857e-05,
+      "loss": 1.5948,
+      "step": 1647
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3657394285714287e-05,
+      "loss": 1.355,
+      "step": 1648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3657125714285715e-05,
+      "loss": 1.6885,
+      "step": 1649
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3656857142857143e-05,
+      "loss": 1.8845,
+      "step": 1650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3656588571428572e-05,
+      "loss": 1.9438,
+      "step": 1651
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3656320000000001e-05,
+      "loss": 1.9125,
+      "step": 1652
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3656051428571428e-05,
+      "loss": 1.8105,
+      "step": 1653
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3655782857142858e-05,
+      "loss": 1.6808,
+      "step": 1654
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3655514285714286e-05,
+      "loss": 1.7247,
+      "step": 1655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3655245714285716e-05,
+      "loss": 1.9039,
+      "step": 1656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3654977142857142e-05,
+      "loss": 1.7309,
+      "step": 1657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3654708571428572e-05,
+      "loss": 1.6977,
+      "step": 1658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.365444e-05,
+      "loss": 1.6592,
+      "step": 1659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3654171428571429e-05,
+      "loss": 1.6445,
+      "step": 1660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3653902857142858e-05,
+      "loss": 2.0704,
+      "step": 1661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.3653634285714287e-05,
+      "loss": 1.4228,
+      "step": 1662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3653365714285715e-05,
+      "loss": 1.7092,
+      "step": 1663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3653097142857143e-05,
+      "loss": 1.6115,
+      "step": 1664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3652828571428573e-05,
+      "loss": 1.7764,
+      "step": 1665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3652560000000001e-05,
+      "loss": 1.7377,
+      "step": 1666
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3652291428571429e-05,
+      "loss": 1.9249,
+      "step": 1667
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3652022857142857e-05,
+      "loss": 2.008,
+      "step": 1668
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3651754285714287e-05,
+      "loss": 1.9014,
+      "step": 1669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3651485714285714e-05,
+      "loss": 1.8582,
+      "step": 1670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3651217142857144e-05,
+      "loss": 2.0293,
+      "step": 1671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3650948571428572e-05,
+      "loss": 1.9712,
+      "step": 1672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.365068e-05,
+      "loss": 1.9194,
+      "step": 1673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.365041142857143e-05,
+      "loss": 1.7227,
+      "step": 1674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3650142857142858e-05,
+      "loss": 1.8768,
+      "step": 1675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3649874285714286e-05,
+      "loss": 2.079,
+      "step": 1676
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3649605714285714e-05,
+      "loss": 1.8381,
+      "step": 1677
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3649337142857144e-05,
+      "loss": 2.0423,
+      "step": 1678
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3649068571428572e-05,
+      "loss": 1.8262,
+      "step": 1679
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.36488e-05,
+      "loss": 1.7301,
+      "step": 1680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3648531428571429e-05,
+      "loss": 1.9467,
+      "step": 1681
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3648262857142859e-05,
+      "loss": 1.599,
+      "step": 1682
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3647994285714285e-05,
+      "loss": 1.5432,
+      "step": 1683
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3647725714285715e-05,
+      "loss": 1.6385,
+      "step": 1684
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3647457142857143e-05,
+      "loss": 1.9589,
+      "step": 1685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3647188571428571e-05,
+      "loss": 1.6894,
+      "step": 1686
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3646920000000001e-05,
+      "loss": 1.6263,
+      "step": 1687
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.364665142857143e-05,
+      "loss": 1.8538,
+      "step": 1688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.364638285714286e-05,
+      "loss": 1.8074,
+      "step": 1689
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3646114285714286e-05,
+      "loss": 1.7674,
+      "step": 1690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3645845714285716e-05,
+      "loss": 1.5605,
+      "step": 1691
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3645577142857144e-05,
+      "loss": 1.7411,
+      "step": 1692
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3645308571428572e-05,
+      "loss": 1.6632,
+      "step": 1693
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.364504e-05,
+      "loss": 1.8566,
+      "step": 1694
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.364477142857143e-05,
+      "loss": 1.8561,
+      "step": 1695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3644502857142856e-05,
+      "loss": 1.6,
+      "step": 1696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3644234285714286e-05,
+      "loss": 1.8589,
+      "step": 1697
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3643965714285714e-05,
+      "loss": 1.6705,
+      "step": 1698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3643697142857143e-05,
+      "loss": 1.6826,
+      "step": 1699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3643428571428573e-05,
+      "loss": 1.5083,
+      "step": 1700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.364316e-05,
+      "loss": 1.7694,
+      "step": 1701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.364289142857143e-05,
+      "loss": 2.0418,
+      "step": 1702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3642622857142857e-05,
+      "loss": 1.6884,
+      "step": 1703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3642354285714287e-05,
+      "loss": 1.4791,
+      "step": 1704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3642085714285715e-05,
+      "loss": 1.7926,
+      "step": 1705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3641817142857143e-05,
+      "loss": 1.7044,
+      "step": 1706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3641548571428571e-05,
+      "loss": 1.5848,
+      "step": 1707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3641280000000001e-05,
+      "loss": 1.77,
+      "step": 1708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3641011428571428e-05,
+      "loss": 1.6991,
+      "step": 1709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3640742857142858e-05,
+      "loss": 1.782,
+      "step": 1710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3640474285714286e-05,
+      "loss": 1.5916,
+      "step": 1711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3640205714285716e-05,
+      "loss": 1.4385,
+      "step": 1712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3639937142857144e-05,
+      "loss": 1.767,
+      "step": 1713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3639668571428572e-05,
+      "loss": 1.5324,
+      "step": 1714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3639400000000002e-05,
+      "loss": 1.5625,
+      "step": 1715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3639131428571428e-05,
+      "loss": 1.6133,
+      "step": 1716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3638862857142858e-05,
+      "loss": 1.7988,
+      "step": 1717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3638594285714286e-05,
+      "loss": 1.8174,
+      "step": 1718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3638325714285715e-05,
+      "loss": 1.7804,
+      "step": 1719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3638057142857143e-05,
+      "loss": 1.7106,
+      "step": 1720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3637788571428573e-05,
+      "loss": 1.7096,
+      "step": 1721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.363752e-05,
+      "loss": 1.6841,
+      "step": 1722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3637251428571429e-05,
+      "loss": 1.6265,
+      "step": 1723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3636982857142857e-05,
+      "loss": 1.7842,
+      "step": 1724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3636714285714287e-05,
+      "loss": 2.0712,
+      "step": 1725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3636445714285715e-05,
+      "loss": 1.9456,
+      "step": 1726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3636177142857143e-05,
+      "loss": 1.5532,
+      "step": 1727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3635908571428573e-05,
+      "loss": 1.7857,
+      "step": 1728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.363564e-05,
+      "loss": 1.7302,
+      "step": 1729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.363537142857143e-05,
+      "loss": 2.0737,
+      "step": 1730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3635102857142858e-05,
+      "loss": 1.6068,
+      "step": 1731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3634834285714286e-05,
+      "loss": 1.8899,
+      "step": 1732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3634565714285714e-05,
+      "loss": 1.7108,
+      "step": 1733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3634297142857144e-05,
+      "loss": 2.0804,
+      "step": 1734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3634028571428572e-05,
+      "loss": 1.5626,
+      "step": 1735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.363376e-05,
+      "loss": 1.8177,
+      "step": 1736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3633491428571429e-05,
+      "loss": 2.0828,
+      "step": 1737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3633222857142858e-05,
+      "loss": 1.5981,
+      "step": 1738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3632954285714287e-05,
+      "loss": 1.6717,
+      "step": 1739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3632685714285715e-05,
+      "loss": 2.0544,
+      "step": 1740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3632417142857145e-05,
+      "loss": 1.6519,
+      "step": 1741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3632148571428571e-05,
+      "loss": 1.7882,
+      "step": 1742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3631880000000001e-05,
+      "loss": 1.7513,
+      "step": 1743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.363161142857143e-05,
+      "loss": 1.7328,
+      "step": 1744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3631342857142857e-05,
+      "loss": 2.0857,
+      "step": 1745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3631074285714286e-05,
+      "loss": 1.5406,
+      "step": 1746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3630805714285715e-05,
+      "loss": 1.422,
+      "step": 1747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3630537142857144e-05,
+      "loss": 1.598,
+      "step": 1748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3630268571428572e-05,
+      "loss": 1.9169,
+      "step": 1749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.363e-05,
+      "loss": 1.5693,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.362973142857143e-05,
+      "loss": 1.6302,
+      "step": 1751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3629462857142858e-05,
+      "loss": 2.0614,
+      "step": 1752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3629194285714286e-05,
+      "loss": 2.0372,
+      "step": 1753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3628925714285716e-05,
+      "loss": 1.8396,
+      "step": 1754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3628657142857143e-05,
+      "loss": 1.6172,
+      "step": 1755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3628388571428572e-05,
+      "loss": 1.8203,
+      "step": 1756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.362812e-05,
+      "loss": 2.068,
+      "step": 1757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3627851428571429e-05,
+      "loss": 1.8126,
+      "step": 1758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3627582857142857e-05,
+      "loss": 2.1504,
+      "step": 1759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3627314285714287e-05,
+      "loss": 1.9159,
+      "step": 1760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3627045714285715e-05,
+      "loss": 1.7683,
+      "step": 1761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3626777142857143e-05,
+      "loss": 1.7522,
+      "step": 1762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3626508571428571e-05,
+      "loss": 1.8757,
+      "step": 1763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3626240000000001e-05,
+      "loss": 1.4435,
+      "step": 1764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3625971428571428e-05,
+      "loss": 1.916,
+      "step": 1765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3625702857142858e-05,
+      "loss": 1.7119,
+      "step": 1766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3625434285714287e-05,
+      "loss": 1.8771,
+      "step": 1767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3625165714285714e-05,
+      "loss": 1.5182,
+      "step": 1768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3624897142857144e-05,
+      "loss": 1.7018,
+      "step": 1769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3624628571428572e-05,
+      "loss": 1.6704,
+      "step": 1770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3624360000000002e-05,
+      "loss": 1.9486,
+      "step": 1771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3624091428571428e-05,
+      "loss": 1.5993,
+      "step": 1772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3623822857142858e-05,
+      "loss": 1.7288,
+      "step": 1773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3623554285714286e-05,
+      "loss": 1.8066,
+      "step": 1774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3623285714285715e-05,
+      "loss": 1.8139,
+      "step": 1775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3623017142857143e-05,
+      "loss": 1.7554,
+      "step": 1776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3622748571428573e-05,
+      "loss": 1.8626,
+      "step": 1777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3622479999999999e-05,
+      "loss": 1.556,
+      "step": 1778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3622211428571429e-05,
+      "loss": 1.7585,
+      "step": 1779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3621942857142859e-05,
+      "loss": 1.6538,
+      "step": 1780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3621674285714285e-05,
+      "loss": 1.8713,
+      "step": 1781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3621405714285715e-05,
+      "loss": 1.923,
+      "step": 1782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3621137142857143e-05,
+      "loss": 1.5371,
+      "step": 1783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3620868571428573e-05,
+      "loss": 1.9358,
+      "step": 1784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.36206e-05,
+      "loss": 1.9701,
+      "step": 1785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.362033142857143e-05,
+      "loss": 1.9346,
+      "step": 1786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3620062857142858e-05,
+      "loss": 1.9754,
+      "step": 1787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3619794285714286e-05,
+      "loss": 1.97,
+      "step": 1788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3619525714285714e-05,
+      "loss": 1.8656,
+      "step": 1789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3619257142857144e-05,
+      "loss": 1.6722,
+      "step": 1790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.361898857142857e-05,
+      "loss": 1.941,
+      "step": 1791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.361872e-05,
+      "loss": 1.7757,
+      "step": 1792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.361845142857143e-05,
+      "loss": 1.7674,
+      "step": 1793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3618182857142858e-05,
+      "loss": 2.0696,
+      "step": 1794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3617914285714287e-05,
+      "loss": 1.9155,
+      "step": 1795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3617645714285715e-05,
+      "loss": 1.4915,
+      "step": 1796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3617377142857145e-05,
+      "loss": 1.7776,
+      "step": 1797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3617108571428571e-05,
+      "loss": 2.0169,
+      "step": 1798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3616840000000001e-05,
+      "loss": 1.801,
+      "step": 1799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3616571428571429e-05,
+      "loss": 2.0129,
+      "step": 1800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3616302857142857e-05,
+      "loss": 1.9245,
+      "step": 1801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3616034285714285e-05,
+      "loss": 1.7425,
+      "step": 1802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3615765714285715e-05,
+      "loss": 1.6261,
+      "step": 1803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3615497142857142e-05,
+      "loss": 1.6301,
+      "step": 1804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3615228571428572e-05,
+      "loss": 1.7658,
+      "step": 1805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3614960000000002e-05,
+      "loss": 1.5076,
+      "step": 1806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.361469142857143e-05,
+      "loss": 1.6738,
+      "step": 1807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3614422857142858e-05,
+      "loss": 1.8069,
+      "step": 1808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3614154285714286e-05,
+      "loss": 1.5567,
+      "step": 1809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3613885714285716e-05,
+      "loss": 1.774,
+      "step": 1810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3613617142857142e-05,
+      "loss": 1.5868,
+      "step": 1811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3613348571428572e-05,
+      "loss": 1.7083,
+      "step": 1812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.361308e-05,
+      "loss": 1.3764,
+      "step": 1813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3612811428571429e-05,
+      "loss": 1.4753,
+      "step": 1814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3612542857142857e-05,
+      "loss": 1.6485,
+      "step": 1815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3612274285714287e-05,
+      "loss": 1.7813,
+      "step": 1816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3612005714285715e-05,
+      "loss": 1.6959,
+      "step": 1817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3611737142857143e-05,
+      "loss": 1.7085,
+      "step": 1818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3611468571428573e-05,
+      "loss": 1.6066,
+      "step": 1819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3611200000000001e-05,
+      "loss": 1.5724,
+      "step": 1820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.361093142857143e-05,
+      "loss": 1.7076,
+      "step": 1821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3610662857142857e-05,
+      "loss": 1.4886,
+      "step": 1822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3610394285714287e-05,
+      "loss": 1.6643,
+      "step": 1823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3610125714285714e-05,
+      "loss": 1.7151,
+      "step": 1824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3609857142857144e-05,
+      "loss": 1.9242,
+      "step": 1825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3609588571428572e-05,
+      "loss": 2.0468,
+      "step": 1826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.360932e-05,
+      "loss": 1.7195,
+      "step": 1827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3609051428571428e-05,
+      "loss": 2.2046,
+      "step": 1828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3608782857142858e-05,
+      "loss": 1.9257,
+      "step": 1829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3608514285714286e-05,
+      "loss": 1.55,
+      "step": 1830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3608245714285714e-05,
+      "loss": 1.776,
+      "step": 1831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3607977142857144e-05,
+      "loss": 1.5624,
+      "step": 1832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3607708571428572e-05,
+      "loss": 1.66,
+      "step": 1833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.360744e-05,
+      "loss": 1.6512,
+      "step": 1834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3607171428571429e-05,
+      "loss": 1.5973,
+      "step": 1835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3606902857142859e-05,
+      "loss": 1.6419,
+      "step": 1836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3606634285714285e-05,
+      "loss": 1.7218,
+      "step": 1837
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3606365714285715e-05,
+      "loss": 1.9203,
+      "step": 1838
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3606097142857143e-05,
+      "loss": 1.8261,
+      "step": 1839
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3605828571428573e-05,
+      "loss": 1.69,
+      "step": 1840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.360556e-05,
+      "loss": 1.558,
+      "step": 1841
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.360529142857143e-05,
+      "loss": 1.6048,
+      "step": 1842
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3605022857142858e-05,
+      "loss": 2.0013,
+      "step": 1843
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3604754285714286e-05,
+      "loss": 1.8888,
+      "step": 1844
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3604485714285716e-05,
+      "loss": 1.588,
+      "step": 1845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3604217142857144e-05,
+      "loss": 1.8088,
+      "step": 1846
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3603948571428572e-05,
+      "loss": 1.6871,
+      "step": 1847
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.360368e-05,
+      "loss": 1.9702,
+      "step": 1848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.360341142857143e-05,
+      "loss": 1.8538,
+      "step": 1849
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3603142857142857e-05,
+      "loss": 1.7112,
+      "step": 1850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3602874285714286e-05,
+      "loss": 1.6244,
+      "step": 1851
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3602605714285715e-05,
+      "loss": 1.4847,
+      "step": 1852
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3602337142857144e-05,
+      "loss": 1.8641,
+      "step": 1853
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3602068571428571e-05,
+      "loss": 1.6634,
+      "step": 1854
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.36018e-05,
+      "loss": 1.9276,
+      "step": 1855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3601531428571429e-05,
+      "loss": 1.6311,
+      "step": 1856
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3601262857142857e-05,
+      "loss": 1.751,
+      "step": 1857
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3600994285714287e-05,
+      "loss": 2.0695,
+      "step": 1858
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3600725714285715e-05,
+      "loss": 1.833,
+      "step": 1859
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3600457142857143e-05,
+      "loss": 1.844,
+      "step": 1860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3600188571428572e-05,
+      "loss": 1.4602,
+      "step": 1861
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3599920000000001e-05,
+      "loss": 1.7133,
+      "step": 1862
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.359965142857143e-05,
+      "loss": 1.5634,
+      "step": 1863
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3599382857142858e-05,
+      "loss": 1.7365,
+      "step": 1864
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3599114285714286e-05,
+      "loss": 1.7296,
+      "step": 1865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3598845714285716e-05,
+      "loss": 1.6657,
+      "step": 1866
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3598577142857142e-05,
+      "loss": 1.8484,
+      "step": 1867
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3598308571428572e-05,
+      "loss": 1.5607,
+      "step": 1868
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.359804e-05,
+      "loss": 1.6294,
+      "step": 1869
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3597771428571429e-05,
+      "loss": 1.8033,
+      "step": 1870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3597502857142857e-05,
+      "loss": 1.6589,
+      "step": 1871
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3597234285714287e-05,
+      "loss": 1.856,
+      "step": 1872
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3596965714285715e-05,
+      "loss": 1.7345,
+      "step": 1873
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3596697142857143e-05,
+      "loss": 1.7726,
+      "step": 1874
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3596428571428573e-05,
+      "loss": 1.7784,
+      "step": 1875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3596160000000001e-05,
+      "loss": 1.8138,
+      "step": 1876
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3595891428571429e-05,
+      "loss": 1.7162,
+      "step": 1877
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3595622857142857e-05,
+      "loss": 1.5182,
+      "step": 1878
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3595354285714287e-05,
+      "loss": 1.7284,
+      "step": 1879
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3595085714285714e-05,
+      "loss": 1.8944,
+      "step": 1880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3594817142857144e-05,
+      "loss": 1.8155,
+      "step": 1881
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3594548571428572e-05,
+      "loss": 1.7326,
+      "step": 1882
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.359428e-05,
+      "loss": 1.5828,
+      "step": 1883
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3594011428571428e-05,
+      "loss": 1.91,
+      "step": 1884
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3593742857142858e-05,
+      "loss": 1.6563,
+      "step": 1885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3593474285714288e-05,
+      "loss": 1.7728,
+      "step": 1886
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3593205714285714e-05,
+      "loss": 1.6191,
+      "step": 1887
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3592937142857144e-05,
+      "loss": 1.8832,
+      "step": 1888
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3592668571428572e-05,
+      "loss": 1.9443,
+      "step": 1889
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.35924e-05,
+      "loss": 1.7142,
+      "step": 1890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3592131428571429e-05,
+      "loss": 1.9746,
+      "step": 1891
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3591862857142859e-05,
+      "loss": 1.6378,
+      "step": 1892
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3591594285714285e-05,
+      "loss": 1.7553,
+      "step": 1893
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3591325714285715e-05,
+      "loss": 1.861,
+      "step": 1894
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3591057142857143e-05,
+      "loss": 1.4878,
+      "step": 1895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3590788571428571e-05,
+      "loss": 1.5607,
+      "step": 1896
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.359052e-05,
+      "loss": 1.8044,
+      "step": 1897
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.359025142857143e-05,
+      "loss": 1.4886,
+      "step": 1898
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.358998285714286e-05,
+      "loss": 1.7805,
+      "step": 1899
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3589714285714286e-05,
+      "loss": 1.4263,
+      "step": 1900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3589445714285716e-05,
+      "loss": 1.5795,
+      "step": 1901
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3589177142857144e-05,
+      "loss": 1.6686,
+      "step": 1902
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3588908571428572e-05,
+      "loss": 1.8764,
+      "step": 1903
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.358864e-05,
+      "loss": 1.5692,
+      "step": 1904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.358837142857143e-05,
+      "loss": 1.6415,
+      "step": 1905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3588102857142856e-05,
+      "loss": 1.622,
+      "step": 1906
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3587834285714286e-05,
+      "loss": 1.5711,
+      "step": 1907
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3587565714285714e-05,
+      "loss": 1.8421,
+      "step": 1908
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3587297142857144e-05,
+      "loss": 1.542,
+      "step": 1909
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.358702857142857e-05,
+      "loss": 1.797,
+      "step": 1910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.358676e-05,
+      "loss": 1.5756,
+      "step": 1911
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.358649142857143e-05,
+      "loss": 1.6527,
+      "step": 1912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3586222857142857e-05,
+      "loss": 1.5933,
+      "step": 1913
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3585954285714287e-05,
+      "loss": 1.911,
+      "step": 1914
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3585685714285715e-05,
+      "loss": 1.9827,
+      "step": 1915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3585417142857143e-05,
+      "loss": 1.8179,
+      "step": 1916
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3585148571428571e-05,
+      "loss": 1.731,
+      "step": 1917
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3584880000000001e-05,
+      "loss": 1.4008,
+      "step": 1918
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3584611428571428e-05,
+      "loss": 1.5889,
+      "step": 1919
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3584342857142858e-05,
+      "loss": 1.9338,
+      "step": 1920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3584074285714286e-05,
+      "loss": 2.0459,
+      "step": 1921
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3583805714285716e-05,
+      "loss": 1.9202,
+      "step": 1922
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3583537142857142e-05,
+      "loss": 1.8946,
+      "step": 1923
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3583268571428572e-05,
+      "loss": 1.7459,
+      "step": 1924
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3583000000000002e-05,
+      "loss": 1.8919,
+      "step": 1925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3582731428571428e-05,
+      "loss": 1.4949,
+      "step": 1926
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3582462857142858e-05,
+      "loss": 1.7347,
+      "step": 1927
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3582194285714286e-05,
+      "loss": 1.6706,
+      "step": 1928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3581925714285715e-05,
+      "loss": 1.5869,
+      "step": 1929
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3581657142857143e-05,
+      "loss": 1.5093,
+      "step": 1930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3581388571428573e-05,
+      "loss": 1.9265,
+      "step": 1931
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3581119999999999e-05,
+      "loss": 1.5121,
+      "step": 1932
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3580851428571429e-05,
+      "loss": 1.6741,
+      "step": 1933
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3580582857142857e-05,
+      "loss": 1.77,
+      "step": 1934
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3580314285714287e-05,
+      "loss": 1.8259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3580045714285714e-05,
+      "loss": 1.8886,
+      "step": 1936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3579777142857143e-05,
+      "loss": 1.6787,
+      "step": 1937
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3579508571428573e-05,
+      "loss": 1.4639,
+      "step": 1938
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.357924e-05,
+      "loss": 1.563,
+      "step": 1939
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.357897142857143e-05,
+      "loss": 1.5913,
+      "step": 1940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3578702857142858e-05,
+      "loss": 1.6602,
+      "step": 1941
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3578434285714286e-05,
+      "loss": 1.8851,
+      "step": 1942
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3578165714285714e-05,
+      "loss": 1.5635,
+      "step": 1943
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3577897142857144e-05,
+      "loss": 1.8397,
+      "step": 1944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3577628571428572e-05,
+      "loss": 1.8517,
+      "step": 1945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.357736e-05,
+      "loss": 1.8043,
+      "step": 1946
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3577091428571429e-05,
+      "loss": 1.633,
+      "step": 1947
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3576822857142858e-05,
+      "loss": 1.7673,
+      "step": 1948
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3576554285714285e-05,
+      "loss": 1.8667,
+      "step": 1949
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3576285714285715e-05,
+      "loss": 1.8577,
+      "step": 1950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3576017142857145e-05,
+      "loss": 1.5341,
+      "step": 1951
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3575748571428571e-05,
+      "loss": 1.8156,
+      "step": 1952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3575480000000001e-05,
+      "loss": 1.7533,
+      "step": 1953
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.357521142857143e-05,
+      "loss": 1.6013,
+      "step": 1954
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3574942857142857e-05,
+      "loss": 1.7472,
+      "step": 1955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3574674285714286e-05,
+      "loss": 1.9502,
+      "step": 1956
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3574405714285715e-05,
+      "loss": 1.704,
+      "step": 1957
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3574137142857144e-05,
+      "loss": 1.8076,
+      "step": 1958
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3573868571428572e-05,
+      "loss": 1.7157,
+      "step": 1959
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.35736e-05,
+      "loss": 1.586,
+      "step": 1960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.357333142857143e-05,
+      "loss": 1.682,
+      "step": 1961
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3573062857142856e-05,
+      "loss": 1.5581,
+      "step": 1962
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3572794285714286e-05,
+      "loss": 1.8666,
+      "step": 1963
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3572525714285716e-05,
+      "loss": 1.6824,
+      "step": 1964
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3572257142857143e-05,
+      "loss": 1.5075,
+      "step": 1965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3571988571428572e-05,
+      "loss": 1.68,
+      "step": 1966
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.357172e-05,
+      "loss": 1.9694,
+      "step": 1967
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.357145142857143e-05,
+      "loss": 1.7245,
+      "step": 1968
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3571182857142857e-05,
+      "loss": 1.455,
+      "step": 1969
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3570914285714287e-05,
+      "loss": 1.8552,
+      "step": 1970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3570645714285715e-05,
+      "loss": 1.5576,
+      "step": 1971
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3570377142857143e-05,
+      "loss": 1.3208,
+      "step": 1972
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3570108571428571e-05,
+      "loss": 1.525,
+      "step": 1973
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3569840000000001e-05,
+      "loss": 1.6206,
+      "step": 1974
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3569571428571428e-05,
+      "loss": 1.6228,
+      "step": 1975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3569302857142858e-05,
+      "loss": 1.6663,
+      "step": 1976
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3569034285714286e-05,
+      "loss": 1.6099,
+      "step": 1977
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3568765714285714e-05,
+      "loss": 1.8368,
+      "step": 1978
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3568497142857144e-05,
+      "loss": 1.6959,
+      "step": 1979
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3568228571428572e-05,
+      "loss": 1.7652,
+      "step": 1980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3567960000000002e-05,
+      "loss": 1.5629,
+      "step": 1981
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3567691428571428e-05,
+      "loss": 1.6386,
+      "step": 1982
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3567422857142858e-05,
+      "loss": 1.3674,
+      "step": 1983
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3567154285714286e-05,
+      "loss": 1.7023,
+      "step": 1984
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3566885714285715e-05,
+      "loss": 1.5596,
+      "step": 1985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3566617142857143e-05,
+      "loss": 1.5618,
+      "step": 1986
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3566348571428573e-05,
+      "loss": 1.3782,
+      "step": 1987
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3566079999999999e-05,
+      "loss": 1.6603,
+      "step": 1988
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3565811428571429e-05,
+      "loss": 1.7969,
+      "step": 1989
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3565542857142857e-05,
+      "loss": 1.6968,
+      "step": 1990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3565274285714287e-05,
+      "loss": 1.8908,
+      "step": 1991
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3565005714285715e-05,
+      "loss": 1.7385,
+      "step": 1992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3564737142857143e-05,
+      "loss": 1.7669,
+      "step": 1993
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3564468571428573e-05,
+      "loss": 1.6685,
+      "step": 1994
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.35642e-05,
+      "loss": 1.6986,
+      "step": 1995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.356393142857143e-05,
+      "loss": 1.7755,
+      "step": 1996
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3563662857142858e-05,
+      "loss": 1.5468,
+      "step": 1997
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3563394285714286e-05,
+      "loss": 1.7591,
+      "step": 1998
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3563125714285714e-05,
+      "loss": 2.0419,
+      "step": 1999
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3562857142857144e-05,
+      "loss": 1.6823,
+      "step": 2000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.356258857142857e-05,
+      "loss": 1.9996,
+      "step": 2001
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.356232e-05,
+      "loss": 1.8707,
+      "step": 2002
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3562051428571428e-05,
+      "loss": 1.9903,
+      "step": 2003
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3561782857142858e-05,
+      "loss": 1.8581,
+      "step": 2004
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3561514285714287e-05,
+      "loss": 1.9281,
+      "step": 2005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3561245714285715e-05,
+      "loss": 1.7859,
+      "step": 2006
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3560977142857145e-05,
+      "loss": 1.7165,
+      "step": 2007
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3560708571428571e-05,
+      "loss": 1.806,
+      "step": 2008
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3560440000000001e-05,
+      "loss": 1.5776,
+      "step": 2009
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3560171428571429e-05,
+      "loss": 1.8249,
+      "step": 2010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3559902857142857e-05,
+      "loss": 2.005,
+      "step": 2011
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3559634285714285e-05,
+      "loss": 1.5948,
+      "step": 2012
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3559365714285715e-05,
+      "loss": 1.7748,
+      "step": 2013
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3559097142857143e-05,
+      "loss": 1.6199,
+      "step": 2014
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3558828571428572e-05,
+      "loss": 1.9052,
+      "step": 2015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.355856e-05,
+      "loss": 1.9966,
+      "step": 2016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.355829142857143e-05,
+      "loss": 1.5475,
+      "step": 2017
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3558022857142858e-05,
+      "loss": 1.6661,
+      "step": 2018
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3557754285714286e-05,
+      "loss": 1.7472,
+      "step": 2019
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3557485714285716e-05,
+      "loss": 1.65,
+      "step": 2020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3557217142857142e-05,
+      "loss": 1.6737,
+      "step": 2021
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3556948571428572e-05,
+      "loss": 1.5245,
+      "step": 2022
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.355668e-05,
+      "loss": 1.7973,
+      "step": 2023
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3556411428571429e-05,
+      "loss": 1.547,
+      "step": 2024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3556142857142857e-05,
+      "loss": 1.8578,
+      "step": 2025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3555874285714287e-05,
+      "loss": 1.6159,
+      "step": 2026
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3555605714285715e-05,
+      "loss": 1.8087,
+      "step": 2027
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3555337142857143e-05,
+      "loss": 1.8348,
+      "step": 2028
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3555068571428571e-05,
+      "loss": 1.8181,
+      "step": 2029
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3554800000000001e-05,
+      "loss": 1.8542,
+      "step": 2030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.355453142857143e-05,
+      "loss": 1.8773,
+      "step": 2031
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3554262857142857e-05,
+      "loss": 1.811,
+      "step": 2032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3553994285714287e-05,
+      "loss": 1.7504,
+      "step": 2033
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3553725714285714e-05,
+      "loss": 1.772,
+      "step": 2034
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3553457142857144e-05,
+      "loss": 1.4439,
+      "step": 2035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3553188571428572e-05,
+      "loss": 1.4082,
+      "step": 2036
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3552920000000002e-05,
+      "loss": 1.6378,
+      "step": 2037
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3552651428571428e-05,
+      "loss": 1.5729,
+      "step": 2038
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3552382857142858e-05,
+      "loss": 1.539,
+      "step": 2039
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3552114285714286e-05,
+      "loss": 1.6778,
+      "step": 2040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3551845714285714e-05,
+      "loss": 1.5558,
+      "step": 2041
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3551577142857143e-05,
+      "loss": 1.7582,
+      "step": 2042
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3551308571428572e-05,
+      "loss": 1.4285,
+      "step": 2043
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.355104e-05,
+      "loss": 1.7349,
+      "step": 2044
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3550771428571429e-05,
+      "loss": 1.3931,
+      "step": 2045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3550502857142859e-05,
+      "loss": 1.7018,
+      "step": 2046
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3550234285714285e-05,
+      "loss": 1.627,
+      "step": 2047
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3549965714285715e-05,
+      "loss": 1.7709,
+      "step": 2048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3549697142857143e-05,
+      "loss": 1.6225,
+      "step": 2049
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3549428571428573e-05,
+      "loss": 2.0351,
+      "step": 2050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.354916e-05,
+      "loss": 1.5032,
+      "step": 2051
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.354889142857143e-05,
+      "loss": 1.5426,
+      "step": 2052
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3548622857142858e-05,
+      "loss": 1.6446,
+      "step": 2053
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3548354285714286e-05,
+      "loss": 1.5394,
+      "step": 2054
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3548085714285714e-05,
+      "loss": 1.6399,
+      "step": 2055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3547817142857144e-05,
+      "loss": 1.9059,
+      "step": 2056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3547548571428572e-05,
+      "loss": 1.7581,
+      "step": 2057
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.354728e-05,
+      "loss": 1.7275,
+      "step": 2058
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.354701142857143e-05,
+      "loss": 1.8859,
+      "step": 2059
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3546742857142858e-05,
+      "loss": 1.7802,
+      "step": 2060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3546474285714286e-05,
+      "loss": 1.6173,
+      "step": 2061
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3546205714285715e-05,
+      "loss": 1.5676,
+      "step": 2062
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3545937142857144e-05,
+      "loss": 1.7755,
+      "step": 2063
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3545668571428571e-05,
+      "loss": 1.663,
+      "step": 2064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.35454e-05,
+      "loss": 2.0342,
+      "step": 2065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3545131428571429e-05,
+      "loss": 1.7277,
+      "step": 2066
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3544862857142857e-05,
+      "loss": 1.6469,
+      "step": 2067
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3544594285714285e-05,
+      "loss": 1.6722,
+      "step": 2068
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3544325714285715e-05,
+      "loss": 1.4994,
+      "step": 2069
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3544057142857143e-05,
+      "loss": 1.911,
+      "step": 2070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3543788571428572e-05,
+      "loss": 1.8725,
+      "step": 2071
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3543520000000001e-05,
+      "loss": 1.7144,
+      "step": 2072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.354325142857143e-05,
+      "loss": 1.507,
+      "step": 2073
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3542982857142858e-05,
+      "loss": 1.8626,
+      "step": 2074
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3542714285714286e-05,
+      "loss": 1.7798,
+      "step": 2075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3542445714285716e-05,
+      "loss": 1.7263,
+      "step": 2076
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3542177142857142e-05,
+      "loss": 1.7651,
+      "step": 2077
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3541908571428572e-05,
+      "loss": 1.7925,
+      "step": 2078
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.354164e-05,
+      "loss": 1.84,
+      "step": 2079
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3541371428571429e-05,
+      "loss": 1.68,
+      "step": 2080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3541102857142857e-05,
+      "loss": 1.6084,
+      "step": 2081
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3540834285714287e-05,
+      "loss": 2.0921,
+      "step": 2082
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3540565714285715e-05,
+      "loss": 1.4601,
+      "step": 2083
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3540297142857143e-05,
+      "loss": 1.4815,
+      "step": 2084
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3540028571428573e-05,
+      "loss": 1.3688,
+      "step": 2085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3539760000000001e-05,
+      "loss": 1.4446,
+      "step": 2086
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3539491428571429e-05,
+      "loss": 1.5724,
+      "step": 2087
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3539222857142857e-05,
+      "loss": 1.5187,
+      "step": 2088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3538954285714287e-05,
+      "loss": 1.5132,
+      "step": 2089
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3538685714285714e-05,
+      "loss": 1.6748,
+      "step": 2090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3538417142857144e-05,
+      "loss": 1.4001,
+      "step": 2091
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3538148571428572e-05,
+      "loss": 1.68,
+      "step": 2092
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.353788e-05,
+      "loss": 1.6315,
+      "step": 2093
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3537611428571428e-05,
+      "loss": 1.4176,
+      "step": 2094
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3537342857142858e-05,
+      "loss": 1.5579,
+      "step": 2095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3537074285714286e-05,
+      "loss": 1.8938,
+      "step": 2096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3536805714285714e-05,
+      "loss": 1.3975,
+      "step": 2097
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3536537142857144e-05,
+      "loss": 1.9001,
+      "step": 2098
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3536268571428572e-05,
+      "loss": 1.736,
+      "step": 2099
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3536e-05,
+      "loss": 1.5975,
+      "step": 2100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3535731428571429e-05,
+      "loss": 1.7682,
+      "step": 2101
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3535462857142859e-05,
+      "loss": 1.6818,
+      "step": 2102
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3535194285714285e-05,
+      "loss": 1.4896,
+      "step": 2103
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3534925714285715e-05,
+      "loss": 1.8814,
+      "step": 2104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3534657142857143e-05,
+      "loss": 1.9377,
+      "step": 2105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3534388571428571e-05,
+      "loss": 1.878,
+      "step": 2106
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.353412e-05,
+      "loss": 1.6906,
+      "step": 2107
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.353385142857143e-05,
+      "loss": 1.4077,
+      "step": 2108
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3533582857142857e-05,
+      "loss": 1.7587,
+      "step": 2109
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3533314285714286e-05,
+      "loss": 1.9579,
+      "step": 2110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3533045714285716e-05,
+      "loss": 1.8393,
+      "step": 2111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3532777142857144e-05,
+      "loss": 1.8024,
+      "step": 2112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3532508571428572e-05,
+      "loss": 1.6266,
+      "step": 2113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.353224e-05,
+      "loss": 1.3631,
+      "step": 2114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.353197142857143e-05,
+      "loss": 1.8065,
+      "step": 2115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3531702857142856e-05,
+      "loss": 1.9871,
+      "step": 2116
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3531434285714286e-05,
+      "loss": 1.7797,
+      "step": 2117
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3531165714285714e-05,
+      "loss": 1.878,
+      "step": 2118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3530897142857144e-05,
+      "loss": 1.709,
+      "step": 2119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.353062857142857e-05,
+      "loss": 1.4451,
+      "step": 2120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.353036e-05,
+      "loss": 1.692,
+      "step": 2121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3530091428571429e-05,
+      "loss": 1.7193,
+      "step": 2122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3529822857142857e-05,
+      "loss": 1.4217,
+      "step": 2123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3529554285714287e-05,
+      "loss": 1.4687,
+      "step": 2124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3529285714285715e-05,
+      "loss": 1.9302,
+      "step": 2125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3529017142857143e-05,
+      "loss": 1.7604,
+      "step": 2126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3528748571428571e-05,
+      "loss": 1.5941,
+      "step": 2127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3528480000000001e-05,
+      "loss": 1.4641,
+      "step": 2128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3528211428571428e-05,
+      "loss": 1.7795,
+      "step": 2129
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3527942857142858e-05,
+      "loss": 1.6657,
+      "step": 2130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3527674285714286e-05,
+      "loss": 1.6334,
+      "step": 2131
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3527405714285716e-05,
+      "loss": 1.6764,
+      "step": 2132
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3527137142857142e-05,
+      "loss": 1.5197,
+      "step": 2133
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3526868571428572e-05,
+      "loss": 1.9,
+      "step": 2134
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.35266e-05,
+      "loss": 1.648,
+      "step": 2135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3526331428571428e-05,
+      "loss": 1.6651,
+      "step": 2136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3526062857142858e-05,
+      "loss": 1.6355,
+      "step": 2137
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3525794285714286e-05,
+      "loss": 1.5923,
+      "step": 2138
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3525525714285715e-05,
+      "loss": 1.8743,
+      "step": 2139
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3525257142857143e-05,
+      "loss": 1.9201,
+      "step": 2140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3524988571428573e-05,
+      "loss": 1.6219,
+      "step": 2141
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.352472e-05,
+      "loss": 1.5597,
+      "step": 2142
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3524451428571429e-05,
+      "loss": 1.6967,
+      "step": 2143
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3524182857142857e-05,
+      "loss": 1.8543,
+      "step": 2144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3523914285714287e-05,
+      "loss": 1.6477,
+      "step": 2145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3523645714285714e-05,
+      "loss": 1.6449,
+      "step": 2146
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3523377142857143e-05,
+      "loss": 1.7202,
+      "step": 2147
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3523108571428572e-05,
+      "loss": 2.0637,
+      "step": 2148
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.352284e-05,
+      "loss": 1.6174,
+      "step": 2149
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.352257142857143e-05,
+      "loss": 1.9565,
+      "step": 2150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3522302857142858e-05,
+      "loss": 1.7399,
+      "step": 2151
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3522034285714286e-05,
+      "loss": 1.5009,
+      "step": 2152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3521765714285714e-05,
+      "loss": 1.6634,
+      "step": 2153
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3521497142857144e-05,
+      "loss": 1.6452,
+      "step": 2154
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3521228571428572e-05,
+      "loss": 1.5503,
+      "step": 2155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.352096e-05,
+      "loss": 1.6363,
+      "step": 2156
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3520691428571429e-05,
+      "loss": 1.6526,
+      "step": 2157
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3520422857142858e-05,
+      "loss": 1.7344,
+      "step": 2158
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3520154285714285e-05,
+      "loss": 1.7289,
+      "step": 2159
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3519885714285715e-05,
+      "loss": 1.4762,
+      "step": 2160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3519617142857143e-05,
+      "loss": 1.8008,
+      "step": 2161
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3519348571428571e-05,
+      "loss": 1.4701,
+      "step": 2162
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3519080000000001e-05,
+      "loss": 1.5655,
+      "step": 2163
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.351881142857143e-05,
+      "loss": 1.8747,
+      "step": 2164
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3518542857142859e-05,
+      "loss": 1.4825,
+      "step": 2165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3518274285714286e-05,
+      "loss": 1.659,
+      "step": 2166
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3518005714285715e-05,
+      "loss": 1.4089,
+      "step": 2167
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3517737142857144e-05,
+      "loss": 1.8208,
+      "step": 2168
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3517468571428572e-05,
+      "loss": 1.5802,
+      "step": 2169
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.35172e-05,
+      "loss": 1.6957,
+      "step": 2170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.351693142857143e-05,
+      "loss": 1.7385,
+      "step": 2171
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3516662857142856e-05,
+      "loss": 1.7011,
+      "step": 2172
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3516394285714286e-05,
+      "loss": 1.482,
+      "step": 2173
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3516125714285714e-05,
+      "loss": 1.8675,
+      "step": 2174
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3515857142857143e-05,
+      "loss": 1.6192,
+      "step": 2175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3515588571428572e-05,
+      "loss": 1.5585,
+      "step": 2176
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.351532e-05,
+      "loss": 1.4129,
+      "step": 2177
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.351505142857143e-05,
+      "loss": 1.5341,
+      "step": 2178
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3514782857142857e-05,
+      "loss": 1.5827,
+      "step": 2179
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3514514285714287e-05,
+      "loss": 1.5166,
+      "step": 2180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3514245714285715e-05,
+      "loss": 1.4839,
+      "step": 2181
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3513977142857143e-05,
+      "loss": 1.457,
+      "step": 2182
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3513708571428571e-05,
+      "loss": 1.4086,
+      "step": 2183
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3513440000000001e-05,
+      "loss": 1.689,
+      "step": 2184
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3513171428571428e-05,
+      "loss": 1.7285,
+      "step": 2185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3512902857142858e-05,
+      "loss": 1.753,
+      "step": 2186
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3512634285714286e-05,
+      "loss": 1.5925,
+      "step": 2187
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3512365714285716e-05,
+      "loss": 1.5791,
+      "step": 2188
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3512097142857144e-05,
+      "loss": 1.5047,
+      "step": 2189
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3511828571428572e-05,
+      "loss": 1.5737,
+      "step": 2190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3511560000000002e-05,
+      "loss": 1.658,
+      "step": 2191
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3511291428571428e-05,
+      "loss": 1.8955,
+      "step": 2192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3511022857142858e-05,
+      "loss": 1.8333,
+      "step": 2193
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3510754285714286e-05,
+      "loss": 1.759,
+      "step": 2194
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3510485714285715e-05,
+      "loss": 1.6872,
+      "step": 2195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3510217142857143e-05,
+      "loss": 1.8972,
+      "step": 2196
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3509948571428573e-05,
+      "loss": 1.6001,
+      "step": 2197
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3509679999999999e-05,
+      "loss": 1.8118,
+      "step": 2198
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3509411428571429e-05,
+      "loss": 1.773,
+      "step": 2199
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3509142857142857e-05,
+      "loss": 2.0506,
+      "step": 2200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3508874285714287e-05,
+      "loss": 1.51,
+      "step": 2201
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3508605714285715e-05,
+      "loss": 1.7563,
+      "step": 2202
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3508337142857143e-05,
+      "loss": 1.6128,
+      "step": 2203
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3508068571428573e-05,
+      "loss": 1.7896,
+      "step": 2204
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.35078e-05,
+      "loss": 1.7014,
+      "step": 2205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.350753142857143e-05,
+      "loss": 1.8438,
+      "step": 2206
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3507262857142858e-05,
+      "loss": 1.9958,
+      "step": 2207
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3506994285714286e-05,
+      "loss": 1.8245,
+      "step": 2208
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3506725714285714e-05,
+      "loss": 1.9898,
+      "step": 2209
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3506457142857144e-05,
+      "loss": 1.688,
+      "step": 2210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3506188571428572e-05,
+      "loss": 1.7744,
+      "step": 2211
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.350592e-05,
+      "loss": 1.75,
+      "step": 2212
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3505651428571428e-05,
+      "loss": 1.5115,
+      "step": 2213
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3505382857142858e-05,
+      "loss": 1.7296,
+      "step": 2214
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3505114285714287e-05,
+      "loss": 1.987,
+      "step": 2215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3504845714285715e-05,
+      "loss": 1.6063,
+      "step": 2216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3504577142857145e-05,
+      "loss": 1.809,
+      "step": 2217
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3504308571428571e-05,
+      "loss": 1.7671,
+      "step": 2218
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3504040000000001e-05,
+      "loss": 1.5635,
+      "step": 2219
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3503771428571429e-05,
+      "loss": 1.5743,
+      "step": 2220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3503502857142857e-05,
+      "loss": 1.9418,
+      "step": 2221
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3503234285714285e-05,
+      "loss": 1.6436,
+      "step": 2222
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3502965714285715e-05,
+      "loss": 1.8638,
+      "step": 2223
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3502697142857143e-05,
+      "loss": 1.8562,
+      "step": 2224
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3502428571428572e-05,
+      "loss": 1.4979,
+      "step": 2225
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.350216e-05,
+      "loss": 1.6243,
+      "step": 2226
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.350189142857143e-05,
+      "loss": 1.8097,
+      "step": 2227
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3501622857142858e-05,
+      "loss": 1.6896,
+      "step": 2228
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3501354285714286e-05,
+      "loss": 1.5105,
+      "step": 2229
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3501085714285716e-05,
+      "loss": 1.4988,
+      "step": 2230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3500817142857142e-05,
+      "loss": 1.7201,
+      "step": 2231
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3500548571428572e-05,
+      "loss": 1.6071,
+      "step": 2232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.350028e-05,
+      "loss": 1.6932,
+      "step": 2233
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.350001142857143e-05,
+      "loss": 2.0402,
+      "step": 2234
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3499742857142857e-05,
+      "loss": 1.4897,
+      "step": 2235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3499474285714287e-05,
+      "loss": 1.619,
+      "step": 2236
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3499205714285715e-05,
+      "loss": 1.8206,
+      "step": 2237
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3498937142857143e-05,
+      "loss": 1.8333,
+      "step": 2238
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3498668571428571e-05,
+      "loss": 1.9075,
+      "step": 2239
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3498400000000001e-05,
+      "loss": 1.4443,
+      "step": 2240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.349813142857143e-05,
+      "loss": 1.735,
+      "step": 2241
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3497862857142857e-05,
+      "loss": 1.6559,
+      "step": 2242
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3497594285714287e-05,
+      "loss": 1.6937,
+      "step": 2243
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3497325714285714e-05,
+      "loss": 1.6468,
+      "step": 2244
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3497057142857144e-05,
+      "loss": 1.8077,
+      "step": 2245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3496788571428572e-05,
+      "loss": 1.5821,
+      "step": 2246
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3496520000000002e-05,
+      "loss": 1.7255,
+      "step": 2247
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3496251428571428e-05,
+      "loss": 1.6308,
+      "step": 2248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3495982857142858e-05,
+      "loss": 1.7555,
+      "step": 2249
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3495714285714286e-05,
+      "loss": 1.5497,
+      "step": 2250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3495445714285714e-05,
+      "loss": 1.5763,
+      "step": 2251
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3495177142857143e-05,
+      "loss": 1.7112,
+      "step": 2252
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3494908571428572e-05,
+      "loss": 1.4655,
+      "step": 2253
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.349464e-05,
+      "loss": 1.4822,
+      "step": 2254
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3494371428571429e-05,
+      "loss": 1.5947,
+      "step": 2255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3494102857142859e-05,
+      "loss": 1.7061,
+      "step": 2256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3493834285714285e-05,
+      "loss": 1.7279,
+      "step": 2257
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3493565714285715e-05,
+      "loss": 1.8042,
+      "step": 2258
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3493297142857143e-05,
+      "loss": 1.6483,
+      "step": 2259
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3493028571428573e-05,
+      "loss": 1.9422,
+      "step": 2260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.349276e-05,
+      "loss": 1.7565,
+      "step": 2261
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.349249142857143e-05,
+      "loss": 1.543,
+      "step": 2262
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3492222857142858e-05,
+      "loss": 1.5299,
+      "step": 2263
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3491954285714286e-05,
+      "loss": 1.5315,
+      "step": 2264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3491685714285714e-05,
+      "loss": 1.4756,
+      "step": 2265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3491417142857144e-05,
+      "loss": 1.6444,
+      "step": 2266
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3491148571428572e-05,
+      "loss": 1.6412,
+      "step": 2267
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.349088e-05,
+      "loss": 1.9154,
+      "step": 2268
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.349061142857143e-05,
+      "loss": 1.6557,
+      "step": 2269
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3490342857142858e-05,
+      "loss": 1.6551,
+      "step": 2270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3490074285714286e-05,
+      "loss": 1.7183,
+      "step": 2271
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3489805714285715e-05,
+      "loss": 1.6479,
+      "step": 2272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3489537142857144e-05,
+      "loss": 2.2474,
+      "step": 2273
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3489268571428571e-05,
+      "loss": 1.98,
+      "step": 2274
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3489e-05,
+      "loss": 1.6315,
+      "step": 2275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3488731428571429e-05,
+      "loss": 1.7629,
+      "step": 2276
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3488462857142857e-05,
+      "loss": 1.4752,
+      "step": 2277
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3488194285714285e-05,
+      "loss": 1.5656,
+      "step": 2278
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3487925714285715e-05,
+      "loss": 1.5891,
+      "step": 2279
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3487657142857143e-05,
+      "loss": 1.5914,
+      "step": 2280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3487388571428572e-05,
+      "loss": 1.8971,
+      "step": 2281
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3487120000000001e-05,
+      "loss": 1.8081,
+      "step": 2282
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.348685142857143e-05,
+      "loss": 1.7581,
+      "step": 2283
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3486582857142858e-05,
+      "loss": 1.5702,
+      "step": 2284
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3486314285714286e-05,
+      "loss": 1.6526,
+      "step": 2285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3486045714285716e-05,
+      "loss": 1.888,
+      "step": 2286
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3485777142857142e-05,
+      "loss": 1.5295,
+      "step": 2287
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3485508571428572e-05,
+      "loss": 1.5644,
+      "step": 2288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.348524e-05,
+      "loss": 1.3205,
+      "step": 2289
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3484971428571429e-05,
+      "loss": 1.6805,
+      "step": 2290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3484702857142857e-05,
+      "loss": 1.5869,
+      "step": 2291
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3484434285714287e-05,
+      "loss": 1.4465,
+      "step": 2292
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3484165714285715e-05,
+      "loss": 1.6444,
+      "step": 2293
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3483897142857143e-05,
+      "loss": 1.4674,
+      "step": 2294
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3483628571428573e-05,
+      "loss": 1.8288,
+      "step": 2295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3483360000000001e-05,
+      "loss": 1.6418,
+      "step": 2296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3483091428571429e-05,
+      "loss": 1.283,
+      "step": 2297
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3482822857142857e-05,
+      "loss": 2.0016,
+      "step": 2298
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3482554285714287e-05,
+      "loss": 1.7865,
+      "step": 2299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3482285714285714e-05,
+      "loss": 1.7996,
+      "step": 2300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3482017142857144e-05,
+      "loss": 1.701,
+      "step": 2301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3481748571428572e-05,
+      "loss": 1.6773,
+      "step": 2302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.348148e-05,
+      "loss": 1.71,
+      "step": 2303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3481211428571428e-05,
+      "loss": 1.9137,
+      "step": 2304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3480942857142858e-05,
+      "loss": 1.9466,
+      "step": 2305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3480674285714286e-05,
+      "loss": 1.4377,
+      "step": 2306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3480405714285714e-05,
+      "loss": 1.9698,
+      "step": 2307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3480137142857144e-05,
+      "loss": 1.9263,
+      "step": 2308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3479868571428572e-05,
+      "loss": 1.7106,
+      "step": 2309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.34796e-05,
+      "loss": 1.5046,
+      "step": 2310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3479331428571429e-05,
+      "loss": 1.6728,
+      "step": 2311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3479062857142859e-05,
+      "loss": 1.8358,
+      "step": 2312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3478794285714285e-05,
+      "loss": 1.5562,
+      "step": 2313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3478525714285715e-05,
+      "loss": 1.4695,
+      "step": 2314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3478257142857143e-05,
+      "loss": 1.8159,
+      "step": 2315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3477988571428573e-05,
+      "loss": 1.9261,
+      "step": 2316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.347772e-05,
+      "loss": 1.7899,
+      "step": 2317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.347745142857143e-05,
+      "loss": 1.6761,
+      "step": 2318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3477182857142857e-05,
+      "loss": 1.7028,
+      "step": 2319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3476914285714286e-05,
+      "loss": 1.9135,
+      "step": 2320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3476645714285716e-05,
+      "loss": 1.4988,
+      "step": 2321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3476377142857144e-05,
+      "loss": 1.9752,
+      "step": 2322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3476108571428572e-05,
+      "loss": 1.6598,
+      "step": 2323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.347584e-05,
+      "loss": 1.8461,
+      "step": 2324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.347557142857143e-05,
+      "loss": 1.7834,
+      "step": 2325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3475302857142856e-05,
+      "loss": 1.6371,
+      "step": 2326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3475034285714286e-05,
+      "loss": 1.71,
+      "step": 2327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3474765714285714e-05,
+      "loss": 1.7907,
+      "step": 2328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3474497142857144e-05,
+      "loss": 1.5801,
+      "step": 2329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.347422857142857e-05,
+      "loss": 1.6516,
+      "step": 2330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.347396e-05,
+      "loss": 1.5391,
+      "step": 2331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3473691428571429e-05,
+      "loss": 1.6889,
+      "step": 2332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3473422857142857e-05,
+      "loss": 1.61,
+      "step": 2333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3473154285714287e-05,
+      "loss": 1.7626,
+      "step": 2334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3472885714285715e-05,
+      "loss": 1.6423,
+      "step": 2335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3472617142857143e-05,
+      "loss": 1.7826,
+      "step": 2336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3472348571428571e-05,
+      "loss": 1.5971,
+      "step": 2337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3472080000000001e-05,
+      "loss": 1.8456,
+      "step": 2338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.347181142857143e-05,
+      "loss": 1.518,
+      "step": 2339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3471542857142858e-05,
+      "loss": 1.8759,
+      "step": 2340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3471274285714286e-05,
+      "loss": 1.6586,
+      "step": 2341
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3471005714285716e-05,
+      "loss": 1.6266,
+      "step": 2342
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3470737142857142e-05,
+      "loss": 1.5837,
+      "step": 2343
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3470468571428572e-05,
+      "loss": 1.426,
+      "step": 2344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.34702e-05,
+      "loss": 1.5362,
+      "step": 2345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3469931428571428e-05,
+      "loss": 1.9062,
+      "step": 2346
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3469662857142858e-05,
+      "loss": 1.6991,
+      "step": 2347
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3469394285714286e-05,
+      "loss": 1.8856,
+      "step": 2348
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3469125714285715e-05,
+      "loss": 2.0194,
+      "step": 2349
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3468857142857143e-05,
+      "loss": 1.5201,
+      "step": 2350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3468588571428573e-05,
+      "loss": 1.3733,
+      "step": 2351
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.346832e-05,
+      "loss": 1.4598,
+      "step": 2352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3468051428571429e-05,
+      "loss": 1.7594,
+      "step": 2353
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3467782857142857e-05,
+      "loss": 2.001,
+      "step": 2354
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3467514285714287e-05,
+      "loss": 1.7962,
+      "step": 2355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3467245714285714e-05,
+      "loss": 1.6736,
+      "step": 2356
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3466977142857143e-05,
+      "loss": 1.9038,
+      "step": 2357
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3466708571428572e-05,
+      "loss": 1.9225,
+      "step": 2358
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.346644e-05,
+      "loss": 1.777,
+      "step": 2359
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.346617142857143e-05,
+      "loss": 1.6686,
+      "step": 2360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3465902857142858e-05,
+      "loss": 1.8485,
+      "step": 2361
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3465634285714288e-05,
+      "loss": 1.6667,
+      "step": 2362
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3465365714285714e-05,
+      "loss": 1.5175,
+      "step": 2363
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3465097142857144e-05,
+      "loss": 1.4703,
+      "step": 2364
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3464828571428572e-05,
+      "loss": 1.6892,
+      "step": 2365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.346456e-05,
+      "loss": 1.6407,
+      "step": 2366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3464291428571429e-05,
+      "loss": 1.6485,
+      "step": 2367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3464022857142858e-05,
+      "loss": 1.5701,
+      "step": 2368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3463754285714285e-05,
+      "loss": 1.74,
+      "step": 2369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3463485714285715e-05,
+      "loss": 1.8456,
+      "step": 2370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3463217142857143e-05,
+      "loss": 1.5618,
+      "step": 2371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3462948571428571e-05,
+      "loss": 1.7661,
+      "step": 2372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3462680000000001e-05,
+      "loss": 1.5368,
+      "step": 2373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.346241142857143e-05,
+      "loss": 1.774,
+      "step": 2374
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3462142857142859e-05,
+      "loss": 1.4463,
+      "step": 2375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3461874285714286e-05,
+      "loss": 1.631,
+      "step": 2376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3461605714285715e-05,
+      "loss": 1.4866,
+      "step": 2377
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3461337142857144e-05,
+      "loss": 1.325,
+      "step": 2378
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3461068571428572e-05,
+      "loss": 1.706,
+      "step": 2379
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.34608e-05,
+      "loss": 1.863,
+      "step": 2380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.346053142857143e-05,
+      "loss": 1.6726,
+      "step": 2381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3460262857142856e-05,
+      "loss": 1.7711,
+      "step": 2382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3459994285714286e-05,
+      "loss": 1.5788,
+      "step": 2383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3459725714285714e-05,
+      "loss": 1.5284,
+      "step": 2384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3459457142857144e-05,
+      "loss": 1.6464,
+      "step": 2385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3459188571428572e-05,
+      "loss": 1.4851,
+      "step": 2386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.345892e-05,
+      "loss": 1.6613,
+      "step": 2387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.345865142857143e-05,
+      "loss": 1.6627,
+      "step": 2388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3458382857142857e-05,
+      "loss": 1.5174,
+      "step": 2389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3458114285714287e-05,
+      "loss": 1.4958,
+      "step": 2390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3457845714285715e-05,
+      "loss": 1.6535,
+      "step": 2391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3457577142857143e-05,
+      "loss": 1.7388,
+      "step": 2392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3457308571428571e-05,
+      "loss": 1.4856,
+      "step": 2393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3457040000000001e-05,
+      "loss": 1.9399,
+      "step": 2394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3456771428571428e-05,
+      "loss": 1.5987,
+      "step": 2395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3456502857142858e-05,
+      "loss": 1.5352,
+      "step": 2396
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3456234285714286e-05,
+      "loss": 1.504,
+      "step": 2397
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3455965714285716e-05,
+      "loss": 1.7725,
+      "step": 2398
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3455697142857144e-05,
+      "loss": 1.3496,
+      "step": 2399
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3455428571428572e-05,
+      "loss": 1.7355,
+      "step": 2400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3455160000000002e-05,
+      "loss": 1.7902,
+      "step": 2401
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3454891428571428e-05,
+      "loss": 1.8523,
+      "step": 2402
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3454622857142858e-05,
+      "loss": 1.6489,
+      "step": 2403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3454354285714286e-05,
+      "loss": 1.4527,
+      "step": 2404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3454085714285715e-05,
+      "loss": 1.8476,
+      "step": 2405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3453817142857143e-05,
+      "loss": 1.4492,
+      "step": 2406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3453548571428573e-05,
+      "loss": 2.1398,
+      "step": 2407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.345328e-05,
+      "loss": 1.7478,
+      "step": 2408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3453011428571429e-05,
+      "loss": 1.6413,
+      "step": 2409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3452742857142857e-05,
+      "loss": 1.7562,
+      "step": 2410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3452474285714287e-05,
+      "loss": 1.8063,
+      "step": 2411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3452205714285715e-05,
+      "loss": 1.7293,
+      "step": 2412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3451937142857143e-05,
+      "loss": 1.6816,
+      "step": 2413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3451668571428573e-05,
+      "loss": 1.7402,
+      "step": 2414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.34514e-05,
+      "loss": 1.5678,
+      "step": 2415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.345113142857143e-05,
+      "loss": 1.591,
+      "step": 2416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3450862857142858e-05,
+      "loss": 1.8527,
+      "step": 2417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3450594285714286e-05,
+      "loss": 1.317,
+      "step": 2418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3450325714285714e-05,
+      "loss": 1.8377,
+      "step": 2419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3450057142857144e-05,
+      "loss": 1.7235,
+      "step": 2420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3449788571428572e-05,
+      "loss": 1.5794,
+      "step": 2421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.344952e-05,
+      "loss": 1.7678,
+      "step": 2422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3449251428571428e-05,
+      "loss": 1.787,
+      "step": 2423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3448982857142858e-05,
+      "loss": 1.421,
+      "step": 2424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3448714285714286e-05,
+      "loss": 1.7174,
+      "step": 2425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3448445714285715e-05,
+      "loss": 1.9671,
+      "step": 2426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3448177142857145e-05,
+      "loss": 1.6533,
+      "step": 2427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3447908571428571e-05,
+      "loss": 1.179,
+      "step": 2428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3447640000000001e-05,
+      "loss": 1.9928,
+      "step": 2429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3447371428571429e-05,
+      "loss": 1.6336,
+      "step": 2430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3447102857142857e-05,
+      "loss": 1.3872,
+      "step": 2431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3446834285714285e-05,
+      "loss": 1.8061,
+      "step": 2432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3446565714285715e-05,
+      "loss": 1.5836,
+      "step": 2433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3446297142857143e-05,
+      "loss": 1.8511,
+      "step": 2434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3446028571428572e-05,
+      "loss": 1.7394,
+      "step": 2435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.344576e-05,
+      "loss": 1.7802,
+      "step": 2436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.344549142857143e-05,
+      "loss": 1.61,
+      "step": 2437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3445222857142858e-05,
+      "loss": 1.7893,
+      "step": 2438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3444954285714286e-05,
+      "loss": 1.6859,
+      "step": 2439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3444685714285716e-05,
+      "loss": 1.5864,
+      "step": 2440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3444417142857142e-05,
+      "loss": 1.4382,
+      "step": 2441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3444148571428572e-05,
+      "loss": 1.8119,
+      "step": 2442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.344388e-05,
+      "loss": 1.5431,
+      "step": 2443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.344361142857143e-05,
+      "loss": 1.5089,
+      "step": 2444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3443342857142857e-05,
+      "loss": 1.7108,
+      "step": 2445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3443074285714287e-05,
+      "loss": 1.7148,
+      "step": 2446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3442805714285715e-05,
+      "loss": 1.6933,
+      "step": 2447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3442537142857143e-05,
+      "loss": 1.8713,
+      "step": 2448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3442268571428571e-05,
+      "loss": 1.4083,
+      "step": 2449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3442000000000001e-05,
+      "loss": 1.7039,
+      "step": 2450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.344173142857143e-05,
+      "loss": 1.6645,
+      "step": 2451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3441462857142857e-05,
+      "loss": 1.6436,
+      "step": 2452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3441194285714287e-05,
+      "loss": 1.6092,
+      "step": 2453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3440925714285714e-05,
+      "loss": 1.4702,
+      "step": 2454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3440657142857144e-05,
+      "loss": 1.8674,
+      "step": 2455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3440388571428572e-05,
+      "loss": 1.8697,
+      "step": 2456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3440120000000002e-05,
+      "loss": 1.6752,
+      "step": 2457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3439851428571428e-05,
+      "loss": 1.8846,
+      "step": 2458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3439582857142858e-05,
+      "loss": 1.4162,
+      "step": 2459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3439314285714286e-05,
+      "loss": 1.7655,
+      "step": 2460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3439045714285714e-05,
+      "loss": 1.9231,
+      "step": 2461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3438777142857143e-05,
+      "loss": 1.5403,
+      "step": 2462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3438508571428572e-05,
+      "loss": 1.506,
+      "step": 2463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.343824e-05,
+      "loss": 1.6663,
+      "step": 2464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3437971428571429e-05,
+      "loss": 1.6899,
+      "step": 2465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3437702857142859e-05,
+      "loss": 1.7237,
+      "step": 2466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3437434285714287e-05,
+      "loss": 1.4322,
+      "step": 2467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3437165714285715e-05,
+      "loss": 1.3831,
+      "step": 2468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3436897142857143e-05,
+      "loss": 1.7515,
+      "step": 2469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3436628571428573e-05,
+      "loss": 1.6967,
+      "step": 2470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.343636e-05,
+      "loss": 1.6311,
+      "step": 2471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.343609142857143e-05,
+      "loss": 1.5153,
+      "step": 2472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3435822857142858e-05,
+      "loss": 1.5266,
+      "step": 2473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3435554285714286e-05,
+      "loss": 1.38,
+      "step": 2474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3435285714285714e-05,
+      "loss": 1.3878,
+      "step": 2475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3435017142857144e-05,
+      "loss": 1.7107,
+      "step": 2476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3434748571428572e-05,
+      "loss": 1.5517,
+      "step": 2477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.343448e-05,
+      "loss": 1.8814,
+      "step": 2478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.343421142857143e-05,
+      "loss": 1.5022,
+      "step": 2479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3433942857142858e-05,
+      "loss": 1.478,
+      "step": 2480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3433674285714286e-05,
+      "loss": 1.2723,
+      "step": 2481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3433405714285715e-05,
+      "loss": 1.739,
+      "step": 2482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3433137142857144e-05,
+      "loss": 1.8378,
+      "step": 2483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3432868571428571e-05,
+      "loss": 1.48,
+      "step": 2484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.34326e-05,
+      "loss": 1.6728,
+      "step": 2485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3432331428571429e-05,
+      "loss": 1.7883,
+      "step": 2486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3432062857142857e-05,
+      "loss": 1.5543,
+      "step": 2487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3431794285714285e-05,
+      "loss": 1.5286,
+      "step": 2488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3431525714285715e-05,
+      "loss": 1.8609,
+      "step": 2489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3431257142857143e-05,
+      "loss": 1.5902,
+      "step": 2490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3430988571428572e-05,
+      "loss": 1.9818,
+      "step": 2491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3430720000000001e-05,
+      "loss": 1.747,
+      "step": 2492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.343045142857143e-05,
+      "loss": 1.5968,
+      "step": 2493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3430182857142858e-05,
+      "loss": 1.7523,
+      "step": 2494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3429914285714286e-05,
+      "loss": 1.6359,
+      "step": 2495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3429645714285716e-05,
+      "loss": 1.7002,
+      "step": 2496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3429377142857142e-05,
+      "loss": 1.4692,
+      "step": 2497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3429108571428572e-05,
+      "loss": 1.4274,
+      "step": 2498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.342884e-05,
+      "loss": 1.7954,
+      "step": 2499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3428571428571429e-05,
+      "loss": 1.7518,
+      "step": 2500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3428302857142857e-05,
+      "loss": 1.3954,
+      "step": 2501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3428034285714287e-05,
+      "loss": 1.71,
+      "step": 2502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3427765714285715e-05,
+      "loss": 1.4763,
+      "step": 2503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3427497142857143e-05,
+      "loss": 1.952,
+      "step": 2504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3427228571428573e-05,
+      "loss": 1.6768,
+      "step": 2505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3426960000000001e-05,
+      "loss": 1.6014,
+      "step": 2506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3426691428571429e-05,
+      "loss": 1.58,
+      "step": 2507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3426422857142857e-05,
+      "loss": 1.5685,
+      "step": 2508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3426154285714287e-05,
+      "loss": 1.5073,
+      "step": 2509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3425885714285714e-05,
+      "loss": 1.7287,
+      "step": 2510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3425617142857144e-05,
+      "loss": 1.3571,
+      "step": 2511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3425348571428572e-05,
+      "loss": 2.033,
+      "step": 2512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3425080000000002e-05,
+      "loss": 1.7286,
+      "step": 2513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3424811428571428e-05,
+      "loss": 1.7188,
+      "step": 2514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3424542857142858e-05,
+      "loss": 1.4826,
+      "step": 2515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3424274285714286e-05,
+      "loss": 1.8422,
+      "step": 2516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3424005714285714e-05,
+      "loss": 1.4427,
+      "step": 2517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3423737142857144e-05,
+      "loss": 1.4845,
+      "step": 2518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3423468571428572e-05,
+      "loss": 1.6554,
+      "step": 2519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.34232e-05,
+      "loss": 1.4259,
+      "step": 2520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3422931428571429e-05,
+      "loss": 1.4976,
+      "step": 2521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3422662857142859e-05,
+      "loss": 1.4723,
+      "step": 2522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3422394285714285e-05,
+      "loss": 1.4539,
+      "step": 2523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3422125714285715e-05,
+      "loss": 1.9463,
+      "step": 2524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3421857142857143e-05,
+      "loss": 1.3315,
+      "step": 2525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3421588571428573e-05,
+      "loss": 1.3967,
+      "step": 2526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.342132e-05,
+      "loss": 1.372,
+      "step": 2527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.342105142857143e-05,
+      "loss": 1.6205,
+      "step": 2528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3420782857142857e-05,
+      "loss": 1.7774,
+      "step": 2529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3420514285714286e-05,
+      "loss": 1.5841,
+      "step": 2530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3420245714285716e-05,
+      "loss": 1.7714,
+      "step": 2531
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3419977142857144e-05,
+      "loss": 1.6225,
+      "step": 2532
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3419708571428572e-05,
+      "loss": 1.6781,
+      "step": 2533
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.341944e-05,
+      "loss": 1.4518,
+      "step": 2534
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.341917142857143e-05,
+      "loss": 1.4836,
+      "step": 2535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3418902857142858e-05,
+      "loss": 1.7601,
+      "step": 2536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3418634285714286e-05,
+      "loss": 2.0256,
+      "step": 2537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3418365714285714e-05,
+      "loss": 1.6063,
+      "step": 2538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3418097142857144e-05,
+      "loss": 1.8379,
+      "step": 2539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.341782857142857e-05,
+      "loss": 1.6306,
+      "step": 2540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.341756e-05,
+      "loss": 1.9229,
+      "step": 2541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3417291428571429e-05,
+      "loss": 1.6249,
+      "step": 2542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3417022857142857e-05,
+      "loss": 1.7046,
+      "step": 2543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3416754285714287e-05,
+      "loss": 1.8222,
+      "step": 2544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3416485714285715e-05,
+      "loss": 1.5591,
+      "step": 2545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3416217142857143e-05,
+      "loss": 1.6969,
+      "step": 2546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3415948571428571e-05,
+      "loss": 1.4812,
+      "step": 2547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3415680000000001e-05,
+      "loss": 1.4633,
+      "step": 2548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.341541142857143e-05,
+      "loss": 1.4923,
+      "step": 2549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3415142857142858e-05,
+      "loss": 1.5898,
+      "step": 2550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3414874285714286e-05,
+      "loss": 1.7971,
+      "step": 2551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3414605714285716e-05,
+      "loss": 1.6571,
+      "step": 2552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3414337142857142e-05,
+      "loss": 1.7328,
+      "step": 2553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3414068571428572e-05,
+      "loss": 1.6277,
+      "step": 2554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.34138e-05,
+      "loss": 1.4558,
+      "step": 2555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3413531428571428e-05,
+      "loss": 1.4741,
+      "step": 2556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3413262857142858e-05,
+      "loss": 1.8035,
+      "step": 2557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3412994285714286e-05,
+      "loss": 1.6624,
+      "step": 2558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3412725714285716e-05,
+      "loss": 1.5067,
+      "step": 2559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3412457142857143e-05,
+      "loss": 1.5257,
+      "step": 2560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3412188571428573e-05,
+      "loss": 1.8939,
+      "step": 2561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.341192e-05,
+      "loss": 1.8425,
+      "step": 2562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3411651428571429e-05,
+      "loss": 1.7567,
+      "step": 2563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3411382857142857e-05,
+      "loss": 1.5345,
+      "step": 2564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3411114285714287e-05,
+      "loss": 1.5619,
+      "step": 2565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3410845714285714e-05,
+      "loss": 1.823,
+      "step": 2566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3410577142857143e-05,
+      "loss": 1.5869,
+      "step": 2567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3410308571428572e-05,
+      "loss": 1.6359,
+      "step": 2568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.341004e-05,
+      "loss": 1.4441,
+      "step": 2569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.340977142857143e-05,
+      "loss": 1.6678,
+      "step": 2570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3409502857142858e-05,
+      "loss": 1.3682,
+      "step": 2571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3409234285714288e-05,
+      "loss": 1.7713,
+      "step": 2572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3408965714285714e-05,
+      "loss": 1.679,
+      "step": 2573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3408697142857144e-05,
+      "loss": 1.3897,
+      "step": 2574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3408428571428572e-05,
+      "loss": 1.8295,
+      "step": 2575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.340816e-05,
+      "loss": 1.8684,
+      "step": 2576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3407891428571429e-05,
+      "loss": 1.8808,
+      "step": 2577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3407622857142858e-05,
+      "loss": 1.9897,
+      "step": 2578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3407354285714285e-05,
+      "loss": 1.8955,
+      "step": 2579
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3407085714285715e-05,
+      "loss": 1.4736,
+      "step": 2580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3406817142857143e-05,
+      "loss": 1.6729,
+      "step": 2581
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3406548571428571e-05,
+      "loss": 1.5762,
+      "step": 2582
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3406280000000001e-05,
+      "loss": 1.8503,
+      "step": 2583
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.340601142857143e-05,
+      "loss": 1.837,
+      "step": 2584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3405742857142859e-05,
+      "loss": 1.7833,
+      "step": 2585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3405474285714286e-05,
+      "loss": 1.788,
+      "step": 2586
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3405205714285715e-05,
+      "loss": 1.5171,
+      "step": 2587
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3404937142857144e-05,
+      "loss": 1.5239,
+      "step": 2588
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3404668571428572e-05,
+      "loss": 1.5091,
+      "step": 2589
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.34044e-05,
+      "loss": 1.5482,
+      "step": 2590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.340413142857143e-05,
+      "loss": 1.5128,
+      "step": 2591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3403862857142856e-05,
+      "loss": 1.8098,
+      "step": 2592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3403594285714286e-05,
+      "loss": 1.4099,
+      "step": 2593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3403325714285714e-05,
+      "loss": 1.8489,
+      "step": 2594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3403057142857144e-05,
+      "loss": 1.5024,
+      "step": 2595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3402788571428572e-05,
+      "loss": 1.5103,
+      "step": 2596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.340252e-05,
+      "loss": 1.791,
+      "step": 2597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.340225142857143e-05,
+      "loss": 1.4489,
+      "step": 2598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3401982857142857e-05,
+      "loss": 1.8699,
+      "step": 2599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3401714285714287e-05,
+      "loss": 1.9358,
+      "step": 2600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3401445714285715e-05,
+      "loss": 1.735,
+      "step": 2601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3401177142857143e-05,
+      "loss": 1.8946,
+      "step": 2602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3400908571428571e-05,
+      "loss": 2.0094,
+      "step": 2603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3400640000000001e-05,
+      "loss": 1.535,
+      "step": 2604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3400371428571428e-05,
+      "loss": 1.4441,
+      "step": 2605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3400102857142858e-05,
+      "loss": 1.4869,
+      "step": 2606
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3399834285714286e-05,
+      "loss": 1.8356,
+      "step": 2607
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3399565714285716e-05,
+      "loss": 1.7774,
+      "step": 2608
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3399297142857144e-05,
+      "loss": 1.5425,
+      "step": 2609
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3399028571428572e-05,
+      "loss": 1.6137,
+      "step": 2610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3398760000000002e-05,
+      "loss": 1.4367,
+      "step": 2611
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3398491428571428e-05,
+      "loss": 1.45,
+      "step": 2612
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3398222857142858e-05,
+      "loss": 1.7462,
+      "step": 2613
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3397954285714286e-05,
+      "loss": 1.6595,
+      "step": 2614
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3397685714285715e-05,
+      "loss": 1.8721,
+      "step": 2615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3397417142857143e-05,
+      "loss": 1.4115,
+      "step": 2616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3397148571428573e-05,
+      "loss": 1.4923,
+      "step": 2617
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.339688e-05,
+      "loss": 1.5849,
+      "step": 2618
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3396611428571429e-05,
+      "loss": 1.872,
+      "step": 2619
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3396342857142857e-05,
+      "loss": 1.3251,
+      "step": 2620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3396074285714287e-05,
+      "loss": 1.7271,
+      "step": 2621
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3395805714285715e-05,
+      "loss": 1.6373,
+      "step": 2622
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3395537142857143e-05,
+      "loss": 1.748,
+      "step": 2623
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3395268571428573e-05,
+      "loss": 1.7301,
+      "step": 2624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3395e-05,
+      "loss": 1.6669,
+      "step": 2625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.339473142857143e-05,
+      "loss": 1.9145,
+      "step": 2626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3394462857142858e-05,
+      "loss": 1.5714,
+      "step": 2627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3394194285714286e-05,
+      "loss": 1.6963,
+      "step": 2628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3393925714285714e-05,
+      "loss": 1.5173,
+      "step": 2629
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3393657142857144e-05,
+      "loss": 1.84,
+      "step": 2630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3393388571428572e-05,
+      "loss": 1.3551,
+      "step": 2631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.339312e-05,
+      "loss": 1.5584,
+      "step": 2632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3392851428571428e-05,
+      "loss": 1.2849,
+      "step": 2633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3392582857142858e-05,
+      "loss": 1.6735,
+      "step": 2634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3392314285714286e-05,
+      "loss": 1.7034,
+      "step": 2635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3392045714285715e-05,
+      "loss": 1.6193,
+      "step": 2636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3391777142857145e-05,
+      "loss": 1.6504,
+      "step": 2637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3391508571428571e-05,
+      "loss": 1.5697,
+      "step": 2638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3391240000000001e-05,
+      "loss": 1.5717,
+      "step": 2639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3390971428571429e-05,
+      "loss": 1.6235,
+      "step": 2640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3390702857142859e-05,
+      "loss": 1.7981,
+      "step": 2641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3390434285714285e-05,
+      "loss": 1.4849,
+      "step": 2642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3390165714285715e-05,
+      "loss": 1.4517,
+      "step": 2643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3389897142857143e-05,
+      "loss": 1.5067,
+      "step": 2644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3389628571428572e-05,
+      "loss": 1.4234,
+      "step": 2645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.338936e-05,
+      "loss": 1.5499,
+      "step": 2646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.338909142857143e-05,
+      "loss": 1.4651,
+      "step": 2647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3388822857142858e-05,
+      "loss": 1.6691,
+      "step": 2648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3388554285714286e-05,
+      "loss": 1.3849,
+      "step": 2649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3388285714285716e-05,
+      "loss": 1.3594,
+      "step": 2650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3388017142857142e-05,
+      "loss": 1.6572,
+      "step": 2651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3387748571428572e-05,
+      "loss": 1.6041,
+      "step": 2652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.338748e-05,
+      "loss": 1.8499,
+      "step": 2653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.338721142857143e-05,
+      "loss": 1.5768,
+      "step": 2654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3386942857142857e-05,
+      "loss": 1.2298,
+      "step": 2655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3386674285714287e-05,
+      "loss": 1.6747,
+      "step": 2656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3386405714285715e-05,
+      "loss": 1.4679,
+      "step": 2657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3386137142857143e-05,
+      "loss": 1.6028,
+      "step": 2658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3385868571428571e-05,
+      "loss": 1.7991,
+      "step": 2659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3385600000000001e-05,
+      "loss": 1.5425,
+      "step": 2660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3385331428571428e-05,
+      "loss": 1.6171,
+      "step": 2661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3385062857142857e-05,
+      "loss": 1.8021,
+      "step": 2662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3384794285714287e-05,
+      "loss": 1.4473,
+      "step": 2663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3384525714285715e-05,
+      "loss": 1.5251,
+      "step": 2664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3384257142857144e-05,
+      "loss": 1.933,
+      "step": 2665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3383988571428572e-05,
+      "loss": 1.6861,
+      "step": 2666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3383720000000002e-05,
+      "loss": 1.6824,
+      "step": 2667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3383451428571428e-05,
+      "loss": 1.4396,
+      "step": 2668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3383182857142858e-05,
+      "loss": 1.4469,
+      "step": 2669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3382914285714286e-05,
+      "loss": 1.6282,
+      "step": 2670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3382645714285714e-05,
+      "loss": 1.1464,
+      "step": 2671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3382377142857143e-05,
+      "loss": 1.5925,
+      "step": 2672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3382108571428572e-05,
+      "loss": 1.8235,
+      "step": 2673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3381839999999999e-05,
+      "loss": 1.501,
+      "step": 2674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3381571428571429e-05,
+      "loss": 1.4449,
+      "step": 2675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3381302857142859e-05,
+      "loss": 1.6444,
+      "step": 2676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3381034285714287e-05,
+      "loss": 1.2735,
+      "step": 2677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3380765714285715e-05,
+      "loss": 1.4571,
+      "step": 2678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3380497142857143e-05,
+      "loss": 1.6287,
+      "step": 2679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3380228571428573e-05,
+      "loss": 1.9011,
+      "step": 2680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.337996e-05,
+      "loss": 1.6103,
+      "step": 2681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.337969142857143e-05,
+      "loss": 1.8879,
+      "step": 2682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3379422857142858e-05,
+      "loss": 1.4774,
+      "step": 2683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3379154285714286e-05,
+      "loss": 1.799,
+      "step": 2684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3378885714285714e-05,
+      "loss": 1.5137,
+      "step": 2685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3378617142857144e-05,
+      "loss": 1.3047,
+      "step": 2686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3378348571428572e-05,
+      "loss": 1.4906,
+      "step": 2687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.337808e-05,
+      "loss": 1.7803,
+      "step": 2688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.337781142857143e-05,
+      "loss": 1.3714,
+      "step": 2689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3377542857142858e-05,
+      "loss": 1.6925,
+      "step": 2690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3377274285714286e-05,
+      "loss": 1.5999,
+      "step": 2691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3377005714285715e-05,
+      "loss": 1.8378,
+      "step": 2692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3376737142857144e-05,
+      "loss": 1.4115,
+      "step": 2693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3376468571428571e-05,
+      "loss": 1.6057,
+      "step": 2694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.33762e-05,
+      "loss": 1.6195,
+      "step": 2695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3375931428571429e-05,
+      "loss": 1.7054,
+      "step": 2696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3375662857142857e-05,
+      "loss": 1.3718,
+      "step": 2697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3375394285714285e-05,
+      "loss": 1.5836,
+      "step": 2698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3375125714285715e-05,
+      "loss": 1.687,
+      "step": 2699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3374857142857143e-05,
+      "loss": 1.5087,
+      "step": 2700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3374588571428572e-05,
+      "loss": 1.6985,
+      "step": 2701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3374320000000001e-05,
+      "loss": 1.4753,
+      "step": 2702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.337405142857143e-05,
+      "loss": 1.8422,
+      "step": 2703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3373782857142858e-05,
+      "loss": 1.8272,
+      "step": 2704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3373514285714286e-05,
+      "loss": 1.7473,
+      "step": 2705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3373245714285716e-05,
+      "loss": 1.6181,
+      "step": 2706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3372977142857142e-05,
+      "loss": 1.5435,
+      "step": 2707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3372708571428572e-05,
+      "loss": 1.5766,
+      "step": 2708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.337244e-05,
+      "loss": 1.6475,
+      "step": 2709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.337217142857143e-05,
+      "loss": 1.5697,
+      "step": 2710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3371902857142857e-05,
+      "loss": 1.7208,
+      "step": 2711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3371634285714287e-05,
+      "loss": 2.0391,
+      "step": 2712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3371365714285715e-05,
+      "loss": 1.8528,
+      "step": 2713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3371097142857143e-05,
+      "loss": 1.7136,
+      "step": 2714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3370828571428573e-05,
+      "loss": 1.5653,
+      "step": 2715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3370560000000001e-05,
+      "loss": 1.4144,
+      "step": 2716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3370291428571429e-05,
+      "loss": 1.7935,
+      "step": 2717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3370022857142857e-05,
+      "loss": 1.4753,
+      "step": 2718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3369754285714287e-05,
+      "loss": 2.0035,
+      "step": 2719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3369485714285714e-05,
+      "loss": 1.5628,
+      "step": 2720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3369217142857144e-05,
+      "loss": 1.2302,
+      "step": 2721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3368948571428572e-05,
+      "loss": 1.6141,
+      "step": 2722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3368680000000002e-05,
+      "loss": 1.8073,
+      "step": 2723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3368411428571428e-05,
+      "loss": 1.8061,
+      "step": 2724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3368142857142858e-05,
+      "loss": 1.6713,
+      "step": 2725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3367874285714286e-05,
+      "loss": 1.8262,
+      "step": 2726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3367605714285714e-05,
+      "loss": 1.731,
+      "step": 2727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3367337142857144e-05,
+      "loss": 1.6048,
+      "step": 2728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3367068571428572e-05,
+      "loss": 1.2786,
+      "step": 2729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.33668e-05,
+      "loss": 1.6897,
+      "step": 2730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3366531428571429e-05,
+      "loss": 1.9313,
+      "step": 2731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3366262857142859e-05,
+      "loss": 1.7404,
+      "step": 2732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3365994285714287e-05,
+      "loss": 1.6807,
+      "step": 2733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3365725714285715e-05,
+      "loss": 1.4582,
+      "step": 2734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3365457142857143e-05,
+      "loss": 1.5427,
+      "step": 2735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3365188571428573e-05,
+      "loss": 1.8101,
+      "step": 2736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.336492e-05,
+      "loss": 1.5317,
+      "step": 2737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.336465142857143e-05,
+      "loss": 1.5231,
+      "step": 2738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3364382857142857e-05,
+      "loss": 1.6639,
+      "step": 2739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3364114285714286e-05,
+      "loss": 1.6273,
+      "step": 2740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3363845714285716e-05,
+      "loss": 1.5962,
+      "step": 2741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3363577142857144e-05,
+      "loss": 1.6038,
+      "step": 2742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3363308571428572e-05,
+      "loss": 1.9698,
+      "step": 2743
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.336304e-05,
+      "loss": 1.4331,
+      "step": 2744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.336277142857143e-05,
+      "loss": 1.5636,
+      "step": 2745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3362502857142858e-05,
+      "loss": 1.652,
+      "step": 2746
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3362234285714286e-05,
+      "loss": 1.2036,
+      "step": 2747
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3361965714285714e-05,
+      "loss": 1.6096,
+      "step": 2748
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3361697142857144e-05,
+      "loss": 1.5865,
+      "step": 2749
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.336142857142857e-05,
+      "loss": 1.5053,
+      "step": 2750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.336116e-05,
+      "loss": 1.7114,
+      "step": 2751
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3360891428571429e-05,
+      "loss": 1.7278,
+      "step": 2752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3360622857142857e-05,
+      "loss": 1.9225,
+      "step": 2753
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3360354285714287e-05,
+      "loss": 1.7923,
+      "step": 2754
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3360085714285715e-05,
+      "loss": 1.3214,
+      "step": 2755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3359817142857143e-05,
+      "loss": 1.8242,
+      "step": 2756
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3359548571428571e-05,
+      "loss": 1.4382,
+      "step": 2757
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3359280000000001e-05,
+      "loss": 1.4819,
+      "step": 2758
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.335901142857143e-05,
+      "loss": 1.81,
+      "step": 2759
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3358742857142858e-05,
+      "loss": 1.4486,
+      "step": 2760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3358474285714286e-05,
+      "loss": 1.4067,
+      "step": 2761
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3358205714285716e-05,
+      "loss": 1.6278,
+      "step": 2762
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3357937142857142e-05,
+      "loss": 1.3068,
+      "step": 2763
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3357668571428572e-05,
+      "loss": 1.379,
+      "step": 2764
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.33574e-05,
+      "loss": 1.6691,
+      "step": 2765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3357131428571428e-05,
+      "loss": 1.6001,
+      "step": 2766
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3356862857142857e-05,
+      "loss": 1.6916,
+      "step": 2767
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3356594285714286e-05,
+      "loss": 1.6502,
+      "step": 2768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3356325714285716e-05,
+      "loss": 1.6397,
+      "step": 2769
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3356057142857143e-05,
+      "loss": 1.4992,
+      "step": 2770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3355788571428573e-05,
+      "loss": 1.5764,
+      "step": 2771
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.335552e-05,
+      "loss": 1.6391,
+      "step": 2772
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3355251428571429e-05,
+      "loss": 1.4819,
+      "step": 2773
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3354982857142857e-05,
+      "loss": 1.7477,
+      "step": 2774
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3354714285714287e-05,
+      "loss": 1.568,
+      "step": 2775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3354445714285714e-05,
+      "loss": 1.5993,
+      "step": 2776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3354177142857143e-05,
+      "loss": 1.4564,
+      "step": 2777
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3353908571428572e-05,
+      "loss": 1.4274,
+      "step": 2778
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.335364e-05,
+      "loss": 1.671,
+      "step": 2779
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3353371428571428e-05,
+      "loss": 1.6339,
+      "step": 2780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3353102857142858e-05,
+      "loss": 1.7992,
+      "step": 2781
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3352834285714288e-05,
+      "loss": 1.6272,
+      "step": 2782
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3352565714285714e-05,
+      "loss": 1.6473,
+      "step": 2783
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3352297142857144e-05,
+      "loss": 1.8648,
+      "step": 2784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3352028571428572e-05,
+      "loss": 1.7794,
+      "step": 2785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.335176e-05,
+      "loss": 1.4307,
+      "step": 2786
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3351491428571429e-05,
+      "loss": 1.4696,
+      "step": 2787
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3351222857142858e-05,
+      "loss": 1.6487,
+      "step": 2788
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3350954285714285e-05,
+      "loss": 1.5651,
+      "step": 2789
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3350685714285715e-05,
+      "loss": 1.5122,
+      "step": 2790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3350417142857143e-05,
+      "loss": 1.4322,
+      "step": 2791
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3350148571428573e-05,
+      "loss": 1.3898,
+      "step": 2792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334988e-05,
+      "loss": 1.5683,
+      "step": 2793
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334961142857143e-05,
+      "loss": 1.8262,
+      "step": 2794
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3349342857142859e-05,
+      "loss": 1.7225,
+      "step": 2795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3349074285714286e-05,
+      "loss": 1.6265,
+      "step": 2796
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3348805714285715e-05,
+      "loss": 1.7204,
+      "step": 2797
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3348537142857144e-05,
+      "loss": 1.5627,
+      "step": 2798
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3348268571428572e-05,
+      "loss": 1.5156,
+      "step": 2799
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3348e-05,
+      "loss": 1.542,
+      "step": 2800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334773142857143e-05,
+      "loss": 1.7727,
+      "step": 2801
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3347462857142856e-05,
+      "loss": 1.5869,
+      "step": 2802
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3347194285714286e-05,
+      "loss": 1.6221,
+      "step": 2803
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3346925714285714e-05,
+      "loss": 1.6013,
+      "step": 2804
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3346657142857144e-05,
+      "loss": 1.9586,
+      "step": 2805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334638857142857e-05,
+      "loss": 1.4226,
+      "step": 2806
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334612e-05,
+      "loss": 1.5682,
+      "step": 2807
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334585142857143e-05,
+      "loss": 1.7827,
+      "step": 2808
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3345582857142857e-05,
+      "loss": 1.6266,
+      "step": 2809
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3345314285714287e-05,
+      "loss": 1.703,
+      "step": 2810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3345045714285715e-05,
+      "loss": 1.6067,
+      "step": 2811
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3344777142857143e-05,
+      "loss": 1.4978,
+      "step": 2812
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3344508571428571e-05,
+      "loss": 2.147,
+      "step": 2813
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3344240000000001e-05,
+      "loss": 1.6463,
+      "step": 2814
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334397142857143e-05,
+      "loss": 1.5052,
+      "step": 2815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3343702857142858e-05,
+      "loss": 1.8496,
+      "step": 2816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3343434285714286e-05,
+      "loss": 1.5466,
+      "step": 2817
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3343165714285716e-05,
+      "loss": 1.5652,
+      "step": 2818
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3342897142857142e-05,
+      "loss": 1.6552,
+      "step": 2819
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3342628571428572e-05,
+      "loss": 1.7578,
+      "step": 2820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3342360000000002e-05,
+      "loss": 1.5544,
+      "step": 2821
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3342091428571428e-05,
+      "loss": 1.7517,
+      "step": 2822
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3341822857142858e-05,
+      "loss": 1.5452,
+      "step": 2823
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3341554285714286e-05,
+      "loss": 1.8788,
+      "step": 2824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3341285714285715e-05,
+      "loss": 1.9284,
+      "step": 2825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3341017142857143e-05,
+      "loss": 1.5334,
+      "step": 2826
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3340748571428573e-05,
+      "loss": 1.6391,
+      "step": 2827
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.334048e-05,
+      "loss": 1.5146,
+      "step": 2828
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3340211428571429e-05,
+      "loss": 1.5008,
+      "step": 2829
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3339942857142857e-05,
+      "loss": 1.6373,
+      "step": 2830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3339674285714287e-05,
+      "loss": 1.3368,
+      "step": 2831
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3339405714285713e-05,
+      "loss": 1.4752,
+      "step": 2832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3339137142857143e-05,
+      "loss": 1.6393,
+      "step": 2833
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3338868571428573e-05,
+      "loss": 1.3087,
+      "step": 2834
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.33386e-05,
+      "loss": 1.498,
+      "step": 2835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.333833142857143e-05,
+      "loss": 1.5413,
+      "step": 2836
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3338062857142858e-05,
+      "loss": 2.1078,
+      "step": 2837
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3337794285714288e-05,
+      "loss": 1.5679,
+      "step": 2838
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3337525714285714e-05,
+      "loss": 1.5685,
+      "step": 2839
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3337257142857144e-05,
+      "loss": 1.6883,
+      "step": 2840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3336988571428572e-05,
+      "loss": 1.3904,
+      "step": 2841
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.333672e-05,
+      "loss": 2.1151,
+      "step": 2842
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3336451428571428e-05,
+      "loss": 1.2144,
+      "step": 2843
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3336182857142858e-05,
+      "loss": 1.8005,
+      "step": 2844
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3335914285714285e-05,
+      "loss": 1.4502,
+      "step": 2845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3335645714285715e-05,
+      "loss": 1.8257,
+      "step": 2846
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3335377142857145e-05,
+      "loss": 1.3503,
+      "step": 2847
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3335108571428571e-05,
+      "loss": 1.9321,
+      "step": 2848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3334840000000001e-05,
+      "loss": 1.3935,
+      "step": 2849
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3334571428571429e-05,
+      "loss": 1.6426,
+      "step": 2850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3334302857142859e-05,
+      "loss": 1.4747,
+      "step": 2851
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3334034285714285e-05,
+      "loss": 1.6549,
+      "step": 2852
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3333765714285715e-05,
+      "loss": 1.4792,
+      "step": 2853
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3333497142857143e-05,
+      "loss": 1.6186,
+      "step": 2854
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3333228571428572e-05,
+      "loss": 1.6638,
+      "step": 2855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.333296e-05,
+      "loss": 1.6386,
+      "step": 2856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.333269142857143e-05,
+      "loss": 1.5869,
+      "step": 2857
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3332422857142856e-05,
+      "loss": 1.5358,
+      "step": 2858
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3332154285714286e-05,
+      "loss": 1.9548,
+      "step": 2859
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3331885714285716e-05,
+      "loss": 1.6682,
+      "step": 2860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3331617142857144e-05,
+      "loss": 1.3873,
+      "step": 2861
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3331348571428572e-05,
+      "loss": 1.5283,
+      "step": 2862
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.333108e-05,
+      "loss": 1.5739,
+      "step": 2863
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.333081142857143e-05,
+      "loss": 1.5259,
+      "step": 2864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3330542857142857e-05,
+      "loss": 1.4545,
+      "step": 2865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3330274285714287e-05,
+      "loss": 1.6559,
+      "step": 2866
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3330005714285715e-05,
+      "loss": 1.7741,
+      "step": 2867
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3329737142857143e-05,
+      "loss": 1.4767,
+      "step": 2868
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3329468571428571e-05,
+      "loss": 1.5638,
+      "step": 2869
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3329200000000001e-05,
+      "loss": 1.7749,
+      "step": 2870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3328931428571428e-05,
+      "loss": 1.7438,
+      "step": 2871
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3328662857142857e-05,
+      "loss": 1.7532,
+      "step": 2872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3328394285714286e-05,
+      "loss": 1.5525,
+      "step": 2873
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3328125714285715e-05,
+      "loss": 1.7042,
+      "step": 2874
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3327857142857144e-05,
+      "loss": 1.6054,
+      "step": 2875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3327588571428572e-05,
+      "loss": 1.6092,
+      "step": 2876
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3327320000000002e-05,
+      "loss": 1.7901,
+      "step": 2877
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3327051428571428e-05,
+      "loss": 1.4482,
+      "step": 2878
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3326782857142858e-05,
+      "loss": 1.3316,
+      "step": 2879
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3326514285714286e-05,
+      "loss": 1.6863,
+      "step": 2880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3326245714285714e-05,
+      "loss": 1.619,
+      "step": 2881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3325977142857143e-05,
+      "loss": 1.6115,
+      "step": 2882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3325708571428572e-05,
+      "loss": 1.7329,
+      "step": 2883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.332544e-05,
+      "loss": 1.9704,
+      "step": 2884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3325171428571429e-05,
+      "loss": 1.609,
+      "step": 2885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3324902857142857e-05,
+      "loss": 1.4905,
+      "step": 2886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.3324634285714287e-05,
+      "loss": 1.668,
+      "step": 2887
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3324365714285715e-05,
+      "loss": 1.7102,
+      "step": 2888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3324097142857143e-05,
+      "loss": 1.5746,
+      "step": 2889
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3323828571428573e-05,
+      "loss": 1.457,
+      "step": 2890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.332356e-05,
+      "loss": 1.7605,
+      "step": 2891
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.332329142857143e-05,
+      "loss": 1.3754,
+      "step": 2892
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3323022857142858e-05,
+      "loss": 1.6487,
+      "step": 2893
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3322754285714286e-05,
+      "loss": 1.5352,
+      "step": 2894
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3322485714285714e-05,
+      "loss": 2.0455,
+      "step": 2895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3322217142857144e-05,
+      "loss": 1.9571,
+      "step": 2896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3321948571428572e-05,
+      "loss": 1.6912,
+      "step": 2897
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.332168e-05,
+      "loss": 1.6579,
+      "step": 2898
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3321411428571428e-05,
+      "loss": 1.3629,
+      "step": 2899
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3321142857142858e-05,
+      "loss": 1.4976,
+      "step": 2900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3320874285714286e-05,
+      "loss": 1.5976,
+      "step": 2901
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3320605714285715e-05,
+      "loss": 1.7154,
+      "step": 2902
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3320337142857144e-05,
+      "loss": 1.7487,
+      "step": 2903
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3320068571428571e-05,
+      "loss": 1.5096,
+      "step": 2904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.33198e-05,
+      "loss": 1.6666,
+      "step": 2905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3319531428571429e-05,
+      "loss": 1.4121,
+      "step": 2906
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3319262857142859e-05,
+      "loss": 1.5666,
+      "step": 2907
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3318994285714285e-05,
+      "loss": 1.5521,
+      "step": 2908
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3318725714285715e-05,
+      "loss": 1.8268,
+      "step": 2909
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3318457142857143e-05,
+      "loss": 1.5747,
+      "step": 2910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3318188571428572e-05,
+      "loss": 1.5752,
+      "step": 2911
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.331792e-05,
+      "loss": 1.9086,
+      "step": 2912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.331765142857143e-05,
+      "loss": 1.5137,
+      "step": 2913
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3317382857142858e-05,
+      "loss": 1.5613,
+      "step": 2914
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3317114285714286e-05,
+      "loss": 1.6368,
+      "step": 2915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3316845714285716e-05,
+      "loss": 1.5103,
+      "step": 2916
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3316577142857142e-05,
+      "loss": 1.7999,
+      "step": 2917
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3316308571428572e-05,
+      "loss": 1.4056,
+      "step": 2918
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.331604e-05,
+      "loss": 1.543,
+      "step": 2919
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.331577142857143e-05,
+      "loss": 1.526,
+      "step": 2920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3315502857142857e-05,
+      "loss": 1.4533,
+      "step": 2921
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3315234285714287e-05,
+      "loss": 1.7657,
+      "step": 2922
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3314965714285715e-05,
+      "loss": 1.5617,
+      "step": 2923
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3314697142857143e-05,
+      "loss": 1.2466,
+      "step": 2924
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3314428571428571e-05,
+      "loss": 1.6266,
+      "step": 2925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3314160000000001e-05,
+      "loss": 1.538,
+      "step": 2926
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3313891428571429e-05,
+      "loss": 1.3845,
+      "step": 2927
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3313622857142857e-05,
+      "loss": 1.4757,
+      "step": 2928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3313354285714287e-05,
+      "loss": 1.4998,
+      "step": 2929
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3313085714285714e-05,
+      "loss": 1.4721,
+      "step": 2930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3312817142857144e-05,
+      "loss": 1.4175,
+      "step": 2931
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3312548571428572e-05,
+      "loss": 1.7042,
+      "step": 2932
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3312280000000002e-05,
+      "loss": 1.4076,
+      "step": 2933
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3312011428571428e-05,
+      "loss": 1.5357,
+      "step": 2934
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3311742857142858e-05,
+      "loss": 1.7059,
+      "step": 2935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3311474285714286e-05,
+      "loss": 1.4439,
+      "step": 2936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3311205714285714e-05,
+      "loss": 1.7064,
+      "step": 2937
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3310937142857142e-05,
+      "loss": 1.8759,
+      "step": 2938
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3310668571428572e-05,
+      "loss": 1.3027,
+      "step": 2939
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.33104e-05,
+      "loss": 1.5767,
+      "step": 2940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3310131428571429e-05,
+      "loss": 1.6599,
+      "step": 2941
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3309862857142859e-05,
+      "loss": 1.68,
+      "step": 2942
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3309594285714287e-05,
+      "loss": 1.6172,
+      "step": 2943
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3309325714285715e-05,
+      "loss": 1.5773,
+      "step": 2944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3309057142857143e-05,
+      "loss": 1.6354,
+      "step": 2945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3308788571428573e-05,
+      "loss": 1.8878,
+      "step": 2946
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.330852e-05,
+      "loss": 1.5194,
+      "step": 2947
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.330825142857143e-05,
+      "loss": 1.6932,
+      "step": 2948
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3307982857142857e-05,
+      "loss": 1.9724,
+      "step": 2949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3307714285714286e-05,
+      "loss": 1.5394,
+      "step": 2950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3307445714285714e-05,
+      "loss": 1.7693,
+      "step": 2951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3307177142857144e-05,
+      "loss": 1.8007,
+      "step": 2952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3306908571428572e-05,
+      "loss": 1.6408,
+      "step": 2953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.330664e-05,
+      "loss": 1.5568,
+      "step": 2954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.330637142857143e-05,
+      "loss": 1.7285,
+      "step": 2955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3306102857142858e-05,
+      "loss": 1.7226,
+      "step": 2956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3305834285714286e-05,
+      "loss": 1.617,
+      "step": 2957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3305565714285714e-05,
+      "loss": 1.4291,
+      "step": 2958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3305297142857144e-05,
+      "loss": 1.6943,
+      "step": 2959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.330502857142857e-05,
+      "loss": 1.623,
+      "step": 2960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.330476e-05,
+      "loss": 1.5221,
+      "step": 2961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3304491428571429e-05,
+      "loss": 1.422,
+      "step": 2962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3304222857142857e-05,
+      "loss": 1.6312,
+      "step": 2963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3303954285714285e-05,
+      "loss": 1.9859,
+      "step": 2964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3303685714285715e-05,
+      "loss": 1.455,
+      "step": 2965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3303417142857145e-05,
+      "loss": 1.8209,
+      "step": 2966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3303148571428571e-05,
+      "loss": 1.6536,
+      "step": 2967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3302880000000001e-05,
+      "loss": 1.4753,
+      "step": 2968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.330261142857143e-05,
+      "loss": 1.3171,
+      "step": 2969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3302342857142858e-05,
+      "loss": 1.4946,
+      "step": 2970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3302074285714286e-05,
+      "loss": 1.6423,
+      "step": 2971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3301805714285716e-05,
+      "loss": 1.9683,
+      "step": 2972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3301537142857142e-05,
+      "loss": 1.5693,
+      "step": 2973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3301268571428572e-05,
+      "loss": 1.7196,
+      "step": 2974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3301e-05,
+      "loss": 1.4263,
+      "step": 2975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3300731428571428e-05,
+      "loss": 1.8467,
+      "step": 2976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3300462857142857e-05,
+      "loss": 1.3569,
+      "step": 2977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3300194285714286e-05,
+      "loss": 1.6782,
+      "step": 2978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3299925714285715e-05,
+      "loss": 1.6485,
+      "step": 2979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3299657142857143e-05,
+      "loss": 1.6045,
+      "step": 2980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3299388571428573e-05,
+      "loss": 1.5642,
+      "step": 2981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.329912e-05,
+      "loss": 1.482,
+      "step": 2982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3298851428571429e-05,
+      "loss": 1.8046,
+      "step": 2983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3298582857142857e-05,
+      "loss": 1.5627,
+      "step": 2984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3298314285714287e-05,
+      "loss": 1.4346,
+      "step": 2985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3298045714285714e-05,
+      "loss": 1.3491,
+      "step": 2986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3297777142857143e-05,
+      "loss": 1.8904,
+      "step": 2987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3297508571428572e-05,
+      "loss": 1.7758,
+      "step": 2988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3297240000000001e-05,
+      "loss": 1.545,
+      "step": 2989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3296971428571428e-05,
+      "loss": 1.7627,
+      "step": 2990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3296702857142858e-05,
+      "loss": 1.7245,
+      "step": 2991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3296434285714286e-05,
+      "loss": 1.6524,
+      "step": 2992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3296165714285714e-05,
+      "loss": 1.6119,
+      "step": 2993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3295897142857144e-05,
+      "loss": 1.2304,
+      "step": 2994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3295628571428572e-05,
+      "loss": 1.6081,
+      "step": 2995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.329536e-05,
+      "loss": 1.6887,
+      "step": 2996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3295091428571429e-05,
+      "loss": 1.6114,
+      "step": 2997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3294822857142858e-05,
+      "loss": 1.7368,
+      "step": 2998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3294554285714285e-05,
+      "loss": 1.1692,
+      "step": 2999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3294285714285715e-05,
+      "loss": 2.0234,
+      "step": 3000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3294017142857143e-05,
+      "loss": 1.826,
+      "step": 3001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3293748571428573e-05,
+      "loss": 1.7101,
+      "step": 3002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.329348e-05,
+      "loss": 1.4132,
+      "step": 3003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.329321142857143e-05,
+      "loss": 1.7067,
+      "step": 3004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3292942857142857e-05,
+      "loss": 1.2018,
+      "step": 3005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3292674285714286e-05,
+      "loss": 1.7566,
+      "step": 3006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3292405714285715e-05,
+      "loss": 1.9331,
+      "step": 3007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3292137142857144e-05,
+      "loss": 1.1833,
+      "step": 3008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3291868571428572e-05,
+      "loss": 1.6859,
+      "step": 3009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.32916e-05,
+      "loss": 1.5158,
+      "step": 3010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.329133142857143e-05,
+      "loss": 1.421,
+      "step": 3011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3291062857142858e-05,
+      "loss": 1.6903,
+      "step": 3012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3290794285714286e-05,
+      "loss": 1.5396,
+      "step": 3013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3290525714285714e-05,
+      "loss": 1.6033,
+      "step": 3014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3290257142857144e-05,
+      "loss": 1.6626,
+      "step": 3015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.328998857142857e-05,
+      "loss": 1.3853,
+      "step": 3016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.328972e-05,
+      "loss": 1.6572,
+      "step": 3017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3289451428571429e-05,
+      "loss": 1.5917,
+      "step": 3018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3289182857142857e-05,
+      "loss": 1.6098,
+      "step": 3019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3288914285714287e-05,
+      "loss": 1.544,
+      "step": 3020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3288645714285715e-05,
+      "loss": 1.2518,
+      "step": 3021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3288377142857143e-05,
+      "loss": 1.5166,
+      "step": 3022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3288108571428571e-05,
+      "loss": 1.7711,
+      "step": 3023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3287840000000001e-05,
+      "loss": 1.5289,
+      "step": 3024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.328757142857143e-05,
+      "loss": 1.3527,
+      "step": 3025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3287302857142858e-05,
+      "loss": 1.9803,
+      "step": 3026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3287034285714286e-05,
+      "loss": 1.3592,
+      "step": 3027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3286765714285716e-05,
+      "loss": 1.3432,
+      "step": 3028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3286497142857142e-05,
+      "loss": 1.6339,
+      "step": 3029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3286228571428572e-05,
+      "loss": 1.5654,
+      "step": 3030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.328596e-05,
+      "loss": 1.4347,
+      "step": 3031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3285691428571428e-05,
+      "loss": 1.6519,
+      "step": 3032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3285422857142858e-05,
+      "loss": 1.65,
+      "step": 3033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3285154285714286e-05,
+      "loss": 1.6895,
+      "step": 3034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3284885714285716e-05,
+      "loss": 1.4846,
+      "step": 3035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3284617142857143e-05,
+      "loss": 1.5541,
+      "step": 3036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3284348571428573e-05,
+      "loss": 1.4638,
+      "step": 3037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.328408e-05,
+      "loss": 1.4572,
+      "step": 3038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3283811428571429e-05,
+      "loss": 1.6278,
+      "step": 3039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3283542857142857e-05,
+      "loss": 1.7957,
+      "step": 3040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3283274285714287e-05,
+      "loss": 1.4851,
+      "step": 3041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3283005714285713e-05,
+      "loss": 1.663,
+      "step": 3042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3282737142857143e-05,
+      "loss": 1.3404,
+      "step": 3043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3282468571428571e-05,
+      "loss": 1.4264,
+      "step": 3044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.32822e-05,
+      "loss": 1.8818,
+      "step": 3045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.328193142857143e-05,
+      "loss": 1.533,
+      "step": 3046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3281662857142858e-05,
+      "loss": 1.804,
+      "step": 3047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3281394285714288e-05,
+      "loss": 1.5887,
+      "step": 3048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3281125714285714e-05,
+      "loss": 1.6153,
+      "step": 3049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3280857142857144e-05,
+      "loss": 1.6933,
+      "step": 3050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3280588571428572e-05,
+      "loss": 1.9419,
+      "step": 3051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.328032e-05,
+      "loss": 1.9228,
+      "step": 3052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3280051428571428e-05,
+      "loss": 1.4685,
+      "step": 3053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3279782857142858e-05,
+      "loss": 1.7822,
+      "step": 3054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3279514285714285e-05,
+      "loss": 1.5846,
+      "step": 3055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3279245714285715e-05,
+      "loss": 1.6942,
+      "step": 3056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3278977142857143e-05,
+      "loss": 1.5152,
+      "step": 3057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3278708571428573e-05,
+      "loss": 1.557,
+      "step": 3058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3278440000000001e-05,
+      "loss": 1.3975,
+      "step": 3059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3278171428571429e-05,
+      "loss": 1.6296,
+      "step": 3060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3277902857142859e-05,
+      "loss": 1.6005,
+      "step": 3061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.3277634285714285e-05,
+      "loss": 1.358,
+      "step": 3062
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3277365714285715e-05,
+      "loss": 1.8281,
+      "step": 3063
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3277097142857143e-05,
+      "loss": 1.5532,
+      "step": 3064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3276828571428572e-05,
+      "loss": 1.4712,
+      "step": 3065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.327656e-05,
+      "loss": 1.344,
+      "step": 3066
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.327629142857143e-05,
+      "loss": 1.8151,
+      "step": 3067
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3276022857142856e-05,
+      "loss": 1.4802,
+      "step": 3068
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3275754285714286e-05,
+      "loss": 1.7678,
+      "step": 3069
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3275485714285714e-05,
+      "loss": 1.4603,
+      "step": 3070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3275217142857144e-05,
+      "loss": 1.3557,
+      "step": 3071
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3274948571428572e-05,
+      "loss": 1.2655,
+      "step": 3072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.327468e-05,
+      "loss": 1.1259,
+      "step": 3073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.327441142857143e-05,
+      "loss": 1.4831,
+      "step": 3074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3274142857142857e-05,
+      "loss": 1.6629,
+      "step": 3075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3273874285714287e-05,
+      "loss": 1.7514,
+      "step": 3076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3273605714285715e-05,
+      "loss": 1.5333,
+      "step": 3077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3273337142857143e-05,
+      "loss": 1.6934,
+      "step": 3078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3273068571428571e-05,
+      "loss": 1.417,
+      "step": 3079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3272800000000001e-05,
+      "loss": 1.672,
+      "step": 3080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3272531428571428e-05,
+      "loss": 1.3711,
+      "step": 3081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3272262857142857e-05,
+      "loss": 1.5246,
+      "step": 3082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3271994285714286e-05,
+      "loss": 1.3131,
+      "step": 3083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3271725714285715e-05,
+      "loss": 1.5792,
+      "step": 3084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3271457142857144e-05,
+      "loss": 1.3958,
+      "step": 3085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3271188571428572e-05,
+      "loss": 1.6483,
+      "step": 3086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3270920000000002e-05,
+      "loss": 1.1827,
+      "step": 3087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3270651428571428e-05,
+      "loss": 1.4604,
+      "step": 3088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3270382857142858e-05,
+      "loss": 1.4532,
+      "step": 3089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3270114285714286e-05,
+      "loss": 1.8146,
+      "step": 3090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3269845714285714e-05,
+      "loss": 1.7526,
+      "step": 3091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3269577142857143e-05,
+      "loss": 1.4571,
+      "step": 3092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3269308571428572e-05,
+      "loss": 1.6214,
+      "step": 3093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.326904e-05,
+      "loss": 1.6556,
+      "step": 3094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3268771428571429e-05,
+      "loss": 1.4657,
+      "step": 3095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3268502857142857e-05,
+      "loss": 1.4702,
+      "step": 3096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3268234285714287e-05,
+      "loss": 1.458,
+      "step": 3097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3267965714285715e-05,
+      "loss": 1.5248,
+      "step": 3098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3267697142857143e-05,
+      "loss": 1.6761,
+      "step": 3099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3267428571428573e-05,
+      "loss": 1.7916,
+      "step": 3100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.326716e-05,
+      "loss": 1.6825,
+      "step": 3101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.326689142857143e-05,
+      "loss": 1.3527,
+      "step": 3102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3266622857142858e-05,
+      "loss": 1.6506,
+      "step": 3103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3266354285714286e-05,
+      "loss": 1.6831,
+      "step": 3104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3266085714285714e-05,
+      "loss": 1.5966,
+      "step": 3105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3265817142857144e-05,
+      "loss": 1.4131,
+      "step": 3106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3265548571428572e-05,
+      "loss": 1.8694,
+      "step": 3107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.326528e-05,
+      "loss": 1.5573,
+      "step": 3108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3265011428571428e-05,
+      "loss": 1.4797,
+      "step": 3109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3264742857142858e-05,
+      "loss": 1.381,
+      "step": 3110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3264474285714286e-05,
+      "loss": 1.4188,
+      "step": 3111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3264205714285715e-05,
+      "loss": 1.4593,
+      "step": 3112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3263937142857144e-05,
+      "loss": 1.734,
+      "step": 3113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3263668571428571e-05,
+      "loss": 1.5839,
+      "step": 3114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.32634e-05,
+      "loss": 1.46,
+      "step": 3115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3263131428571429e-05,
+      "loss": 1.7103,
+      "step": 3116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3262862857142859e-05,
+      "loss": 1.5353,
+      "step": 3117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3262594285714285e-05,
+      "loss": 1.5153,
+      "step": 3118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3262325714285715e-05,
+      "loss": 1.4536,
+      "step": 3119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3262057142857143e-05,
+      "loss": 1.5917,
+      "step": 3120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3261788571428572e-05,
+      "loss": 1.5932,
+      "step": 3121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.326152e-05,
+      "loss": 1.6805,
+      "step": 3122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.326125142857143e-05,
+      "loss": 1.6125,
+      "step": 3123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3260982857142858e-05,
+      "loss": 1.446,
+      "step": 3124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3260714285714286e-05,
+      "loss": 1.3307,
+      "step": 3125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3260445714285716e-05,
+      "loss": 1.6183,
+      "step": 3126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3260177142857142e-05,
+      "loss": 1.32,
+      "step": 3127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3259908571428572e-05,
+      "loss": 1.6903,
+      "step": 3128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.325964e-05,
+      "loss": 1.2808,
+      "step": 3129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.325937142857143e-05,
+      "loss": 1.5742,
+      "step": 3130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3259102857142857e-05,
+      "loss": 1.8832,
+      "step": 3131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3258834285714287e-05,
+      "loss": 1.5291,
+      "step": 3132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3258565714285715e-05,
+      "loss": 1.475,
+      "step": 3133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3258297142857143e-05,
+      "loss": 1.3797,
+      "step": 3134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3258028571428571e-05,
+      "loss": 1.7438,
+      "step": 3135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3257760000000001e-05,
+      "loss": 1.8077,
+      "step": 3136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3257491428571429e-05,
+      "loss": 1.3852,
+      "step": 3137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3257222857142857e-05,
+      "loss": 1.5936,
+      "step": 3138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3256954285714287e-05,
+      "loss": 1.8876,
+      "step": 3139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3256685714285715e-05,
+      "loss": 1.3722,
+      "step": 3140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3256417142857144e-05,
+      "loss": 1.6424,
+      "step": 3141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3256148571428572e-05,
+      "loss": 1.4867,
+      "step": 3142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3255880000000002e-05,
+      "loss": 1.6143,
+      "step": 3143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3255611428571428e-05,
+      "loss": 1.6241,
+      "step": 3144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3255342857142858e-05,
+      "loss": 1.8632,
+      "step": 3145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3255074285714286e-05,
+      "loss": 1.6536,
+      "step": 3146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3254805714285714e-05,
+      "loss": 1.6856,
+      "step": 3147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3254537142857142e-05,
+      "loss": 1.675,
+      "step": 3148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3254268571428572e-05,
+      "loss": 1.4297,
+      "step": 3149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3254e-05,
+      "loss": 1.6016,
+      "step": 3150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3253731428571429e-05,
+      "loss": 1.6561,
+      "step": 3151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3253462857142859e-05,
+      "loss": 1.4952,
+      "step": 3152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3253194285714287e-05,
+      "loss": 1.7648,
+      "step": 3153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3252925714285715e-05,
+      "loss": 1.3275,
+      "step": 3154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3252657142857143e-05,
+      "loss": 1.5656,
+      "step": 3155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3252388571428573e-05,
+      "loss": 1.5091,
+      "step": 3156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.325212e-05,
+      "loss": 1.5186,
+      "step": 3157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.325185142857143e-05,
+      "loss": 1.5924,
+      "step": 3158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3251582857142857e-05,
+      "loss": 1.2555,
+      "step": 3159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3251314285714286e-05,
+      "loss": 1.354,
+      "step": 3160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3251045714285714e-05,
+      "loss": 1.4454,
+      "step": 3161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3250777142857144e-05,
+      "loss": 1.5749,
+      "step": 3162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3250508571428572e-05,
+      "loss": 1.8201,
+      "step": 3163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.325024e-05,
+      "loss": 1.52,
+      "step": 3164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.324997142857143e-05,
+      "loss": 1.4862,
+      "step": 3165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3249702857142858e-05,
+      "loss": 1.4768,
+      "step": 3166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3249434285714286e-05,
+      "loss": 1.6381,
+      "step": 3167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3249165714285714e-05,
+      "loss": 1.7309,
+      "step": 3168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3248897142857144e-05,
+      "loss": 1.4992,
+      "step": 3169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.324862857142857e-05,
+      "loss": 1.5683,
+      "step": 3170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.324836e-05,
+      "loss": 1.4049,
+      "step": 3171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3248091428571429e-05,
+      "loss": 1.6537,
+      "step": 3172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3247822857142857e-05,
+      "loss": 1.6546,
+      "step": 3173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3247554285714285e-05,
+      "loss": 1.9073,
+      "step": 3174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3247285714285715e-05,
+      "loss": 1.576,
+      "step": 3175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3247017142857143e-05,
+      "loss": 1.526,
+      "step": 3176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3246748571428571e-05,
+      "loss": 1.6128,
+      "step": 3177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3246480000000001e-05,
+      "loss": 1.0502,
+      "step": 3178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.324621142857143e-05,
+      "loss": 1.7352,
+      "step": 3179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3245942857142858e-05,
+      "loss": 1.5857,
+      "step": 3180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3245674285714286e-05,
+      "loss": 1.3463,
+      "step": 3181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3245405714285716e-05,
+      "loss": 1.7058,
+      "step": 3182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3245137142857142e-05,
+      "loss": 1.3449,
+      "step": 3183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3244868571428572e-05,
+      "loss": 1.6091,
+      "step": 3184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.32446e-05,
+      "loss": 1.6006,
+      "step": 3185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.324433142857143e-05,
+      "loss": 1.4424,
+      "step": 3186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3244062857142857e-05,
+      "loss": 1.2716,
+      "step": 3187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3243794285714286e-05,
+      "loss": 1.3738,
+      "step": 3188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3243525714285715e-05,
+      "loss": 1.4023,
+      "step": 3189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3243257142857143e-05,
+      "loss": 1.4529,
+      "step": 3190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3242988571428573e-05,
+      "loss": 1.9088,
+      "step": 3191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.324272e-05,
+      "loss": 1.7416,
+      "step": 3192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3242451428571429e-05,
+      "loss": 1.8189,
+      "step": 3193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3242182857142857e-05,
+      "loss": 1.7276,
+      "step": 3194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3241914285714287e-05,
+      "loss": 1.669,
+      "step": 3195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3241645714285714e-05,
+      "loss": 1.8909,
+      "step": 3196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3241377142857143e-05,
+      "loss": 1.6636,
+      "step": 3197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3241108571428572e-05,
+      "loss": 1.3683,
+      "step": 3198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3240840000000001e-05,
+      "loss": 1.6909,
+      "step": 3199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3240571428571428e-05,
+      "loss": 1.6701,
+      "step": 3200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3240302857142858e-05,
+      "loss": 1.1792,
+      "step": 3201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3240034285714286e-05,
+      "loss": 1.568,
+      "step": 3202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3239765714285714e-05,
+      "loss": 1.3743,
+      "step": 3203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3239497142857144e-05,
+      "loss": 1.5019,
+      "step": 3204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3239228571428572e-05,
+      "loss": 1.6731,
+      "step": 3205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.323896e-05,
+      "loss": 1.6591,
+      "step": 3206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3238691428571429e-05,
+      "loss": 1.4223,
+      "step": 3207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3238422857142858e-05,
+      "loss": 1.3473,
+      "step": 3208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3238154285714287e-05,
+      "loss": 1.3372,
+      "step": 3209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3237885714285715e-05,
+      "loss": 1.9441,
+      "step": 3210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3237617142857143e-05,
+      "loss": 1.7092,
+      "step": 3211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3237348571428573e-05,
+      "loss": 1.3633,
+      "step": 3212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.323708e-05,
+      "loss": 1.2222,
+      "step": 3213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.323681142857143e-05,
+      "loss": 1.8544,
+      "step": 3214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3236542857142857e-05,
+      "loss": 1.4481,
+      "step": 3215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3236274285714286e-05,
+      "loss": 1.3986,
+      "step": 3216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3236005714285715e-05,
+      "loss": 1.3608,
+      "step": 3217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3235737142857144e-05,
+      "loss": 1.4771,
+      "step": 3218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3235468571428572e-05,
+      "loss": 1.7901,
+      "step": 3219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.32352e-05,
+      "loss": 1.4339,
+      "step": 3220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.323493142857143e-05,
+      "loss": 1.3217,
+      "step": 3221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3234662857142858e-05,
+      "loss": 1.5726,
+      "step": 3222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3234394285714286e-05,
+      "loss": 1.4082,
+      "step": 3223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3234125714285714e-05,
+      "loss": 1.5969,
+      "step": 3224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3233857142857144e-05,
+      "loss": 1.5697,
+      "step": 3225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.323358857142857e-05,
+      "loss": 1.5352,
+      "step": 3226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.323332e-05,
+      "loss": 1.4698,
+      "step": 3227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3233051428571429e-05,
+      "loss": 1.5576,
+      "step": 3228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3232782857142857e-05,
+      "loss": 1.2709,
+      "step": 3229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3232514285714287e-05,
+      "loss": 1.5469,
+      "step": 3230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3232245714285715e-05,
+      "loss": 1.385,
+      "step": 3231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3231977142857145e-05,
+      "loss": 1.5272,
+      "step": 3232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3231708571428571e-05,
+      "loss": 1.4886,
+      "step": 3233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3231440000000001e-05,
+      "loss": 1.6225,
+      "step": 3234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.323117142857143e-05,
+      "loss": 1.4309,
+      "step": 3235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3230902857142858e-05,
+      "loss": 1.8515,
+      "step": 3236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.3230634285714286e-05,
+      "loss": 1.619,
+      "step": 3237
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3230365714285716e-05,
+      "loss": 1.7974,
+      "step": 3238
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3230097142857142e-05,
+      "loss": 1.6332,
+      "step": 3239
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3229828571428572e-05,
+      "loss": 1.4388,
+      "step": 3240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.322956e-05,
+      "loss": 1.5682,
+      "step": 3241
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3229291428571428e-05,
+      "loss": 1.8256,
+      "step": 3242
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3229022857142858e-05,
+      "loss": 1.6962,
+      "step": 3243
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3228754285714286e-05,
+      "loss": 1.553,
+      "step": 3244
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3228485714285716e-05,
+      "loss": 1.5693,
+      "step": 3245
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3228217142857143e-05,
+      "loss": 1.3585,
+      "step": 3246
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3227948571428573e-05,
+      "loss": 1.2559,
+      "step": 3247
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.322768e-05,
+      "loss": 1.8649,
+      "step": 3248
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3227411428571429e-05,
+      "loss": 1.7418,
+      "step": 3249
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3227142857142857e-05,
+      "loss": 1.77,
+      "step": 3250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3226874285714287e-05,
+      "loss": 1.2926,
+      "step": 3251
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3226605714285713e-05,
+      "loss": 1.666,
+      "step": 3252
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3226337142857143e-05,
+      "loss": 1.4896,
+      "step": 3253
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3226068571428571e-05,
+      "loss": 1.4381,
+      "step": 3254
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.32258e-05,
+      "loss": 1.6521,
+      "step": 3255
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.322553142857143e-05,
+      "loss": 1.8554,
+      "step": 3256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3225262857142858e-05,
+      "loss": 1.639,
+      "step": 3257
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3224994285714288e-05,
+      "loss": 1.8375,
+      "step": 3258
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3224725714285714e-05,
+      "loss": 1.6193,
+      "step": 3259
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3224457142857144e-05,
+      "loss": 1.9114,
+      "step": 3260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3224188571428572e-05,
+      "loss": 1.4346,
+      "step": 3261
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.322392e-05,
+      "loss": 1.758,
+      "step": 3262
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3223651428571428e-05,
+      "loss": 1.4867,
+      "step": 3263
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3223382857142858e-05,
+      "loss": 1.6903,
+      "step": 3264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3223114285714285e-05,
+      "loss": 1.5994,
+      "step": 3265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3222845714285715e-05,
+      "loss": 1.7506,
+      "step": 3266
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3222577142857143e-05,
+      "loss": 1.4942,
+      "step": 3267
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3222308571428573e-05,
+      "loss": 1.3247,
+      "step": 3268
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3222040000000001e-05,
+      "loss": 1.4269,
+      "step": 3269
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3221771428571429e-05,
+      "loss": 1.6549,
+      "step": 3270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3221502857142859e-05,
+      "loss": 1.4431,
+      "step": 3271
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3221234285714285e-05,
+      "loss": 1.6104,
+      "step": 3272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3220965714285715e-05,
+      "loss": 1.4376,
+      "step": 3273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3220697142857143e-05,
+      "loss": 1.3939,
+      "step": 3274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3220428571428572e-05,
+      "loss": 1.5173,
+      "step": 3275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.322016e-05,
+      "loss": 1.4494,
+      "step": 3276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.321989142857143e-05,
+      "loss": 1.6499,
+      "step": 3277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3219622857142856e-05,
+      "loss": 1.4778,
+      "step": 3278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3219354285714286e-05,
+      "loss": 1.2695,
+      "step": 3279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3219085714285714e-05,
+      "loss": 1.4279,
+      "step": 3280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3218817142857144e-05,
+      "loss": 1.4755,
+      "step": 3281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3218548571428572e-05,
+      "loss": 1.5533,
+      "step": 3282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.321828e-05,
+      "loss": 1.8319,
+      "step": 3283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.321801142857143e-05,
+      "loss": 1.6499,
+      "step": 3284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3217742857142857e-05,
+      "loss": 1.4567,
+      "step": 3285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3217474285714287e-05,
+      "loss": 1.395,
+      "step": 3286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3217205714285715e-05,
+      "loss": 1.4905,
+      "step": 3287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3216937142857143e-05,
+      "loss": 1.5999,
+      "step": 3288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3216668571428571e-05,
+      "loss": 1.6818,
+      "step": 3289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3216400000000001e-05,
+      "loss": 1.494,
+      "step": 3290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.321613142857143e-05,
+      "loss": 1.4406,
+      "step": 3291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3215862857142857e-05,
+      "loss": 1.9872,
+      "step": 3292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3215594285714286e-05,
+      "loss": 1.4895,
+      "step": 3293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3215325714285715e-05,
+      "loss": 1.6269,
+      "step": 3294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3215057142857144e-05,
+      "loss": 1.2138,
+      "step": 3295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3214788571428572e-05,
+      "loss": 1.7011,
+      "step": 3296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3214520000000002e-05,
+      "loss": 1.5639,
+      "step": 3297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3214251428571428e-05,
+      "loss": 1.2755,
+      "step": 3298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3213982857142858e-05,
+      "loss": 1.3812,
+      "step": 3299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3213714285714286e-05,
+      "loss": 1.6701,
+      "step": 3300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3213445714285714e-05,
+      "loss": 1.6563,
+      "step": 3301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3213177142857143e-05,
+      "loss": 1.2014,
+      "step": 3302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3212908571428572e-05,
+      "loss": 1.3306,
+      "step": 3303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.321264e-05,
+      "loss": 1.5658,
+      "step": 3304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3212371428571429e-05,
+      "loss": 1.1387,
+      "step": 3305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3212102857142857e-05,
+      "loss": 1.5182,
+      "step": 3306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3211834285714287e-05,
+      "loss": 1.7107,
+      "step": 3307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3211565714285715e-05,
+      "loss": 1.7984,
+      "step": 3308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3211297142857143e-05,
+      "loss": 1.7886,
+      "step": 3309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3211028571428573e-05,
+      "loss": 1.4786,
+      "step": 3310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.321076e-05,
+      "loss": 1.4264,
+      "step": 3311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.321049142857143e-05,
+      "loss": 1.6629,
+      "step": 3312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3210222857142858e-05,
+      "loss": 1.487,
+      "step": 3313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3209954285714287e-05,
+      "loss": 1.5823,
+      "step": 3314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3209685714285714e-05,
+      "loss": 1.4001,
+      "step": 3315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3209417142857144e-05,
+      "loss": 1.5676,
+      "step": 3316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3209148571428572e-05,
+      "loss": 1.3613,
+      "step": 3317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.320888e-05,
+      "loss": 1.4463,
+      "step": 3318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3208611428571428e-05,
+      "loss": 1.3957,
+      "step": 3319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3208342857142858e-05,
+      "loss": 1.7806,
+      "step": 3320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3208074285714286e-05,
+      "loss": 1.6624,
+      "step": 3321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3207805714285715e-05,
+      "loss": 1.7248,
+      "step": 3322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3207537142857144e-05,
+      "loss": 1.4357,
+      "step": 3323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3207268571428571e-05,
+      "loss": 1.8314,
+      "step": 3324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3207e-05,
+      "loss": 1.3825,
+      "step": 3325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3206731428571429e-05,
+      "loss": 1.5088,
+      "step": 3326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3206462857142859e-05,
+      "loss": 1.5613,
+      "step": 3327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3206194285714285e-05,
+      "loss": 1.6077,
+      "step": 3328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3205925714285715e-05,
+      "loss": 1.539,
+      "step": 3329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3205657142857143e-05,
+      "loss": 1.4171,
+      "step": 3330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3205388571428572e-05,
+      "loss": 1.5021,
+      "step": 3331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.320512e-05,
+      "loss": 1.976,
+      "step": 3332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.320485142857143e-05,
+      "loss": 1.6725,
+      "step": 3333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3204582857142858e-05,
+      "loss": 1.5011,
+      "step": 3334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3204314285714286e-05,
+      "loss": 1.6669,
+      "step": 3335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3204045714285716e-05,
+      "loss": 1.7592,
+      "step": 3336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3203777142857144e-05,
+      "loss": 1.611,
+      "step": 3337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3203508571428572e-05,
+      "loss": 1.5207,
+      "step": 3338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.320324e-05,
+      "loss": 1.5257,
+      "step": 3339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.320297142857143e-05,
+      "loss": 1.3835,
+      "step": 3340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3202702857142857e-05,
+      "loss": 1.7039,
+      "step": 3341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3202434285714287e-05,
+      "loss": 1.6052,
+      "step": 3342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3202165714285715e-05,
+      "loss": 1.6807,
+      "step": 3343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3201897142857143e-05,
+      "loss": 1.6923,
+      "step": 3344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3201628571428571e-05,
+      "loss": 1.7022,
+      "step": 3345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3201360000000001e-05,
+      "loss": 1.4424,
+      "step": 3346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3201091428571429e-05,
+      "loss": 1.624,
+      "step": 3347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3200822857142857e-05,
+      "loss": 1.5347,
+      "step": 3348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3200554285714287e-05,
+      "loss": 1.5222,
+      "step": 3349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3200285714285715e-05,
+      "loss": 1.476,
+      "step": 3350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3200017142857144e-05,
+      "loss": 1.3179,
+      "step": 3351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3199748571428572e-05,
+      "loss": 1.5618,
+      "step": 3352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3199480000000002e-05,
+      "loss": 1.479,
+      "step": 3353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3199211428571428e-05,
+      "loss": 1.6328,
+      "step": 3354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3198942857142858e-05,
+      "loss": 1.3678,
+      "step": 3355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3198674285714286e-05,
+      "loss": 1.4535,
+      "step": 3356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3198405714285714e-05,
+      "loss": 1.3088,
+      "step": 3357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3198137142857142e-05,
+      "loss": 1.7397,
+      "step": 3358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3197868571428572e-05,
+      "loss": 1.5527,
+      "step": 3359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.31976e-05,
+      "loss": 1.4079,
+      "step": 3360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3197331428571429e-05,
+      "loss": 1.5734,
+      "step": 3361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3197062857142859e-05,
+      "loss": 1.539,
+      "step": 3362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3196794285714287e-05,
+      "loss": 1.2999,
+      "step": 3363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3196525714285715e-05,
+      "loss": 1.4606,
+      "step": 3364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3196257142857143e-05,
+      "loss": 1.6369,
+      "step": 3365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3195988571428573e-05,
+      "loss": 1.5875,
+      "step": 3366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.319572e-05,
+      "loss": 1.6517,
+      "step": 3367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.319545142857143e-05,
+      "loss": 1.8179,
+      "step": 3368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3195182857142857e-05,
+      "loss": 1.9906,
+      "step": 3369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3194914285714286e-05,
+      "loss": 1.6353,
+      "step": 3370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3194645714285714e-05,
+      "loss": 1.5154,
+      "step": 3371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3194377142857144e-05,
+      "loss": 1.316,
+      "step": 3372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3194108571428572e-05,
+      "loss": 1.6464,
+      "step": 3373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.319384e-05,
+      "loss": 1.672,
+      "step": 3374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.319357142857143e-05,
+      "loss": 1.2586,
+      "step": 3375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3193302857142858e-05,
+      "loss": 1.4131,
+      "step": 3376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3193034285714286e-05,
+      "loss": 1.8497,
+      "step": 3377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3192765714285714e-05,
+      "loss": 1.6219,
+      "step": 3378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3192497142857144e-05,
+      "loss": 2.1319,
+      "step": 3379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.319222857142857e-05,
+      "loss": 1.4703,
+      "step": 3380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.319196e-05,
+      "loss": 1.4447,
+      "step": 3381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3191691428571429e-05,
+      "loss": 1.3897,
+      "step": 3382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3191422857142859e-05,
+      "loss": 1.3006,
+      "step": 3383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3191154285714285e-05,
+      "loss": 1.7211,
+      "step": 3384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3190885714285715e-05,
+      "loss": 1.2066,
+      "step": 3385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3190617142857143e-05,
+      "loss": 1.4367,
+      "step": 3386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3190348571428571e-05,
+      "loss": 1.3468,
+      "step": 3387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3190080000000001e-05,
+      "loss": 1.4728,
+      "step": 3388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.318981142857143e-05,
+      "loss": 1.3097,
+      "step": 3389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3189542857142858e-05,
+      "loss": 1.548,
+      "step": 3390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3189274285714286e-05,
+      "loss": 1.653,
+      "step": 3391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3189005714285716e-05,
+      "loss": 1.3593,
+      "step": 3392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3188737142857142e-05,
+      "loss": 1.5443,
+      "step": 3393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3188468571428572e-05,
+      "loss": 1.4186,
+      "step": 3394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.31882e-05,
+      "loss": 1.2268,
+      "step": 3395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.318793142857143e-05,
+      "loss": 1.6646,
+      "step": 3396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3187662857142857e-05,
+      "loss": 1.3457,
+      "step": 3397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3187394285714286e-05,
+      "loss": 1.3095,
+      "step": 3398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3187125714285715e-05,
+      "loss": 1.8471,
+      "step": 3399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3186857142857143e-05,
+      "loss": 1.4292,
+      "step": 3400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3186588571428573e-05,
+      "loss": 1.4446,
+      "step": 3401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.318632e-05,
+      "loss": 1.4623,
+      "step": 3402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3186051428571429e-05,
+      "loss": 1.5652,
+      "step": 3403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3185782857142857e-05,
+      "loss": 1.5665,
+      "step": 3404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3185514285714287e-05,
+      "loss": 1.4208,
+      "step": 3405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3185245714285714e-05,
+      "loss": 1.4373,
+      "step": 3406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3184977142857143e-05,
+      "loss": 1.3511,
+      "step": 3407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3184708571428572e-05,
+      "loss": 1.4804,
+      "step": 3408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3184440000000001e-05,
+      "loss": 1.545,
+      "step": 3409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3184171428571428e-05,
+      "loss": 1.4271,
+      "step": 3410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3183902857142858e-05,
+      "loss": 1.6068,
+      "step": 3411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.3183634285714286e-05,
+      "loss": 1.4843,
+      "step": 3412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3183365714285714e-05,
+      "loss": 1.5972,
+      "step": 3413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3183097142857144e-05,
+      "loss": 1.4863,
+      "step": 3414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3182828571428572e-05,
+      "loss": 1.7098,
+      "step": 3415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.318256e-05,
+      "loss": 1.4093,
+      "step": 3416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3182291428571429e-05,
+      "loss": 1.5094,
+      "step": 3417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3182022857142858e-05,
+      "loss": 1.6086,
+      "step": 3418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3181754285714287e-05,
+      "loss": 1.3015,
+      "step": 3419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3181485714285715e-05,
+      "loss": 1.5158,
+      "step": 3420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3181217142857143e-05,
+      "loss": 1.7866,
+      "step": 3421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3180948571428573e-05,
+      "loss": 1.8119,
+      "step": 3422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.318068e-05,
+      "loss": 1.5856,
+      "step": 3423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.318041142857143e-05,
+      "loss": 1.34,
+      "step": 3424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3180142857142857e-05,
+      "loss": 1.4026,
+      "step": 3425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3179874285714286e-05,
+      "loss": 1.5005,
+      "step": 3426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3179605714285715e-05,
+      "loss": 1.4882,
+      "step": 3427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3179337142857144e-05,
+      "loss": 1.5775,
+      "step": 3428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3179068571428572e-05,
+      "loss": 1.7337,
+      "step": 3429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.31788e-05,
+      "loss": 1.357,
+      "step": 3430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.317853142857143e-05,
+      "loss": 1.5053,
+      "step": 3431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3178262857142858e-05,
+      "loss": 1.7803,
+      "step": 3432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3177994285714286e-05,
+      "loss": 1.3656,
+      "step": 3433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3177725714285714e-05,
+      "loss": 1.7175,
+      "step": 3434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3177457142857144e-05,
+      "loss": 1.5692,
+      "step": 3435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.317718857142857e-05,
+      "loss": 1.908,
+      "step": 3436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.317692e-05,
+      "loss": 1.8291,
+      "step": 3437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3176651428571429e-05,
+      "loss": 1.91,
+      "step": 3438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3176382857142857e-05,
+      "loss": 1.5215,
+      "step": 3439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3176114285714287e-05,
+      "loss": 1.7688,
+      "step": 3440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3175845714285715e-05,
+      "loss": 1.5381,
+      "step": 3441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3175577142857145e-05,
+      "loss": 1.7913,
+      "step": 3442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3175308571428571e-05,
+      "loss": 1.4445,
+      "step": 3443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3175040000000001e-05,
+      "loss": 1.8701,
+      "step": 3444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.317477142857143e-05,
+      "loss": 1.4273,
+      "step": 3445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3174502857142858e-05,
+      "loss": 1.6004,
+      "step": 3446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3174234285714286e-05,
+      "loss": 1.5025,
+      "step": 3447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3173965714285716e-05,
+      "loss": 1.6261,
+      "step": 3448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3173697142857142e-05,
+      "loss": 1.5127,
+      "step": 3449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3173428571428572e-05,
+      "loss": 1.5062,
+      "step": 3450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.317316e-05,
+      "loss": 1.6124,
+      "step": 3451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3172891428571428e-05,
+      "loss": 1.5754,
+      "step": 3452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3172622857142858e-05,
+      "loss": 1.5676,
+      "step": 3453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3172354285714286e-05,
+      "loss": 1.4938,
+      "step": 3454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3172085714285716e-05,
+      "loss": 1.1856,
+      "step": 3455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3171817142857143e-05,
+      "loss": 1.4292,
+      "step": 3456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3171548571428573e-05,
+      "loss": 1.5141,
+      "step": 3457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.317128e-05,
+      "loss": 1.445,
+      "step": 3458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3171011428571429e-05,
+      "loss": 1.4742,
+      "step": 3459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3170742857142857e-05,
+      "loss": 1.6557,
+      "step": 3460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3170474285714287e-05,
+      "loss": 1.2509,
+      "step": 3461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3170205714285713e-05,
+      "loss": 1.3432,
+      "step": 3462
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3169937142857143e-05,
+      "loss": 1.5454,
+      "step": 3463
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3169668571428571e-05,
+      "loss": 1.6976,
+      "step": 3464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3169400000000001e-05,
+      "loss": 1.4631,
+      "step": 3465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.316913142857143e-05,
+      "loss": 1.4309,
+      "step": 3466
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3168862857142858e-05,
+      "loss": 1.4087,
+      "step": 3467
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3168594285714288e-05,
+      "loss": 1.5676,
+      "step": 3468
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3168325714285714e-05,
+      "loss": 1.6047,
+      "step": 3469
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3168057142857144e-05,
+      "loss": 1.4416,
+      "step": 3470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3167788571428572e-05,
+      "loss": 1.8768,
+      "step": 3471
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.316752e-05,
+      "loss": 1.7568,
+      "step": 3472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3167251428571428e-05,
+      "loss": 1.4451,
+      "step": 3473
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3166982857142858e-05,
+      "loss": 1.6459,
+      "step": 3474
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3166714285714285e-05,
+      "loss": 1.8339,
+      "step": 3475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3166445714285715e-05,
+      "loss": 1.3319,
+      "step": 3476
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3166177142857143e-05,
+      "loss": 1.5402,
+      "step": 3477
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3165908571428573e-05,
+      "loss": 1.3269,
+      "step": 3478
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3165640000000001e-05,
+      "loss": 1.4686,
+      "step": 3479
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3165371428571429e-05,
+      "loss": 1.4465,
+      "step": 3480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3165102857142859e-05,
+      "loss": 1.2721,
+      "step": 3481
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3164834285714285e-05,
+      "loss": 1.6115,
+      "step": 3482
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3164565714285715e-05,
+      "loss": 1.4469,
+      "step": 3483
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3164297142857143e-05,
+      "loss": 1.519,
+      "step": 3484
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3164028571428572e-05,
+      "loss": 1.6914,
+      "step": 3485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.316376e-05,
+      "loss": 1.7585,
+      "step": 3486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.316349142857143e-05,
+      "loss": 1.4301,
+      "step": 3487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3163222857142858e-05,
+      "loss": 1.4801,
+      "step": 3488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3162954285714286e-05,
+      "loss": 1.306,
+      "step": 3489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3162685714285714e-05,
+      "loss": 1.8856,
+      "step": 3490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3162417142857144e-05,
+      "loss": 1.64,
+      "step": 3491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3162148571428572e-05,
+      "loss": 1.5039,
+      "step": 3492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.316188e-05,
+      "loss": 1.4026,
+      "step": 3493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.316161142857143e-05,
+      "loss": 1.4184,
+      "step": 3494
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3161342857142857e-05,
+      "loss": 1.7557,
+      "step": 3495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3161074285714287e-05,
+      "loss": 1.7543,
+      "step": 3496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3160805714285715e-05,
+      "loss": 1.7149,
+      "step": 3497
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3160537142857143e-05,
+      "loss": 1.7299,
+      "step": 3498
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3160268571428571e-05,
+      "loss": 1.5163,
+      "step": 3499
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3160000000000001e-05,
+      "loss": 1.5803,
+      "step": 3500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.315973142857143e-05,
+      "loss": 1.319,
+      "step": 3501
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3159462857142857e-05,
+      "loss": 1.9599,
+      "step": 3502
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3159194285714286e-05,
+      "loss": 1.4604,
+      "step": 3503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3158925714285715e-05,
+      "loss": 1.5635,
+      "step": 3504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3158657142857144e-05,
+      "loss": 1.3962,
+      "step": 3505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3158388571428572e-05,
+      "loss": 1.6721,
+      "step": 3506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3158120000000002e-05,
+      "loss": 1.4302,
+      "step": 3507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3157851428571428e-05,
+      "loss": 1.4,
+      "step": 3508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3157582857142858e-05,
+      "loss": 1.4524,
+      "step": 3509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3157314285714286e-05,
+      "loss": 1.5959,
+      "step": 3510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3157045714285716e-05,
+      "loss": 1.3502,
+      "step": 3511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3156777142857143e-05,
+      "loss": 1.8582,
+      "step": 3512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3156508571428572e-05,
+      "loss": 1.5046,
+      "step": 3513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.315624e-05,
+      "loss": 1.8255,
+      "step": 3514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3155971428571429e-05,
+      "loss": 1.6231,
+      "step": 3515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3155702857142857e-05,
+      "loss": 1.3723,
+      "step": 3516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3155434285714287e-05,
+      "loss": 1.5,
+      "step": 3517
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3155165714285715e-05,
+      "loss": 1.4436,
+      "step": 3518
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3154897142857143e-05,
+      "loss": 1.8851,
+      "step": 3519
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3154628571428573e-05,
+      "loss": 1.5163,
+      "step": 3520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.315436e-05,
+      "loss": 1.7992,
+      "step": 3521
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.315409142857143e-05,
+      "loss": 1.2765,
+      "step": 3522
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3153822857142858e-05,
+      "loss": 1.4572,
+      "step": 3523
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3153554285714287e-05,
+      "loss": 1.5728,
+      "step": 3524
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3153285714285714e-05,
+      "loss": 1.7624,
+      "step": 3525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3153017142857144e-05,
+      "loss": 1.6587,
+      "step": 3526
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3152748571428572e-05,
+      "loss": 1.4525,
+      "step": 3527
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.315248e-05,
+      "loss": 1.7867,
+      "step": 3528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3152211428571428e-05,
+      "loss": 1.5872,
+      "step": 3529
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3151942857142858e-05,
+      "loss": 1.6493,
+      "step": 3530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3151674285714286e-05,
+      "loss": 1.6795,
+      "step": 3531
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3151405714285715e-05,
+      "loss": 1.5303,
+      "step": 3532
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3151137142857144e-05,
+      "loss": 1.6657,
+      "step": 3533
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3150868571428573e-05,
+      "loss": 1.7144,
+      "step": 3534
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.31506e-05,
+      "loss": 1.4871,
+      "step": 3535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3150331428571429e-05,
+      "loss": 1.4083,
+      "step": 3536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3150062857142859e-05,
+      "loss": 1.7638,
+      "step": 3537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3149794285714285e-05,
+      "loss": 1.8435,
+      "step": 3538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3149525714285715e-05,
+      "loss": 1.7061,
+      "step": 3539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3149257142857143e-05,
+      "loss": 1.9106,
+      "step": 3540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3148988571428572e-05,
+      "loss": 1.4115,
+      "step": 3541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.314872e-05,
+      "loss": 1.4924,
+      "step": 3542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.314845142857143e-05,
+      "loss": 1.5033,
+      "step": 3543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3148182857142858e-05,
+      "loss": 1.3285,
+      "step": 3544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3147914285714286e-05,
+      "loss": 1.4787,
+      "step": 3545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3147645714285716e-05,
+      "loss": 1.4187,
+      "step": 3546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3147377142857144e-05,
+      "loss": 1.4035,
+      "step": 3547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3147108571428572e-05,
+      "loss": 1.3108,
+      "step": 3548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.314684e-05,
+      "loss": 1.3705,
+      "step": 3549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.314657142857143e-05,
+      "loss": 1.5193,
+      "step": 3550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3146302857142857e-05,
+      "loss": 1.3737,
+      "step": 3551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3146034285714287e-05,
+      "loss": 1.2811,
+      "step": 3552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3145765714285715e-05,
+      "loss": 1.7377,
+      "step": 3553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3145497142857143e-05,
+      "loss": 1.2461,
+      "step": 3554
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3145228571428571e-05,
+      "loss": 1.6918,
+      "step": 3555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3144960000000001e-05,
+      "loss": 1.525,
+      "step": 3556
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3144691428571429e-05,
+      "loss": 1.8051,
+      "step": 3557
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3144422857142857e-05,
+      "loss": 1.4331,
+      "step": 3558
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3144154285714287e-05,
+      "loss": 1.4485,
+      "step": 3559
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3143885714285715e-05,
+      "loss": 1.7161,
+      "step": 3560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3143617142857144e-05,
+      "loss": 1.5422,
+      "step": 3561
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3143348571428572e-05,
+      "loss": 1.5491,
+      "step": 3562
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3143080000000002e-05,
+      "loss": 1.6023,
+      "step": 3563
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3142811428571428e-05,
+      "loss": 1.4612,
+      "step": 3564
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3142542857142858e-05,
+      "loss": 1.7154,
+      "step": 3565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3142274285714286e-05,
+      "loss": 1.2782,
+      "step": 3566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3142005714285714e-05,
+      "loss": 1.5969,
+      "step": 3567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3141737142857142e-05,
+      "loss": 1.6132,
+      "step": 3568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3141468571428572e-05,
+      "loss": 1.4727,
+      "step": 3569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.31412e-05,
+      "loss": 1.5756,
+      "step": 3570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3140931428571429e-05,
+      "loss": 1.5732,
+      "step": 3571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3140662857142859e-05,
+      "loss": 1.3921,
+      "step": 3572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3140394285714287e-05,
+      "loss": 1.4807,
+      "step": 3573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3140125714285715e-05,
+      "loss": 1.5447,
+      "step": 3574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3139857142857143e-05,
+      "loss": 1.6675,
+      "step": 3575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3139588571428573e-05,
+      "loss": 1.6191,
+      "step": 3576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.313932e-05,
+      "loss": 1.5285,
+      "step": 3577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.313905142857143e-05,
+      "loss": 1.4853,
+      "step": 3578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3138782857142857e-05,
+      "loss": 1.5067,
+      "step": 3579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3138514285714286e-05,
+      "loss": 1.4723,
+      "step": 3580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3138245714285714e-05,
+      "loss": 1.3743,
+      "step": 3581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3137977142857144e-05,
+      "loss": 1.6484,
+      "step": 3582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3137708571428572e-05,
+      "loss": 1.4813,
+      "step": 3583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.313744e-05,
+      "loss": 1.5247,
+      "step": 3584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.313717142857143e-05,
+      "loss": 1.5954,
+      "step": 3585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3136902857142858e-05,
+      "loss": 1.5725,
+      "step": 3586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3136634285714286e-05,
+      "loss": 1.5787,
+      "step": 3587
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3136365714285714e-05,
+      "loss": 1.2147,
+      "step": 3588
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3136097142857144e-05,
+      "loss": 1.3815,
+      "step": 3589
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.313582857142857e-05,
+      "loss": 1.3882,
+      "step": 3590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.313556e-05,
+      "loss": 1.6661,
+      "step": 3591
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3135291428571429e-05,
+      "loss": 1.4656,
+      "step": 3592
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3135022857142859e-05,
+      "loss": 1.5636,
+      "step": 3593
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3134754285714285e-05,
+      "loss": 1.5375,
+      "step": 3594
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3134485714285715e-05,
+      "loss": 1.4738,
+      "step": 3595
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3134217142857143e-05,
+      "loss": 1.765,
+      "step": 3596
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3133948571428571e-05,
+      "loss": 1.5906,
+      "step": 3597
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3133680000000001e-05,
+      "loss": 1.9849,
+      "step": 3598
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.313341142857143e-05,
+      "loss": 1.5716,
+      "step": 3599
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3133142857142858e-05,
+      "loss": 1.7824,
+      "step": 3600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3132874285714286e-05,
+      "loss": 1.1908,
+      "step": 3601
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3132605714285716e-05,
+      "loss": 1.4537,
+      "step": 3602
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3132337142857142e-05,
+      "loss": 1.9325,
+      "step": 3603
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3132068571428572e-05,
+      "loss": 1.4101,
+      "step": 3604
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.31318e-05,
+      "loss": 1.5652,
+      "step": 3605
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.313153142857143e-05,
+      "loss": 1.6374,
+      "step": 3606
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3131262857142857e-05,
+      "loss": 1.477,
+      "step": 3607
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3130994285714286e-05,
+      "loss": 1.3854,
+      "step": 3608
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3130725714285715e-05,
+      "loss": 1.5448,
+      "step": 3609
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3130457142857143e-05,
+      "loss": 1.4683,
+      "step": 3610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3130188571428573e-05,
+      "loss": 1.5123,
+      "step": 3611
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.312992e-05,
+      "loss": 1.8483,
+      "step": 3612
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3129651428571429e-05,
+      "loss": 1.435,
+      "step": 3613
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3129382857142857e-05,
+      "loss": 1.5509,
+      "step": 3614
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3129114285714287e-05,
+      "loss": 1.497,
+      "step": 3615
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3128845714285715e-05,
+      "loss": 1.6605,
+      "step": 3616
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3128577142857143e-05,
+      "loss": 1.2734,
+      "step": 3617
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3128308571428572e-05,
+      "loss": 1.3832,
+      "step": 3618
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3128040000000001e-05,
+      "loss": 1.6312,
+      "step": 3619
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3127771428571428e-05,
+      "loss": 1.6829,
+      "step": 3620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3127502857142858e-05,
+      "loss": 1.6245,
+      "step": 3621
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3127234285714286e-05,
+      "loss": 1.2728,
+      "step": 3622
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3126965714285714e-05,
+      "loss": 1.6229,
+      "step": 3623
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3126697142857144e-05,
+      "loss": 1.6212,
+      "step": 3624
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3126428571428572e-05,
+      "loss": 1.5313,
+      "step": 3625
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.312616e-05,
+      "loss": 1.535,
+      "step": 3626
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3125891428571429e-05,
+      "loss": 1.4467,
+      "step": 3627
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3125622857142858e-05,
+      "loss": 1.4351,
+      "step": 3628
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3125354285714287e-05,
+      "loss": 1.5432,
+      "step": 3629
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3125085714285715e-05,
+      "loss": 1.4616,
+      "step": 3630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3124817142857143e-05,
+      "loss": 1.4897,
+      "step": 3631
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3124548571428573e-05,
+      "loss": 1.6909,
+      "step": 3632
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.312428e-05,
+      "loss": 1.2294,
+      "step": 3633
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.312401142857143e-05,
+      "loss": 1.5209,
+      "step": 3634
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3123742857142857e-05,
+      "loss": 1.7465,
+      "step": 3635
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3123474285714286e-05,
+      "loss": 1.5636,
+      "step": 3636
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3123205714285715e-05,
+      "loss": 1.634,
+      "step": 3637
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3122937142857144e-05,
+      "loss": 1.6991,
+      "step": 3638
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3122668571428573e-05,
+      "loss": 1.7618,
+      "step": 3639
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.31224e-05,
+      "loss": 1.246,
+      "step": 3640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.312213142857143e-05,
+      "loss": 1.7381,
+      "step": 3641
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3121862857142858e-05,
+      "loss": 1.485,
+      "step": 3642
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3121594285714286e-05,
+      "loss": 1.4353,
+      "step": 3643
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3121325714285714e-05,
+      "loss": 1.7538,
+      "step": 3644
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3121057142857144e-05,
+      "loss": 1.4478,
+      "step": 3645
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.312078857142857e-05,
+      "loss": 1.9169,
+      "step": 3646
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.312052e-05,
+      "loss": 1.6022,
+      "step": 3647
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3120251428571429e-05,
+      "loss": 1.6723,
+      "step": 3648
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3119982857142857e-05,
+      "loss": 1.4795,
+      "step": 3649
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3119714285714287e-05,
+      "loss": 1.5482,
+      "step": 3650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3119445714285715e-05,
+      "loss": 1.692,
+      "step": 3651
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3119177142857145e-05,
+      "loss": 1.5376,
+      "step": 3652
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3118908571428571e-05,
+      "loss": 1.3857,
+      "step": 3653
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3118640000000001e-05,
+      "loss": 1.7744,
+      "step": 3654
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.311837142857143e-05,
+      "loss": 1.5312,
+      "step": 3655
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3118102857142858e-05,
+      "loss": 1.3157,
+      "step": 3656
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3117834285714286e-05,
+      "loss": 1.5383,
+      "step": 3657
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3117565714285716e-05,
+      "loss": 1.6265,
+      "step": 3658
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3117297142857142e-05,
+      "loss": 1.7085,
+      "step": 3659
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3117028571428572e-05,
+      "loss": 1.2425,
+      "step": 3660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.311676e-05,
+      "loss": 1.3142,
+      "step": 3661
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.311649142857143e-05,
+      "loss": 1.4897,
+      "step": 3662
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3116222857142858e-05,
+      "loss": 1.7596,
+      "step": 3663
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3115954285714286e-05,
+      "loss": 1.3675,
+      "step": 3664
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3115685714285716e-05,
+      "loss": 1.3236,
+      "step": 3665
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3115417142857143e-05,
+      "loss": 1.7811,
+      "step": 3666
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3115148571428573e-05,
+      "loss": 1.4307,
+      "step": 3667
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.311488e-05,
+      "loss": 1.4412,
+      "step": 3668
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3114611428571429e-05,
+      "loss": 1.472,
+      "step": 3669
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3114342857142857e-05,
+      "loss": 1.743,
+      "step": 3670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3114074285714287e-05,
+      "loss": 1.6075,
+      "step": 3671
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3113805714285713e-05,
+      "loss": 1.0537,
+      "step": 3672
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3113537142857143e-05,
+      "loss": 1.7536,
+      "step": 3673
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3113268571428571e-05,
+      "loss": 1.6271,
+      "step": 3674
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3113000000000001e-05,
+      "loss": 1.4164,
+      "step": 3675
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3112731428571428e-05,
+      "loss": 1.6342,
+      "step": 3676
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3112462857142858e-05,
+      "loss": 1.7984,
+      "step": 3677
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3112194285714288e-05,
+      "loss": 1.588,
+      "step": 3678
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3111925714285714e-05,
+      "loss": 1.1383,
+      "step": 3679
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3111657142857144e-05,
+      "loss": 1.2842,
+      "step": 3680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3111388571428572e-05,
+      "loss": 1.6444,
+      "step": 3681
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.311112e-05,
+      "loss": 1.4955,
+      "step": 3682
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3110851428571428e-05,
+      "loss": 1.8042,
+      "step": 3683
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3110582857142858e-05,
+      "loss": 1.5447,
+      "step": 3684
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3110314285714286e-05,
+      "loss": 1.5349,
+      "step": 3685
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3110045714285715e-05,
+      "loss": 1.7089,
+      "step": 3686
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3109777142857143e-05,
+      "loss": 1.2248,
+      "step": 3687
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3109508571428573e-05,
+      "loss": 1.2673,
+      "step": 3688
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.310924e-05,
+      "loss": 1.5333,
+      "step": 3689
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3108971428571429e-05,
+      "loss": 1.3871,
+      "step": 3690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3108702857142859e-05,
+      "loss": 1.5425,
+      "step": 3691
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3108434285714285e-05,
+      "loss": 1.4919,
+      "step": 3692
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3108165714285715e-05,
+      "loss": 1.3241,
+      "step": 3693
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3107897142857143e-05,
+      "loss": 1.3859,
+      "step": 3694
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3107628571428572e-05,
+      "loss": 1.3422,
+      "step": 3695
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.310736e-05,
+      "loss": 1.4782,
+      "step": 3696
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.310709142857143e-05,
+      "loss": 1.6458,
+      "step": 3697
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3106822857142858e-05,
+      "loss": 1.6785,
+      "step": 3698
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3106554285714286e-05,
+      "loss": 1.1655,
+      "step": 3699
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3106285714285714e-05,
+      "loss": 1.4764,
+      "step": 3700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3106017142857144e-05,
+      "loss": 1.3953,
+      "step": 3701
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.310574857142857e-05,
+      "loss": 1.7339,
+      "step": 3702
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.310548e-05,
+      "loss": 1.5637,
+      "step": 3703
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.310521142857143e-05,
+      "loss": 1.3224,
+      "step": 3704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3104942857142857e-05,
+      "loss": 1.4416,
+      "step": 3705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3104674285714287e-05,
+      "loss": 1.6575,
+      "step": 3706
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3104405714285715e-05,
+      "loss": 1.5171,
+      "step": 3707
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3104137142857145e-05,
+      "loss": 1.6243,
+      "step": 3708
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3103868571428571e-05,
+      "loss": 1.2378,
+      "step": 3709
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3103600000000001e-05,
+      "loss": 1.6216,
+      "step": 3710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.310333142857143e-05,
+      "loss": 1.6084,
+      "step": 3711
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3103062857142857e-05,
+      "loss": 1.6872,
+      "step": 3712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3102794285714286e-05,
+      "loss": 1.819,
+      "step": 3713
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3102525714285715e-05,
+      "loss": 1.6268,
+      "step": 3714
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3102257142857142e-05,
+      "loss": 1.6079,
+      "step": 3715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3101988571428572e-05,
+      "loss": 1.5549,
+      "step": 3716
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3101720000000002e-05,
+      "loss": 1.4899,
+      "step": 3717
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3101451428571428e-05,
+      "loss": 1.4336,
+      "step": 3718
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3101182857142858e-05,
+      "loss": 1.525,
+      "step": 3719
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3100914285714286e-05,
+      "loss": 1.6642,
+      "step": 3720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3100645714285716e-05,
+      "loss": 1.5995,
+      "step": 3721
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3100377142857143e-05,
+      "loss": 1.5555,
+      "step": 3722
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3100108571428572e-05,
+      "loss": 1.5341,
+      "step": 3723
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.309984e-05,
+      "loss": 1.7247,
+      "step": 3724
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3099571428571429e-05,
+      "loss": 1.4123,
+      "step": 3725
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3099302857142857e-05,
+      "loss": 1.6886,
+      "step": 3726
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3099034285714287e-05,
+      "loss": 1.8511,
+      "step": 3727
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3098765714285713e-05,
+      "loss": 1.3984,
+      "step": 3728
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3098497142857143e-05,
+      "loss": 1.6904,
+      "step": 3729
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3098228571428573e-05,
+      "loss": 1.711,
+      "step": 3730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3097960000000001e-05,
+      "loss": 1.8956,
+      "step": 3731
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.309769142857143e-05,
+      "loss": 1.4546,
+      "step": 3732
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3097422857142858e-05,
+      "loss": 1.3426,
+      "step": 3733
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3097154285714287e-05,
+      "loss": 1.4832,
+      "step": 3734
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3096885714285714e-05,
+      "loss": 1.9373,
+      "step": 3735
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3096617142857144e-05,
+      "loss": 1.4158,
+      "step": 3736
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3096348571428572e-05,
+      "loss": 1.501,
+      "step": 3737
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.309608e-05,
+      "loss": 1.4133,
+      "step": 3738
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3095811428571428e-05,
+      "loss": 1.8626,
+      "step": 3739
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3095542857142858e-05,
+      "loss": 1.5082,
+      "step": 3740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3095274285714285e-05,
+      "loss": 2.0045,
+      "step": 3741
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3095005714285715e-05,
+      "loss": 1.2816,
+      "step": 3742
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3094737142857144e-05,
+      "loss": 1.9285,
+      "step": 3743
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3094468571428573e-05,
+      "loss": 1.7162,
+      "step": 3744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.30942e-05,
+      "loss": 1.737,
+      "step": 3745
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3093931428571429e-05,
+      "loss": 1.2,
+      "step": 3746
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3093662857142859e-05,
+      "loss": 1.8734,
+      "step": 3747
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3093394285714285e-05,
+      "loss": 1.6856,
+      "step": 3748
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3093125714285715e-05,
+      "loss": 1.7338,
+      "step": 3749
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3092857142857143e-05,
+      "loss": 1.7483,
+      "step": 3750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3092588571428572e-05,
+      "loss": 1.6312,
+      "step": 3751
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.309232e-05,
+      "loss": 1.5838,
+      "step": 3752
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.309205142857143e-05,
+      "loss": 1.6105,
+      "step": 3753
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3091782857142856e-05,
+      "loss": 1.5659,
+      "step": 3754
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3091514285714286e-05,
+      "loss": 1.4779,
+      "step": 3755
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3091245714285716e-05,
+      "loss": 1.5233,
+      "step": 3756
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3090977142857144e-05,
+      "loss": 1.5604,
+      "step": 3757
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3090708571428572e-05,
+      "loss": 1.629,
+      "step": 3758
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.309044e-05,
+      "loss": 1.7903,
+      "step": 3759
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.309017142857143e-05,
+      "loss": 1.72,
+      "step": 3760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3089902857142857e-05,
+      "loss": 1.7991,
+      "step": 3761
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.3089634285714287e-05,
+      "loss": 1.3413,
+      "step": 3762
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3089365714285715e-05,
+      "loss": 1.5071,
+      "step": 3763
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3089097142857143e-05,
+      "loss": 1.5845,
+      "step": 3764
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3088828571428571e-05,
+      "loss": 1.7407,
+      "step": 3765
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3088560000000001e-05,
+      "loss": 1.6016,
+      "step": 3766
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3088291428571429e-05,
+      "loss": 1.5934,
+      "step": 3767
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3088022857142857e-05,
+      "loss": 1.5202,
+      "step": 3768
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3087754285714287e-05,
+      "loss": 1.7878,
+      "step": 3769
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3087485714285715e-05,
+      "loss": 1.7032,
+      "step": 3770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3087217142857144e-05,
+      "loss": 1.9245,
+      "step": 3771
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3086948571428572e-05,
+      "loss": 1.3912,
+      "step": 3772
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3086680000000002e-05,
+      "loss": 1.5787,
+      "step": 3773
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3086411428571428e-05,
+      "loss": 1.1642,
+      "step": 3774
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3086142857142858e-05,
+      "loss": 1.4998,
+      "step": 3775
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3085874285714286e-05,
+      "loss": 1.5282,
+      "step": 3776
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3085605714285714e-05,
+      "loss": 1.8443,
+      "step": 3777
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3085337142857142e-05,
+      "loss": 1.6187,
+      "step": 3778
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3085068571428572e-05,
+      "loss": 1.6835,
+      "step": 3779
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.30848e-05,
+      "loss": 1.6343,
+      "step": 3780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3084531428571429e-05,
+      "loss": 1.8123,
+      "step": 3781
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3084262857142857e-05,
+      "loss": 1.779,
+      "step": 3782
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3083994285714287e-05,
+      "loss": 1.7144,
+      "step": 3783
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3083725714285715e-05,
+      "loss": 1.9026,
+      "step": 3784
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3083457142857143e-05,
+      "loss": 1.3573,
+      "step": 3785
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3083188571428573e-05,
+      "loss": 1.5983,
+      "step": 3786
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.308292e-05,
+      "loss": 1.1373,
+      "step": 3787
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.308265142857143e-05,
+      "loss": 1.4906,
+      "step": 3788
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3082382857142857e-05,
+      "loss": 1.5901,
+      "step": 3789
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3082114285714287e-05,
+      "loss": 1.4375,
+      "step": 3790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3081845714285714e-05,
+      "loss": 1.3654,
+      "step": 3791
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3081577142857144e-05,
+      "loss": 1.4675,
+      "step": 3792
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3081308571428572e-05,
+      "loss": 1.6253,
+      "step": 3793
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.308104e-05,
+      "loss": 1.0451,
+      "step": 3794
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3080771428571428e-05,
+      "loss": 1.6413,
+      "step": 3795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3080502857142858e-05,
+      "loss": 1.531,
+      "step": 3796
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3080234285714286e-05,
+      "loss": 1.6671,
+      "step": 3797
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3079965714285714e-05,
+      "loss": 1.5759,
+      "step": 3798
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3079697142857144e-05,
+      "loss": 1.4251,
+      "step": 3799
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.307942857142857e-05,
+      "loss": 1.3089,
+      "step": 3800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.307916e-05,
+      "loss": 1.8513,
+      "step": 3801
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3078891428571429e-05,
+      "loss": 1.3747,
+      "step": 3802
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3078622857142859e-05,
+      "loss": 1.2476,
+      "step": 3803
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3078354285714285e-05,
+      "loss": 1.5802,
+      "step": 3804
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3078085714285715e-05,
+      "loss": 1.2556,
+      "step": 3805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3077817142857143e-05,
+      "loss": 1.7352,
+      "step": 3806
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3077548571428571e-05,
+      "loss": 1.6747,
+      "step": 3807
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.307728e-05,
+      "loss": 1.3548,
+      "step": 3808
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.307701142857143e-05,
+      "loss": 1.2631,
+      "step": 3809
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3076742857142858e-05,
+      "loss": 1.9047,
+      "step": 3810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3076474285714286e-05,
+      "loss": 1.5974,
+      "step": 3811
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3076205714285716e-05,
+      "loss": 1.6233,
+      "step": 3812
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3075937142857144e-05,
+      "loss": 1.6701,
+      "step": 3813
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3075668571428572e-05,
+      "loss": 1.2834,
+      "step": 3814
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.30754e-05,
+      "loss": 1.7689,
+      "step": 3815
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.307513142857143e-05,
+      "loss": 1.719,
+      "step": 3816
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3074862857142857e-05,
+      "loss": 1.2847,
+      "step": 3817
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3074594285714286e-05,
+      "loss": 1.5537,
+      "step": 3818
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3074325714285715e-05,
+      "loss": 1.3135,
+      "step": 3819
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3074057142857143e-05,
+      "loss": 1.8694,
+      "step": 3820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3073788571428571e-05,
+      "loss": 1.1233,
+      "step": 3821
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.307352e-05,
+      "loss": 1.6365,
+      "step": 3822
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3073251428571429e-05,
+      "loss": 1.6148,
+      "step": 3823
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3072982857142857e-05,
+      "loss": 1.8208,
+      "step": 3824
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3072714285714287e-05,
+      "loss": 1.7149,
+      "step": 3825
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3072445714285715e-05,
+      "loss": 1.2849,
+      "step": 3826
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3072177142857143e-05,
+      "loss": 1.4818,
+      "step": 3827
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3071908571428572e-05,
+      "loss": 1.4459,
+      "step": 3828
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3071640000000001e-05,
+      "loss": 1.5817,
+      "step": 3829
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3071371428571428e-05,
+      "loss": 1.4718,
+      "step": 3830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3071102857142858e-05,
+      "loss": 1.9448,
+      "step": 3831
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3070834285714286e-05,
+      "loss": 1.3915,
+      "step": 3832
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3070565714285714e-05,
+      "loss": 1.7504,
+      "step": 3833
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3070297142857142e-05,
+      "loss": 1.5255,
+      "step": 3834
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3070028571428572e-05,
+      "loss": 1.914,
+      "step": 3835
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3069760000000002e-05,
+      "loss": 1.6631,
+      "step": 3836
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3069491428571429e-05,
+      "loss": 1.6401,
+      "step": 3837
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3069222857142858e-05,
+      "loss": 1.3302,
+      "step": 3838
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3068954285714287e-05,
+      "loss": 1.6255,
+      "step": 3839
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3068685714285715e-05,
+      "loss": 1.3173,
+      "step": 3840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3068417142857143e-05,
+      "loss": 1.5266,
+      "step": 3841
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3068148571428573e-05,
+      "loss": 1.2862,
+      "step": 3842
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306788e-05,
+      "loss": 1.6976,
+      "step": 3843
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306761142857143e-05,
+      "loss": 1.5423,
+      "step": 3844
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3067342857142857e-05,
+      "loss": 1.9706,
+      "step": 3845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3067074285714286e-05,
+      "loss": 1.4082,
+      "step": 3846
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3066805714285714e-05,
+      "loss": 1.2687,
+      "step": 3847
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3066537142857144e-05,
+      "loss": 1.7646,
+      "step": 3848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3066268571428573e-05,
+      "loss": 1.4447,
+      "step": 3849
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3066e-05,
+      "loss": 1.4122,
+      "step": 3850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306573142857143e-05,
+      "loss": 1.4558,
+      "step": 3851
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3065462857142858e-05,
+      "loss": 1.7273,
+      "step": 3852
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3065194285714286e-05,
+      "loss": 1.3217,
+      "step": 3853
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3064925714285714e-05,
+      "loss": 1.4398,
+      "step": 3854
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3064657142857144e-05,
+      "loss": 1.5143,
+      "step": 3855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306438857142857e-05,
+      "loss": 1.6145,
+      "step": 3856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306412e-05,
+      "loss": 1.2546,
+      "step": 3857
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3063851428571429e-05,
+      "loss": 2.0641,
+      "step": 3858
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3063582857142859e-05,
+      "loss": 1.4939,
+      "step": 3859
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3063314285714285e-05,
+      "loss": 1.5773,
+      "step": 3860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3063045714285715e-05,
+      "loss": 1.5489,
+      "step": 3861
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3062777142857145e-05,
+      "loss": 1.3649,
+      "step": 3862
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3062508571428571e-05,
+      "loss": 1.4234,
+      "step": 3863
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3062240000000001e-05,
+      "loss": 1.4962,
+      "step": 3864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306197142857143e-05,
+      "loss": 1.1107,
+      "step": 3865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3061702857142858e-05,
+      "loss": 1.2306,
+      "step": 3866
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3061434285714286e-05,
+      "loss": 1.4347,
+      "step": 3867
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3061165714285716e-05,
+      "loss": 1.2475,
+      "step": 3868
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3060897142857142e-05,
+      "loss": 1.8069,
+      "step": 3869
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3060628571428572e-05,
+      "loss": 1.5097,
+      "step": 3870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306036e-05,
+      "loss": 1.2383,
+      "step": 3871
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.306009142857143e-05,
+      "loss": 1.6493,
+      "step": 3872
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3059822857142856e-05,
+      "loss": 1.5378,
+      "step": 3873
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3059554285714286e-05,
+      "loss": 1.337,
+      "step": 3874
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3059285714285716e-05,
+      "loss": 1.5438,
+      "step": 3875
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3059017142857143e-05,
+      "loss": 1.6761,
+      "step": 3876
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3058748571428573e-05,
+      "loss": 1.569,
+      "step": 3877
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.305848e-05,
+      "loss": 1.2791,
+      "step": 3878
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3058211428571429e-05,
+      "loss": 1.7175,
+      "step": 3879
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3057942857142857e-05,
+      "loss": 1.1809,
+      "step": 3880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3057674285714287e-05,
+      "loss": 1.8712,
+      "step": 3881
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3057405714285715e-05,
+      "loss": 1.2542,
+      "step": 3882
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3057137142857143e-05,
+      "loss": 1.6787,
+      "step": 3883
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3056868571428571e-05,
+      "loss": 1.5749,
+      "step": 3884
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3056600000000001e-05,
+      "loss": 1.5045,
+      "step": 3885
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3056331428571428e-05,
+      "loss": 1.2756,
+      "step": 3886
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3056062857142858e-05,
+      "loss": 1.5679,
+      "step": 3887
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3055794285714286e-05,
+      "loss": 1.2576,
+      "step": 3888
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3055525714285714e-05,
+      "loss": 1.8345,
+      "step": 3889
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3055257142857144e-05,
+      "loss": 1.838,
+      "step": 3890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3054988571428572e-05,
+      "loss": 1.7856,
+      "step": 3891
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.305472e-05,
+      "loss": 1.4483,
+      "step": 3892
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3054451428571428e-05,
+      "loss": 1.4486,
+      "step": 3893
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3054182857142858e-05,
+      "loss": 1.337,
+      "step": 3894
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3053914285714286e-05,
+      "loss": 1.9062,
+      "step": 3895
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3053645714285715e-05,
+      "loss": 1.5175,
+      "step": 3896
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3053377142857143e-05,
+      "loss": 1.493,
+      "step": 3897
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3053108571428573e-05,
+      "loss": 1.5209,
+      "step": 3898
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.305284e-05,
+      "loss": 1.5844,
+      "step": 3899
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3052571428571429e-05,
+      "loss": 1.7632,
+      "step": 3900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3052302857142857e-05,
+      "loss": 1.3095,
+      "step": 3901
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3052034285714285e-05,
+      "loss": 1.8141,
+      "step": 3902
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3051765714285715e-05,
+      "loss": 1.6191,
+      "step": 3903
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3051497142857143e-05,
+      "loss": 1.3118,
+      "step": 3904
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3051228571428572e-05,
+      "loss": 1.3943,
+      "step": 3905
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.305096e-05,
+      "loss": 1.5955,
+      "step": 3906
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.305069142857143e-05,
+      "loss": 1.7545,
+      "step": 3907
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3050422857142858e-05,
+      "loss": 1.4385,
+      "step": 3908
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3050154285714286e-05,
+      "loss": 1.4287,
+      "step": 3909
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3049885714285714e-05,
+      "loss": 1.4361,
+      "step": 3910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3049617142857144e-05,
+      "loss": 1.5278,
+      "step": 3911
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.304934857142857e-05,
+      "loss": 1.6052,
+      "step": 3912
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.304908e-05,
+      "loss": 1.3772,
+      "step": 3913
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3048811428571429e-05,
+      "loss": 1.4749,
+      "step": 3914
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3048542857142857e-05,
+      "loss": 1.6733,
+      "step": 3915
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3048274285714287e-05,
+      "loss": 1.3523,
+      "step": 3916
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3048005714285715e-05,
+      "loss": 1.558,
+      "step": 3917
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3047737142857145e-05,
+      "loss": 1.4482,
+      "step": 3918
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3047468571428571e-05,
+      "loss": 1.4077,
+      "step": 3919
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3047200000000001e-05,
+      "loss": 1.6463,
+      "step": 3920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.304693142857143e-05,
+      "loss": 1.4048,
+      "step": 3921
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3046662857142857e-05,
+      "loss": 1.1845,
+      "step": 3922
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3046394285714286e-05,
+      "loss": 1.5025,
+      "step": 3923
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3046125714285715e-05,
+      "loss": 1.4569,
+      "step": 3924
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3045857142857142e-05,
+      "loss": 1.4138,
+      "step": 3925
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3045588571428572e-05,
+      "loss": 1.464,
+      "step": 3926
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.304532e-05,
+      "loss": 1.5731,
+      "step": 3927
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3045051428571428e-05,
+      "loss": 1.5247,
+      "step": 3928
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3044782857142858e-05,
+      "loss": 1.6776,
+      "step": 3929
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3044514285714286e-05,
+      "loss": 1.3538,
+      "step": 3930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3044245714285716e-05,
+      "loss": 1.3815,
+      "step": 3931
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3043977142857143e-05,
+      "loss": 1.3369,
+      "step": 3932
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3043708571428572e-05,
+      "loss": 1.404,
+      "step": 3933
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.304344e-05,
+      "loss": 1.4716,
+      "step": 3934
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3043171428571429e-05,
+      "loss": 1.2852,
+      "step": 3935
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3042902857142857e-05,
+      "loss": 1.6974,
+      "step": 3936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.3042634285714287e-05,
+      "loss": 1.3815,
+      "step": 3937
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3042365714285713e-05,
+      "loss": 1.8042,
+      "step": 3938
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3042097142857143e-05,
+      "loss": 1.3684,
+      "step": 3939
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3041828571428571e-05,
+      "loss": 1.4757,
+      "step": 3940
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3041560000000001e-05,
+      "loss": 1.5647,
+      "step": 3941
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.304129142857143e-05,
+      "loss": 1.6417,
+      "step": 3942
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3041022857142858e-05,
+      "loss": 1.3416,
+      "step": 3943
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3040754285714287e-05,
+      "loss": 1.2077,
+      "step": 3944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3040485714285714e-05,
+      "loss": 1.3699,
+      "step": 3945
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3040217142857144e-05,
+      "loss": 1.6391,
+      "step": 3946
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3039948571428572e-05,
+      "loss": 1.3179,
+      "step": 3947
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.303968e-05,
+      "loss": 1.5528,
+      "step": 3948
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3039411428571428e-05,
+      "loss": 1.4249,
+      "step": 3949
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3039142857142858e-05,
+      "loss": 1.4843,
+      "step": 3950
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3038874285714285e-05,
+      "loss": 1.6083,
+      "step": 3951
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3038605714285715e-05,
+      "loss": 1.6098,
+      "step": 3952
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3038337142857143e-05,
+      "loss": 1.6622,
+      "step": 3953
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3038068571428573e-05,
+      "loss": 1.5741,
+      "step": 3954
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.30378e-05,
+      "loss": 1.5065,
+      "step": 3955
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3037531428571429e-05,
+      "loss": 1.5841,
+      "step": 3956
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3037262857142859e-05,
+      "loss": 1.5423,
+      "step": 3957
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3036994285714285e-05,
+      "loss": 1.9617,
+      "step": 3958
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3036725714285715e-05,
+      "loss": 1.6512,
+      "step": 3959
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3036457142857143e-05,
+      "loss": 1.5461,
+      "step": 3960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3036188571428572e-05,
+      "loss": 1.5639,
+      "step": 3961
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.303592e-05,
+      "loss": 1.2349,
+      "step": 3962
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.303565142857143e-05,
+      "loss": 1.5178,
+      "step": 3963
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3035382857142858e-05,
+      "loss": 1.2719,
+      "step": 3964
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3035114285714286e-05,
+      "loss": 1.4827,
+      "step": 3965
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3034845714285714e-05,
+      "loss": 1.4445,
+      "step": 3966
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3034577142857144e-05,
+      "loss": 1.5714,
+      "step": 3967
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3034308571428572e-05,
+      "loss": 1.8449,
+      "step": 3968
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.303404e-05,
+      "loss": 1.3577,
+      "step": 3969
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.303377142857143e-05,
+      "loss": 1.5688,
+      "step": 3970
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3033502857142857e-05,
+      "loss": 1.4165,
+      "step": 3971
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3033234285714287e-05,
+      "loss": 1.6895,
+      "step": 3972
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3032965714285715e-05,
+      "loss": 1.7034,
+      "step": 3973
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3032697142857143e-05,
+      "loss": 1.5013,
+      "step": 3974
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3032428571428571e-05,
+      "loss": 1.7972,
+      "step": 3975
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3032160000000001e-05,
+      "loss": 1.6881,
+      "step": 3976
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3031891428571429e-05,
+      "loss": 1.2657,
+      "step": 3977
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3031622857142857e-05,
+      "loss": 1.7004,
+      "step": 3978
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3031354285714285e-05,
+      "loss": 1.4443,
+      "step": 3979
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3031085714285715e-05,
+      "loss": 1.577,
+      "step": 3980
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3030817142857144e-05,
+      "loss": 1.4572,
+      "step": 3981
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3030548571428572e-05,
+      "loss": 1.4195,
+      "step": 3982
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3030280000000002e-05,
+      "loss": 1.668,
+      "step": 3983
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3030011428571428e-05,
+      "loss": 1.5238,
+      "step": 3984
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3029742857142858e-05,
+      "loss": 1.6659,
+      "step": 3985
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3029474285714286e-05,
+      "loss": 1.6506,
+      "step": 3986
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3029205714285716e-05,
+      "loss": 1.1567,
+      "step": 3987
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3028937142857142e-05,
+      "loss": 1.2158,
+      "step": 3988
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3028668571428572e-05,
+      "loss": 1.8385,
+      "step": 3989
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.30284e-05,
+      "loss": 1.4197,
+      "step": 3990
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3028131428571429e-05,
+      "loss": 1.4514,
+      "step": 3991
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3027862857142857e-05,
+      "loss": 1.5381,
+      "step": 3992
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3027594285714287e-05,
+      "loss": 1.6166,
+      "step": 3993
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3027325714285715e-05,
+      "loss": 1.5815,
+      "step": 3994
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3027057142857143e-05,
+      "loss": 1.3129,
+      "step": 3995
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3026788571428573e-05,
+      "loss": 1.6645,
+      "step": 3996
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.302652e-05,
+      "loss": 1.4861,
+      "step": 3997
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.302625142857143e-05,
+      "loss": 1.4397,
+      "step": 3998
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3025982857142857e-05,
+      "loss": 1.4455,
+      "step": 3999
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3025714285714287e-05,
+      "loss": 1.6555,
+      "step": 4000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3025445714285714e-05,
+      "loss": 1.677,
+      "step": 4001
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3025177142857144e-05,
+      "loss": 1.6187,
+      "step": 4002
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3024908571428572e-05,
+      "loss": 1.8877,
+      "step": 4003
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.302464e-05,
+      "loss": 1.6509,
+      "step": 4004
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3024371428571428e-05,
+      "loss": 1.899,
+      "step": 4005
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3024102857142858e-05,
+      "loss": 1.4859,
+      "step": 4006
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3023834285714286e-05,
+      "loss": 1.3919,
+      "step": 4007
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3023565714285714e-05,
+      "loss": 2.0133,
+      "step": 4008
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3023297142857144e-05,
+      "loss": 1.6178,
+      "step": 4009
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3023028571428572e-05,
+      "loss": 1.3157,
+      "step": 4010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.302276e-05,
+      "loss": 1.3976,
+      "step": 4011
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3022491428571429e-05,
+      "loss": 1.4875,
+      "step": 4012
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3022222857142859e-05,
+      "loss": 1.6407,
+      "step": 4013
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3021954285714285e-05,
+      "loss": 1.5942,
+      "step": 4014
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3021685714285715e-05,
+      "loss": 1.3869,
+      "step": 4015
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3021417142857143e-05,
+      "loss": 1.5504,
+      "step": 4016
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3021148571428571e-05,
+      "loss": 1.4713,
+      "step": 4017
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.302088e-05,
+      "loss": 1.616,
+      "step": 4018
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.302061142857143e-05,
+      "loss": 1.4439,
+      "step": 4019
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3020342857142858e-05,
+      "loss": 1.5975,
+      "step": 4020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3020074285714286e-05,
+      "loss": 1.8668,
+      "step": 4021
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3019805714285716e-05,
+      "loss": 1.8004,
+      "step": 4022
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3019537142857144e-05,
+      "loss": 1.5435,
+      "step": 4023
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3019268571428572e-05,
+      "loss": 1.2734,
+      "step": 4024
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3019e-05,
+      "loss": 1.8786,
+      "step": 4025
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.301873142857143e-05,
+      "loss": 1.4893,
+      "step": 4026
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3018462857142857e-05,
+      "loss": 1.0436,
+      "step": 4027
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3018194285714286e-05,
+      "loss": 1.4991,
+      "step": 4028
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3017925714285715e-05,
+      "loss": 1.2822,
+      "step": 4029
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3017657142857143e-05,
+      "loss": 1.5375,
+      "step": 4030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3017388571428571e-05,
+      "loss": 1.8881,
+      "step": 4031
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.301712e-05,
+      "loss": 1.4076,
+      "step": 4032
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3016851428571429e-05,
+      "loss": 1.3578,
+      "step": 4033
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3016582857142857e-05,
+      "loss": 1.4229,
+      "step": 4034
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3016314285714287e-05,
+      "loss": 1.7216,
+      "step": 4035
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3016045714285715e-05,
+      "loss": 1.9793,
+      "step": 4036
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3015777142857143e-05,
+      "loss": 1.5627,
+      "step": 4037
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3015508571428572e-05,
+      "loss": 1.7722,
+      "step": 4038
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3015240000000001e-05,
+      "loss": 1.4649,
+      "step": 4039
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3014971428571428e-05,
+      "loss": 1.4334,
+      "step": 4040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3014702857142858e-05,
+      "loss": 1.3868,
+      "step": 4041
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3014434285714286e-05,
+      "loss": 1.2204,
+      "step": 4042
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3014165714285714e-05,
+      "loss": 1.4182,
+      "step": 4043
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3013897142857142e-05,
+      "loss": 1.5852,
+      "step": 4044
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3013628571428572e-05,
+      "loss": 1.4458,
+      "step": 4045
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.301336e-05,
+      "loss": 1.4798,
+      "step": 4046
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3013091428571429e-05,
+      "loss": 1.2739,
+      "step": 4047
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3012822857142858e-05,
+      "loss": 1.4233,
+      "step": 4048
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3012554285714287e-05,
+      "loss": 1.1161,
+      "step": 4049
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3012285714285715e-05,
+      "loss": 1.6112,
+      "step": 4050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3012017142857143e-05,
+      "loss": 1.5981,
+      "step": 4051
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3011748571428573e-05,
+      "loss": 1.4758,
+      "step": 4052
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.301148e-05,
+      "loss": 1.5841,
+      "step": 4053
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.301121142857143e-05,
+      "loss": 1.5319,
+      "step": 4054
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3010942857142857e-05,
+      "loss": 1.1445,
+      "step": 4055
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3010674285714287e-05,
+      "loss": 1.7413,
+      "step": 4056
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3010405714285714e-05,
+      "loss": 1.7402,
+      "step": 4057
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3010137142857144e-05,
+      "loss": 1.4227,
+      "step": 4058
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3009868571428572e-05,
+      "loss": 1.5543,
+      "step": 4059
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.30096e-05,
+      "loss": 1.3152,
+      "step": 4060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.300933142857143e-05,
+      "loss": 1.4899,
+      "step": 4061
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3009062857142858e-05,
+      "loss": 1.6823,
+      "step": 4062
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3008794285714286e-05,
+      "loss": 1.7017,
+      "step": 4063
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3008525714285714e-05,
+      "loss": 1.5184,
+      "step": 4064
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3008257142857144e-05,
+      "loss": 1.3532,
+      "step": 4065
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.300798857142857e-05,
+      "loss": 1.3669,
+      "step": 4066
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.300772e-05,
+      "loss": 1.3416,
+      "step": 4067
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3007451428571429e-05,
+      "loss": 1.5232,
+      "step": 4068
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3007182857142859e-05,
+      "loss": 1.4711,
+      "step": 4069
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3006914285714285e-05,
+      "loss": 1.854,
+      "step": 4070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3006645714285715e-05,
+      "loss": 1.2841,
+      "step": 4071
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3006377142857143e-05,
+      "loss": 1.7386,
+      "step": 4072
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3006108571428571e-05,
+      "loss": 1.6451,
+      "step": 4073
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3005840000000001e-05,
+      "loss": 1.2604,
+      "step": 4074
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.300557142857143e-05,
+      "loss": 1.511,
+      "step": 4075
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3005302857142858e-05,
+      "loss": 1.5638,
+      "step": 4076
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3005034285714286e-05,
+      "loss": 1.6241,
+      "step": 4077
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3004765714285716e-05,
+      "loss": 1.6516,
+      "step": 4078
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3004497142857142e-05,
+      "loss": 1.2747,
+      "step": 4079
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3004228571428572e-05,
+      "loss": 1.4905,
+      "step": 4080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.300396e-05,
+      "loss": 1.4328,
+      "step": 4081
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.300369142857143e-05,
+      "loss": 1.3783,
+      "step": 4082
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3003422857142856e-05,
+      "loss": 1.6489,
+      "step": 4083
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3003154285714286e-05,
+      "loss": 1.5401,
+      "step": 4084
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3002885714285714e-05,
+      "loss": 1.4105,
+      "step": 4085
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3002617142857143e-05,
+      "loss": 1.5647,
+      "step": 4086
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3002348571428573e-05,
+      "loss": 1.5336,
+      "step": 4087
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.300208e-05,
+      "loss": 1.3891,
+      "step": 4088
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3001811428571429e-05,
+      "loss": 1.4263,
+      "step": 4089
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3001542857142857e-05,
+      "loss": 1.7578,
+      "step": 4090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3001274285714287e-05,
+      "loss": 1.3897,
+      "step": 4091
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3001005714285715e-05,
+      "loss": 1.7861,
+      "step": 4092
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3000737142857143e-05,
+      "loss": 1.5431,
+      "step": 4093
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3000468571428571e-05,
+      "loss": 1.5153,
+      "step": 4094
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3000200000000001e-05,
+      "loss": 1.9796,
+      "step": 4095
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2999931428571428e-05,
+      "loss": 1.4917,
+      "step": 4096
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2999662857142858e-05,
+      "loss": 1.3609,
+      "step": 4097
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2999394285714286e-05,
+      "loss": 1.5343,
+      "step": 4098
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2999125714285714e-05,
+      "loss": 1.4861,
+      "step": 4099
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2998857142857144e-05,
+      "loss": 1.4553,
+      "step": 4100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2998588571428572e-05,
+      "loss": 1.3766,
+      "step": 4101
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.299832e-05,
+      "loss": 1.3088,
+      "step": 4102
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2998051428571428e-05,
+      "loss": 1.569,
+      "step": 4103
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2997782857142858e-05,
+      "loss": 1.3797,
+      "step": 4104
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2997514285714286e-05,
+      "loss": 1.6024,
+      "step": 4105
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2997245714285715e-05,
+      "loss": 1.5751,
+      "step": 4106
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2996977142857143e-05,
+      "loss": 1.6636,
+      "step": 4107
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2996708571428573e-05,
+      "loss": 1.8192,
+      "step": 4108
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.299644e-05,
+      "loss": 1.4834,
+      "step": 4109
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2996171428571429e-05,
+      "loss": 1.602,
+      "step": 4110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2995902857142857e-05,
+      "loss": 1.4936,
+      "step": 4111
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.2995634285714285e-05,
+      "loss": 1.3824,
+      "step": 4112
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2995365714285715e-05,
+      "loss": 1.5448,
+      "step": 4113
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2995097142857143e-05,
+      "loss": 1.0537,
+      "step": 4114
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2994828571428573e-05,
+      "loss": 1.5132,
+      "step": 4115
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.299456e-05,
+      "loss": 1.4155,
+      "step": 4116
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.299429142857143e-05,
+      "loss": 1.6105,
+      "step": 4117
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2994022857142858e-05,
+      "loss": 1.3218,
+      "step": 4118
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2993754285714286e-05,
+      "loss": 1.5083,
+      "step": 4119
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2993485714285714e-05,
+      "loss": 1.4496,
+      "step": 4120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2993217142857144e-05,
+      "loss": 1.805,
+      "step": 4121
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.299294857142857e-05,
+      "loss": 1.5228,
+      "step": 4122
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.299268e-05,
+      "loss": 1.5007,
+      "step": 4123
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2992411428571429e-05,
+      "loss": 1.4985,
+      "step": 4124
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2992142857142857e-05,
+      "loss": 1.6093,
+      "step": 4125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2991874285714287e-05,
+      "loss": 1.7935,
+      "step": 4126
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2991605714285715e-05,
+      "loss": 1.5353,
+      "step": 4127
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2991337142857145e-05,
+      "loss": 1.5546,
+      "step": 4128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2991068571428571e-05,
+      "loss": 1.6972,
+      "step": 4129
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2990800000000001e-05,
+      "loss": 1.411,
+      "step": 4130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.299053142857143e-05,
+      "loss": 1.3559,
+      "step": 4131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2990262857142857e-05,
+      "loss": 1.7643,
+      "step": 4132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2989994285714286e-05,
+      "loss": 1.6341,
+      "step": 4133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2989725714285715e-05,
+      "loss": 1.329,
+      "step": 4134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2989457142857142e-05,
+      "loss": 1.7393,
+      "step": 4135
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2989188571428572e-05,
+      "loss": 1.2548,
+      "step": 4136
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.298892e-05,
+      "loss": 1.474,
+      "step": 4137
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.298865142857143e-05,
+      "loss": 1.5254,
+      "step": 4138
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2988382857142858e-05,
+      "loss": 2.0204,
+      "step": 4139
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2988114285714286e-05,
+      "loss": 1.3795,
+      "step": 4140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2987845714285716e-05,
+      "loss": 1.8091,
+      "step": 4141
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2987577142857143e-05,
+      "loss": 1.369,
+      "step": 4142
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2987308571428572e-05,
+      "loss": 1.7245,
+      "step": 4143
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.298704e-05,
+      "loss": 1.281,
+      "step": 4144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2986771428571429e-05,
+      "loss": 1.4888,
+      "step": 4145
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2986502857142857e-05,
+      "loss": 1.2639,
+      "step": 4146
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2986234285714287e-05,
+      "loss": 1.544,
+      "step": 4147
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2985965714285713e-05,
+      "loss": 1.4693,
+      "step": 4148
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2985697142857143e-05,
+      "loss": 1.4324,
+      "step": 4149
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2985428571428571e-05,
+      "loss": 1.4754,
+      "step": 4150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2985160000000001e-05,
+      "loss": 1.7051,
+      "step": 4151
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.298489142857143e-05,
+      "loss": 1.5257,
+      "step": 4152
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2984622857142858e-05,
+      "loss": 1.9046,
+      "step": 4153
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2984354285714287e-05,
+      "loss": 1.3475,
+      "step": 4154
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2984085714285714e-05,
+      "loss": 1.2043,
+      "step": 4155
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2983817142857144e-05,
+      "loss": 1.4249,
+      "step": 4156
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2983548571428572e-05,
+      "loss": 1.6527,
+      "step": 4157
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.298328e-05,
+      "loss": 1.5467,
+      "step": 4158
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2983011428571428e-05,
+      "loss": 1.5541,
+      "step": 4159
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2982742857142858e-05,
+      "loss": 1.5984,
+      "step": 4160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2982474285714286e-05,
+      "loss": 1.3373,
+      "step": 4161
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2982205714285715e-05,
+      "loss": 1.3262,
+      "step": 4162
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2981937142857143e-05,
+      "loss": 1.7588,
+      "step": 4163
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2981668571428573e-05,
+      "loss": 1.7464,
+      "step": 4164
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.29814e-05,
+      "loss": 1.339,
+      "step": 4165
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2981131428571429e-05,
+      "loss": 1.2753,
+      "step": 4166
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2980862857142859e-05,
+      "loss": 1.1878,
+      "step": 4167
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2980594285714285e-05,
+      "loss": 1.344,
+      "step": 4168
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2980325714285715e-05,
+      "loss": 1.3677,
+      "step": 4169
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2980057142857143e-05,
+      "loss": 1.4273,
+      "step": 4170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2979788571428572e-05,
+      "loss": 1.4639,
+      "step": 4171
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.297952e-05,
+      "loss": 1.4224,
+      "step": 4172
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.297925142857143e-05,
+      "loss": 1.5689,
+      "step": 4173
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2978982857142858e-05,
+      "loss": 1.6884,
+      "step": 4174
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2978714285714286e-05,
+      "loss": 1.8249,
+      "step": 4175
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2978445714285714e-05,
+      "loss": 1.7976,
+      "step": 4176
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2978177142857144e-05,
+      "loss": 1.6196,
+      "step": 4177
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2977908571428572e-05,
+      "loss": 1.5897,
+      "step": 4178
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.297764e-05,
+      "loss": 1.7056,
+      "step": 4179
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.297737142857143e-05,
+      "loss": 1.5128,
+      "step": 4180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2977102857142857e-05,
+      "loss": 1.1498,
+      "step": 4181
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2976834285714287e-05,
+      "loss": 1.6009,
+      "step": 4182
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2976565714285715e-05,
+      "loss": 1.5421,
+      "step": 4183
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2976297142857145e-05,
+      "loss": 1.2678,
+      "step": 4184
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2976028571428571e-05,
+      "loss": 1.7153,
+      "step": 4185
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2975760000000001e-05,
+      "loss": 1.1514,
+      "step": 4186
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2975491428571429e-05,
+      "loss": 1.1585,
+      "step": 4187
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2975222857142857e-05,
+      "loss": 1.4284,
+      "step": 4188
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2974954285714285e-05,
+      "loss": 1.2203,
+      "step": 4189
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2974685714285715e-05,
+      "loss": 1.6669,
+      "step": 4190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2974417142857144e-05,
+      "loss": 1.3592,
+      "step": 4191
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2974148571428572e-05,
+      "loss": 1.3719,
+      "step": 4192
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2973880000000002e-05,
+      "loss": 1.2702,
+      "step": 4193
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2973611428571428e-05,
+      "loss": 1.2926,
+      "step": 4194
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2973342857142858e-05,
+      "loss": 1.4438,
+      "step": 4195
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2973074285714286e-05,
+      "loss": 1.5756,
+      "step": 4196
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2972805714285716e-05,
+      "loss": 1.3695,
+      "step": 4197
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2972537142857142e-05,
+      "loss": 1.6301,
+      "step": 4198
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2972268571428572e-05,
+      "loss": 1.7061,
+      "step": 4199
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2972e-05,
+      "loss": 1.5538,
+      "step": 4200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2971731428571429e-05,
+      "loss": 1.6098,
+      "step": 4201
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2971462857142857e-05,
+      "loss": 1.5748,
+      "step": 4202
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2971194285714287e-05,
+      "loss": 1.5857,
+      "step": 4203
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2970925714285715e-05,
+      "loss": 1.5027,
+      "step": 4204
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2970657142857143e-05,
+      "loss": 1.389,
+      "step": 4205
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2970388571428573e-05,
+      "loss": 1.4815,
+      "step": 4206
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2970120000000001e-05,
+      "loss": 1.4904,
+      "step": 4207
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.296985142857143e-05,
+      "loss": 1.5141,
+      "step": 4208
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2969582857142857e-05,
+      "loss": 1.636,
+      "step": 4209
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2969314285714287e-05,
+      "loss": 1.0359,
+      "step": 4210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2969045714285714e-05,
+      "loss": 1.2318,
+      "step": 4211
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2968777142857144e-05,
+      "loss": 1.6769,
+      "step": 4212
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2968508571428572e-05,
+      "loss": 1.4853,
+      "step": 4213
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.296824e-05,
+      "loss": 1.5001,
+      "step": 4214
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2967971428571428e-05,
+      "loss": 1.5735,
+      "step": 4215
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2967702857142858e-05,
+      "loss": 1.6201,
+      "step": 4216
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2967434285714286e-05,
+      "loss": 1.3236,
+      "step": 4217
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2967165714285714e-05,
+      "loss": 1.5792,
+      "step": 4218
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2966897142857144e-05,
+      "loss": 1.8131,
+      "step": 4219
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2966628571428572e-05,
+      "loss": 1.592,
+      "step": 4220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.296636e-05,
+      "loss": 1.3216,
+      "step": 4221
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2966091428571429e-05,
+      "loss": 1.7528,
+      "step": 4222
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2965822857142859e-05,
+      "loss": 1.5228,
+      "step": 4223
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2965554285714285e-05,
+      "loss": 1.3622,
+      "step": 4224
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2965285714285715e-05,
+      "loss": 1.5982,
+      "step": 4225
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2965017142857143e-05,
+      "loss": 1.2226,
+      "step": 4226
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2964748571428571e-05,
+      "loss": 1.7503,
+      "step": 4227
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.296448e-05,
+      "loss": 1.6579,
+      "step": 4228
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.296421142857143e-05,
+      "loss": 1.3657,
+      "step": 4229
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2963942857142858e-05,
+      "loss": 1.331,
+      "step": 4230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2963674285714286e-05,
+      "loss": 1.5581,
+      "step": 4231
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2963405714285716e-05,
+      "loss": 1.4199,
+      "step": 4232
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2963137142857144e-05,
+      "loss": 1.5767,
+      "step": 4233
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2962868571428572e-05,
+      "loss": 1.5392,
+      "step": 4234
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.29626e-05,
+      "loss": 1.3217,
+      "step": 4235
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.296233142857143e-05,
+      "loss": 1.5264,
+      "step": 4236
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2962062857142857e-05,
+      "loss": 1.7306,
+      "step": 4237
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2961794285714286e-05,
+      "loss": 1.5982,
+      "step": 4238
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2961525714285715e-05,
+      "loss": 1.3198,
+      "step": 4239
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2961257142857143e-05,
+      "loss": 1.5892,
+      "step": 4240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2960988571428571e-05,
+      "loss": 1.479,
+      "step": 4241
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.296072e-05,
+      "loss": 1.6147,
+      "step": 4242
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2960451428571429e-05,
+      "loss": 1.496,
+      "step": 4243
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2960182857142857e-05,
+      "loss": 1.9057,
+      "step": 4244
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2959914285714287e-05,
+      "loss": 1.4463,
+      "step": 4245
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2959645714285715e-05,
+      "loss": 1.4502,
+      "step": 4246
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2959377142857143e-05,
+      "loss": 1.5387,
+      "step": 4247
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2959108571428572e-05,
+      "loss": 1.8028,
+      "step": 4248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2958840000000001e-05,
+      "loss": 1.1043,
+      "step": 4249
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2958571428571428e-05,
+      "loss": 1.4923,
+      "step": 4250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2958302857142858e-05,
+      "loss": 1.2514,
+      "step": 4251
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2958034285714286e-05,
+      "loss": 1.3939,
+      "step": 4252
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2957765714285714e-05,
+      "loss": 1.0931,
+      "step": 4253
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2957497142857142e-05,
+      "loss": 1.4381,
+      "step": 4254
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2957228571428572e-05,
+      "loss": 1.4336,
+      "step": 4255
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.295696e-05,
+      "loss": 1.4906,
+      "step": 4256
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2956691428571429e-05,
+      "loss": 1.1959,
+      "step": 4257
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2956422857142858e-05,
+      "loss": 1.3948,
+      "step": 4258
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2956154285714287e-05,
+      "loss": 1.2639,
+      "step": 4259
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2955885714285715e-05,
+      "loss": 1.5083,
+      "step": 4260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2955617142857143e-05,
+      "loss": 1.5424,
+      "step": 4261
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2955348571428573e-05,
+      "loss": 1.3557,
+      "step": 4262
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.295508e-05,
+      "loss": 1.6545,
+      "step": 4263
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.295481142857143e-05,
+      "loss": 1.5894,
+      "step": 4264
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2954542857142857e-05,
+      "loss": 1.5422,
+      "step": 4265
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2954274285714287e-05,
+      "loss": 1.4399,
+      "step": 4266
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2954005714285714e-05,
+      "loss": 1.4622,
+      "step": 4267
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2953737142857144e-05,
+      "loss": 1.6432,
+      "step": 4268
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2953468571428572e-05,
+      "loss": 1.7302,
+      "step": 4269
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.29532e-05,
+      "loss": 1.3866,
+      "step": 4270
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.295293142857143e-05,
+      "loss": 1.763,
+      "step": 4271
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2952662857142858e-05,
+      "loss": 1.5558,
+      "step": 4272
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2952394285714286e-05,
+      "loss": 1.1018,
+      "step": 4273
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2952125714285714e-05,
+      "loss": 1.5794,
+      "step": 4274
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2951857142857144e-05,
+      "loss": 1.4124,
+      "step": 4275
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.295158857142857e-05,
+      "loss": 1.5045,
+      "step": 4276
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.295132e-05,
+      "loss": 1.5051,
+      "step": 4277
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2951051428571429e-05,
+      "loss": 1.6879,
+      "step": 4278
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2950782857142859e-05,
+      "loss": 1.5138,
+      "step": 4279
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2950514285714285e-05,
+      "loss": 1.2741,
+      "step": 4280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2950245714285715e-05,
+      "loss": 1.3775,
+      "step": 4281
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2949977142857143e-05,
+      "loss": 1.3361,
+      "step": 4282
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2949708571428571e-05,
+      "loss": 1.7531,
+      "step": 4283
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2949440000000001e-05,
+      "loss": 1.519,
+      "step": 4284
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.294917142857143e-05,
+      "loss": 1.5626,
+      "step": 4285
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2948902857142858e-05,
+      "loss": 1.5867,
+      "step": 4286
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.2948634285714286e-05,
+      "loss": 1.414,
+      "step": 4287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2948365714285716e-05,
+      "loss": 1.3506,
+      "step": 4288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2948097142857144e-05,
+      "loss": 1.2561,
+      "step": 4289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2947828571428572e-05,
+      "loss": 1.836,
+      "step": 4290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.294756e-05,
+      "loss": 1.4477,
+      "step": 4291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.294729142857143e-05,
+      "loss": 1.1883,
+      "step": 4292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2947022857142856e-05,
+      "loss": 1.5414,
+      "step": 4293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2946754285714286e-05,
+      "loss": 1.6755,
+      "step": 4294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2946485714285714e-05,
+      "loss": 1.6158,
+      "step": 4295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2946217142857143e-05,
+      "loss": 1.4298,
+      "step": 4296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2945948571428573e-05,
+      "loss": 1.6705,
+      "step": 4297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.294568e-05,
+      "loss": 1.588,
+      "step": 4298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2945411428571429e-05,
+      "loss": 1.4436,
+      "step": 4299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2945142857142857e-05,
+      "loss": 1.6782,
+      "step": 4300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2944874285714287e-05,
+      "loss": 1.4911,
+      "step": 4301
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2944605714285715e-05,
+      "loss": 1.5063,
+      "step": 4302
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2944337142857143e-05,
+      "loss": 1.2652,
+      "step": 4303
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2944068571428571e-05,
+      "loss": 1.6712,
+      "step": 4304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2943800000000001e-05,
+      "loss": 1.3496,
+      "step": 4305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2943531428571428e-05,
+      "loss": 1.335,
+      "step": 4306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2943262857142858e-05,
+      "loss": 1.1564,
+      "step": 4307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2942994285714286e-05,
+      "loss": 1.5081,
+      "step": 4308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2942725714285714e-05,
+      "loss": 1.4153,
+      "step": 4309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2942457142857144e-05,
+      "loss": 1.3526,
+      "step": 4310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2942188571428572e-05,
+      "loss": 1.5147,
+      "step": 4311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2941920000000002e-05,
+      "loss": 1.2926,
+      "step": 4312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2941651428571428e-05,
+      "loss": 1.7854,
+      "step": 4313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2941382857142858e-05,
+      "loss": 1.4254,
+      "step": 4314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2941114285714286e-05,
+      "loss": 1.409,
+      "step": 4315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2940845714285715e-05,
+      "loss": 1.5212,
+      "step": 4316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2940577142857143e-05,
+      "loss": 1.408,
+      "step": 4317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2940308571428573e-05,
+      "loss": 1.5597,
+      "step": 4318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.294004e-05,
+      "loss": 1.3579,
+      "step": 4319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2939771428571429e-05,
+      "loss": 1.3751,
+      "step": 4320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2939502857142857e-05,
+      "loss": 1.4461,
+      "step": 4321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2939234285714285e-05,
+      "loss": 1.5139,
+      "step": 4322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2938965714285715e-05,
+      "loss": 1.5937,
+      "step": 4323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2938697142857143e-05,
+      "loss": 1.468,
+      "step": 4324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2938428571428573e-05,
+      "loss": 1.841,
+      "step": 4325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293816e-05,
+      "loss": 1.9148,
+      "step": 4326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293789142857143e-05,
+      "loss": 1.6185,
+      "step": 4327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2937622857142858e-05,
+      "loss": 1.4139,
+      "step": 4328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2937354285714286e-05,
+      "loss": 1.7665,
+      "step": 4329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2937085714285714e-05,
+      "loss": 1.7707,
+      "step": 4330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2936817142857144e-05,
+      "loss": 1.7023,
+      "step": 4331
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293654857142857e-05,
+      "loss": 1.6724,
+      "step": 4332
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293628e-05,
+      "loss": 1.6838,
+      "step": 4333
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2936011428571429e-05,
+      "loss": 1.3522,
+      "step": 4334
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2935742857142858e-05,
+      "loss": 1.4067,
+      "step": 4335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2935474285714287e-05,
+      "loss": 1.6705,
+      "step": 4336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2935205714285715e-05,
+      "loss": 1.7657,
+      "step": 4337
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2934937142857145e-05,
+      "loss": 1.5149,
+      "step": 4338
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2934668571428571e-05,
+      "loss": 1.3123,
+      "step": 4339
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2934400000000001e-05,
+      "loss": 1.7111,
+      "step": 4340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293413142857143e-05,
+      "loss": 1.2678,
+      "step": 4341
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2933862857142857e-05,
+      "loss": 1.5529,
+      "step": 4342
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2933594285714286e-05,
+      "loss": 1.4476,
+      "step": 4343
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2933325714285715e-05,
+      "loss": 1.4102,
+      "step": 4344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2933057142857142e-05,
+      "loss": 1.3785,
+      "step": 4345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2932788571428572e-05,
+      "loss": 1.2974,
+      "step": 4346
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293252e-05,
+      "loss": 1.3641,
+      "step": 4347
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293225142857143e-05,
+      "loss": 1.3135,
+      "step": 4348
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2931982857142858e-05,
+      "loss": 1.4694,
+      "step": 4349
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2931714285714286e-05,
+      "loss": 1.6582,
+      "step": 4350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2931445714285716e-05,
+      "loss": 1.555,
+      "step": 4351
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2931177142857143e-05,
+      "loss": 1.7469,
+      "step": 4352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2930908571428572e-05,
+      "loss": 1.4413,
+      "step": 4353
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.293064e-05,
+      "loss": 1.6978,
+      "step": 4354
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2930371428571429e-05,
+      "loss": 1.7116,
+      "step": 4355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2930102857142857e-05,
+      "loss": 1.7508,
+      "step": 4356
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2929834285714287e-05,
+      "loss": 1.796,
+      "step": 4357
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2929565714285715e-05,
+      "loss": 1.3295,
+      "step": 4358
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2929297142857143e-05,
+      "loss": 1.5193,
+      "step": 4359
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2929028571428571e-05,
+      "loss": 1.7857,
+      "step": 4360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2928760000000001e-05,
+      "loss": 1.2772,
+      "step": 4361
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.292849142857143e-05,
+      "loss": 1.4371,
+      "step": 4362
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2928222857142858e-05,
+      "loss": 1.619,
+      "step": 4363
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2927954285714287e-05,
+      "loss": 1.437,
+      "step": 4364
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2927685714285714e-05,
+      "loss": 1.2556,
+      "step": 4365
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2927417142857144e-05,
+      "loss": 1.6882,
+      "step": 4366
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2927148571428572e-05,
+      "loss": 1.2848,
+      "step": 4367
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.292688e-05,
+      "loss": 1.4912,
+      "step": 4368
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2926611428571428e-05,
+      "loss": 1.2925,
+      "step": 4369
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2926342857142858e-05,
+      "loss": 1.3648,
+      "step": 4370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2926074285714286e-05,
+      "loss": 1.8293,
+      "step": 4371
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2925805714285715e-05,
+      "loss": 1.7943,
+      "step": 4372
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2925537142857143e-05,
+      "loss": 1.4395,
+      "step": 4373
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2925268571428573e-05,
+      "loss": 1.391,
+      "step": 4374
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2925e-05,
+      "loss": 1.5893,
+      "step": 4375
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2924731428571429e-05,
+      "loss": 1.6008,
+      "step": 4376
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2924462857142859e-05,
+      "loss": 1.5062,
+      "step": 4377
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2924194285714285e-05,
+      "loss": 1.6377,
+      "step": 4378
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2923925714285715e-05,
+      "loss": 1.4713,
+      "step": 4379
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2923657142857143e-05,
+      "loss": 1.374,
+      "step": 4380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2923388571428573e-05,
+      "loss": 1.5892,
+      "step": 4381
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.292312e-05,
+      "loss": 1.3649,
+      "step": 4382
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.292285142857143e-05,
+      "loss": 1.1252,
+      "step": 4383
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2922582857142858e-05,
+      "loss": 1.3914,
+      "step": 4384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2922314285714286e-05,
+      "loss": 1.4616,
+      "step": 4385
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2922045714285714e-05,
+      "loss": 1.6449,
+      "step": 4386
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2921777142857144e-05,
+      "loss": 1.3422,
+      "step": 4387
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2921508571428572e-05,
+      "loss": 1.4588,
+      "step": 4388
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.292124e-05,
+      "loss": 1.7742,
+      "step": 4389
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.292097142857143e-05,
+      "loss": 1.1105,
+      "step": 4390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2920702857142857e-05,
+      "loss": 1.3926,
+      "step": 4391
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2920434285714287e-05,
+      "loss": 1.8495,
+      "step": 4392
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2920165714285715e-05,
+      "loss": 1.5672,
+      "step": 4393
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2919897142857145e-05,
+      "loss": 1.3712,
+      "step": 4394
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2919628571428571e-05,
+      "loss": 1.5786,
+      "step": 4395
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2919360000000001e-05,
+      "loss": 1.566,
+      "step": 4396
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2919091428571429e-05,
+      "loss": 1.2875,
+      "step": 4397
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2918822857142857e-05,
+      "loss": 1.6032,
+      "step": 4398
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2918554285714285e-05,
+      "loss": 1.4741,
+      "step": 4399
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2918285714285715e-05,
+      "loss": 1.2699,
+      "step": 4400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2918017142857144e-05,
+      "loss": 1.6955,
+      "step": 4401
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2917748571428572e-05,
+      "loss": 1.5074,
+      "step": 4402
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2917480000000002e-05,
+      "loss": 1.4148,
+      "step": 4403
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2917211428571428e-05,
+      "loss": 1.5169,
+      "step": 4404
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2916942857142858e-05,
+      "loss": 1.1821,
+      "step": 4405
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2916674285714286e-05,
+      "loss": 1.1287,
+      "step": 4406
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2916405714285716e-05,
+      "loss": 1.6918,
+      "step": 4407
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2916137142857142e-05,
+      "loss": 1.2848,
+      "step": 4408
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2915868571428572e-05,
+      "loss": 1.7106,
+      "step": 4409
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.29156e-05,
+      "loss": 1.6963,
+      "step": 4410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2915331428571429e-05,
+      "loss": 1.4111,
+      "step": 4411
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2915062857142857e-05,
+      "loss": 1.5966,
+      "step": 4412
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2914794285714287e-05,
+      "loss": 1.2766,
+      "step": 4413
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2914525714285715e-05,
+      "loss": 1.4829,
+      "step": 4414
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2914257142857143e-05,
+      "loss": 1.4137,
+      "step": 4415
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2913988571428573e-05,
+      "loss": 1.5755,
+      "step": 4416
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2913720000000001e-05,
+      "loss": 1.8777,
+      "step": 4417
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.291345142857143e-05,
+      "loss": 1.7096,
+      "step": 4418
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2913182857142857e-05,
+      "loss": 1.5521,
+      "step": 4419
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2912914285714287e-05,
+      "loss": 1.5698,
+      "step": 4420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2912645714285714e-05,
+      "loss": 1.542,
+      "step": 4421
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2912377142857144e-05,
+      "loss": 1.2712,
+      "step": 4422
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2912108571428572e-05,
+      "loss": 1.7101,
+      "step": 4423
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.291184e-05,
+      "loss": 1.3956,
+      "step": 4424
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2911571428571428e-05,
+      "loss": 1.4642,
+      "step": 4425
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2911302857142858e-05,
+      "loss": 1.7903,
+      "step": 4426
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2911034285714286e-05,
+      "loss": 1.5146,
+      "step": 4427
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2910765714285714e-05,
+      "loss": 1.2058,
+      "step": 4428
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2910497142857144e-05,
+      "loss": 1.7646,
+      "step": 4429
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2910228571428572e-05,
+      "loss": 1.1172,
+      "step": 4430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.290996e-05,
+      "loss": 1.5545,
+      "step": 4431
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2909691428571429e-05,
+      "loss": 1.3564,
+      "step": 4432
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2909422857142859e-05,
+      "loss": 1.5973,
+      "step": 4433
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2909154285714285e-05,
+      "loss": 1.593,
+      "step": 4434
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2908885714285715e-05,
+      "loss": 1.3172,
+      "step": 4435
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2908617142857143e-05,
+      "loss": 1.7103,
+      "step": 4436
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2908348571428571e-05,
+      "loss": 1.3748,
+      "step": 4437
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.290808e-05,
+      "loss": 1.9295,
+      "step": 4438
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.290781142857143e-05,
+      "loss": 1.3615,
+      "step": 4439
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2907542857142858e-05,
+      "loss": 1.5192,
+      "step": 4440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2907274285714286e-05,
+      "loss": 1.4734,
+      "step": 4441
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2907005714285716e-05,
+      "loss": 1.6967,
+      "step": 4442
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2906737142857144e-05,
+      "loss": 1.1852,
+      "step": 4443
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2906468571428572e-05,
+      "loss": 1.3486,
+      "step": 4444
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.29062e-05,
+      "loss": 1.3775,
+      "step": 4445
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.290593142857143e-05,
+      "loss": 1.2367,
+      "step": 4446
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2905662857142857e-05,
+      "loss": 1.5368,
+      "step": 4447
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2905394285714286e-05,
+      "loss": 1.7195,
+      "step": 4448
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2905125714285715e-05,
+      "loss": 1.4417,
+      "step": 4449
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2904857142857143e-05,
+      "loss": 1.295,
+      "step": 4450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2904588571428571e-05,
+      "loss": 1.2161,
+      "step": 4451
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.290432e-05,
+      "loss": 1.275,
+      "step": 4452
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2904051428571429e-05,
+      "loss": 1.4919,
+      "step": 4453
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2903782857142857e-05,
+      "loss": 1.56,
+      "step": 4454
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2903514285714287e-05,
+      "loss": 1.7722,
+      "step": 4455
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2903245714285715e-05,
+      "loss": 1.3722,
+      "step": 4456
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2902977142857143e-05,
+      "loss": 1.1153,
+      "step": 4457
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2902708571428572e-05,
+      "loss": 1.2057,
+      "step": 4458
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2902440000000001e-05,
+      "loss": 1.2512,
+      "step": 4459
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2902171428571428e-05,
+      "loss": 1.4838,
+      "step": 4460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2901902857142858e-05,
+      "loss": 1.6095,
+      "step": 4461
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.2901634285714286e-05,
+      "loss": 1.2173,
+      "step": 4462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2901365714285716e-05,
+      "loss": 1.7234,
+      "step": 4463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2901097142857142e-05,
+      "loss": 1.4451,
+      "step": 4464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2900828571428572e-05,
+      "loss": 1.8727,
+      "step": 4465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.290056e-05,
+      "loss": 1.2101,
+      "step": 4466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2900291428571429e-05,
+      "loss": 1.3802,
+      "step": 4467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2900022857142858e-05,
+      "loss": 1.5901,
+      "step": 4468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2899754285714287e-05,
+      "loss": 1.7466,
+      "step": 4469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2899485714285715e-05,
+      "loss": 1.686,
+      "step": 4470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2899217142857143e-05,
+      "loss": 1.2661,
+      "step": 4471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2898948571428573e-05,
+      "loss": 1.7438,
+      "step": 4472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.289868e-05,
+      "loss": 1.3219,
+      "step": 4473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.289841142857143e-05,
+      "loss": 1.4515,
+      "step": 4474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2898142857142857e-05,
+      "loss": 1.5601,
+      "step": 4475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2897874285714287e-05,
+      "loss": 1.5085,
+      "step": 4476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2897605714285714e-05,
+      "loss": 1.3604,
+      "step": 4477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2897337142857144e-05,
+      "loss": 1.2595,
+      "step": 4478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2897068571428572e-05,
+      "loss": 1.6256,
+      "step": 4479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.28968e-05,
+      "loss": 1.6564,
+      "step": 4480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.289653142857143e-05,
+      "loss": 1.5998,
+      "step": 4481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2896262857142858e-05,
+      "loss": 1.6617,
+      "step": 4482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2895994285714286e-05,
+      "loss": 1.6131,
+      "step": 4483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2895725714285714e-05,
+      "loss": 1.7678,
+      "step": 4484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2895457142857144e-05,
+      "loss": 1.6833,
+      "step": 4485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2895188571428572e-05,
+      "loss": 1.5972,
+      "step": 4486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.289492e-05,
+      "loss": 1.4541,
+      "step": 4487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2894651428571429e-05,
+      "loss": 1.5563,
+      "step": 4488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2894382857142859e-05,
+      "loss": 1.7931,
+      "step": 4489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2894114285714285e-05,
+      "loss": 1.2446,
+      "step": 4490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2893845714285715e-05,
+      "loss": 1.4216,
+      "step": 4491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2893577142857143e-05,
+      "loss": 1.7063,
+      "step": 4492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2893308571428571e-05,
+      "loss": 1.3148,
+      "step": 4493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2893040000000001e-05,
+      "loss": 1.4646,
+      "step": 4494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.289277142857143e-05,
+      "loss": 1.5742,
+      "step": 4495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2892502857142858e-05,
+      "loss": 1.7568,
+      "step": 4496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2892234285714286e-05,
+      "loss": 1.2006,
+      "step": 4497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2891965714285716e-05,
+      "loss": 1.5427,
+      "step": 4498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2891697142857144e-05,
+      "loss": 1.3818,
+      "step": 4499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2891428571428572e-05,
+      "loss": 1.3433,
+      "step": 4500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.289116e-05,
+      "loss": 0.9108,
+      "step": 4501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.289089142857143e-05,
+      "loss": 1.8431,
+      "step": 4502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2890622857142856e-05,
+      "loss": 2.1152,
+      "step": 4503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2890354285714286e-05,
+      "loss": 1.1715,
+      "step": 4504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2890085714285714e-05,
+      "loss": 1.6097,
+      "step": 4505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2889817142857143e-05,
+      "loss": 1.3243,
+      "step": 4506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2889548571428573e-05,
+      "loss": 1.6044,
+      "step": 4507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.288928e-05,
+      "loss": 1.4272,
+      "step": 4508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.288901142857143e-05,
+      "loss": 1.5941,
+      "step": 4509
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2888742857142857e-05,
+      "loss": 1.3867,
+      "step": 4510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2888474285714287e-05,
+      "loss": 1.6303,
+      "step": 4511
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2888205714285715e-05,
+      "loss": 1.7948,
+      "step": 4512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2887937142857143e-05,
+      "loss": 1.3868,
+      "step": 4513
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2887668571428571e-05,
+      "loss": 1.7335,
+      "step": 4514
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2887400000000001e-05,
+      "loss": 1.242,
+      "step": 4515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2887131428571428e-05,
+      "loss": 1.2459,
+      "step": 4516
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2886862857142858e-05,
+      "loss": 1.0573,
+      "step": 4517
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2886594285714286e-05,
+      "loss": 1.4558,
+      "step": 4518
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2886325714285714e-05,
+      "loss": 1.1824,
+      "step": 4519
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2886057142857144e-05,
+      "loss": 1.2684,
+      "step": 4520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2885788571428572e-05,
+      "loss": 1.554,
+      "step": 4521
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2885520000000002e-05,
+      "loss": 1.5259,
+      "step": 4522
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2885251428571428e-05,
+      "loss": 1.4532,
+      "step": 4523
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2884982857142858e-05,
+      "loss": 1.3131,
+      "step": 4524
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2884714285714286e-05,
+      "loss": 1.3366,
+      "step": 4525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2884445714285715e-05,
+      "loss": 1.5377,
+      "step": 4526
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2884177142857143e-05,
+      "loss": 1.5288,
+      "step": 4527
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2883908571428573e-05,
+      "loss": 1.2986,
+      "step": 4528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.288364e-05,
+      "loss": 1.3848,
+      "step": 4529
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2883371428571429e-05,
+      "loss": 1.7479,
+      "step": 4530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2883102857142857e-05,
+      "loss": 1.6035,
+      "step": 4531
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2882834285714287e-05,
+      "loss": 1.4308,
+      "step": 4532
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2882565714285715e-05,
+      "loss": 1.3038,
+      "step": 4533
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2882297142857143e-05,
+      "loss": 1.5654,
+      "step": 4534
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2882028571428573e-05,
+      "loss": 1.5111,
+      "step": 4535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.288176e-05,
+      "loss": 1.6582,
+      "step": 4536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.288149142857143e-05,
+      "loss": 1.4593,
+      "step": 4537
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2881222857142858e-05,
+      "loss": 1.6127,
+      "step": 4538
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2880954285714286e-05,
+      "loss": 1.3783,
+      "step": 4539
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2880685714285714e-05,
+      "loss": 0.9621,
+      "step": 4540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2880417142857144e-05,
+      "loss": 1.3913,
+      "step": 4541
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.288014857142857e-05,
+      "loss": 1.5772,
+      "step": 4542
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.287988e-05,
+      "loss": 1.3304,
+      "step": 4543
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2879611428571429e-05,
+      "loss": 1.4931,
+      "step": 4544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2879342857142858e-05,
+      "loss": 1.7036,
+      "step": 4545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2879074285714287e-05,
+      "loss": 1.1967,
+      "step": 4546
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2878805714285715e-05,
+      "loss": 1.9025,
+      "step": 4547
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2878537142857145e-05,
+      "loss": 1.4643,
+      "step": 4548
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2878268571428571e-05,
+      "loss": 1.6424,
+      "step": 4549
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2878000000000001e-05,
+      "loss": 1.5321,
+      "step": 4550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.287773142857143e-05,
+      "loss": 1.4388,
+      "step": 4551
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2877462857142857e-05,
+      "loss": 1.8411,
+      "step": 4552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2877194285714286e-05,
+      "loss": 1.3957,
+      "step": 4553
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2876925714285715e-05,
+      "loss": 1.5463,
+      "step": 4554
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2876657142857144e-05,
+      "loss": 1.4528,
+      "step": 4555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2876388571428572e-05,
+      "loss": 1.2532,
+      "step": 4556
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.287612e-05,
+      "loss": 1.7481,
+      "step": 4557
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.287585142857143e-05,
+      "loss": 1.4039,
+      "step": 4558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2875582857142858e-05,
+      "loss": 1.3735,
+      "step": 4559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2875314285714286e-05,
+      "loss": 1.8458,
+      "step": 4560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2875045714285716e-05,
+      "loss": 1.2974,
+      "step": 4561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2874777142857143e-05,
+      "loss": 1.3629,
+      "step": 4562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2874508571428572e-05,
+      "loss": 1.2566,
+      "step": 4563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.287424e-05,
+      "loss": 1.493,
+      "step": 4564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2873971428571429e-05,
+      "loss": 1.2581,
+      "step": 4565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2873702857142857e-05,
+      "loss": 1.3546,
+      "step": 4566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2873434285714287e-05,
+      "loss": 1.296,
+      "step": 4567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2873165714285715e-05,
+      "loss": 1.6065,
+      "step": 4568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2872897142857143e-05,
+      "loss": 1.4981,
+      "step": 4569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2872628571428571e-05,
+      "loss": 1.5334,
+      "step": 4570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2872360000000001e-05,
+      "loss": 1.7664,
+      "step": 4571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2872091428571428e-05,
+      "loss": 1.4861,
+      "step": 4572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2871822857142858e-05,
+      "loss": 1.8593,
+      "step": 4573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2871554285714287e-05,
+      "loss": 1.3123,
+      "step": 4574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2871285714285714e-05,
+      "loss": 1.2419,
+      "step": 4575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2871017142857144e-05,
+      "loss": 1.5117,
+      "step": 4576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2870748571428572e-05,
+      "loss": 1.4871,
+      "step": 4577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.287048e-05,
+      "loss": 1.4224,
+      "step": 4578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2870211428571428e-05,
+      "loss": 1.421,
+      "step": 4579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2869942857142858e-05,
+      "loss": 1.5137,
+      "step": 4580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2869674285714286e-05,
+      "loss": 1.3621,
+      "step": 4581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2869405714285715e-05,
+      "loss": 1.1133,
+      "step": 4582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2869137142857143e-05,
+      "loss": 1.6965,
+      "step": 4583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2868868571428573e-05,
+      "loss": 1.3817,
+      "step": 4584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2868599999999999e-05,
+      "loss": 1.3048,
+      "step": 4585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2868331428571429e-05,
+      "loss": 1.4356,
+      "step": 4586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2868062857142859e-05,
+      "loss": 1.5534,
+      "step": 4587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2867794285714285e-05,
+      "loss": 1.3702,
+      "step": 4588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2867525714285715e-05,
+      "loss": 1.3325,
+      "step": 4589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2867257142857143e-05,
+      "loss": 1.5825,
+      "step": 4590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2866988571428573e-05,
+      "loss": 1.5944,
+      "step": 4591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.286672e-05,
+      "loss": 1.1712,
+      "step": 4592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.286645142857143e-05,
+      "loss": 1.5677,
+      "step": 4593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2866182857142858e-05,
+      "loss": 0.9795,
+      "step": 4594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2865914285714286e-05,
+      "loss": 1.508,
+      "step": 4595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2865645714285714e-05,
+      "loss": 1.2136,
+      "step": 4596
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2865377142857144e-05,
+      "loss": 1.5157,
+      "step": 4597
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.286510857142857e-05,
+      "loss": 1.2936,
+      "step": 4598
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.286484e-05,
+      "loss": 1.4528,
+      "step": 4599
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.286457142857143e-05,
+      "loss": 1.5557,
+      "step": 4600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2864302857142857e-05,
+      "loss": 1.5559,
+      "step": 4601
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2864034285714287e-05,
+      "loss": 1.3292,
+      "step": 4602
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2863765714285715e-05,
+      "loss": 1.7499,
+      "step": 4603
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2863497142857145e-05,
+      "loss": 1.1137,
+      "step": 4604
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2863228571428571e-05,
+      "loss": 1.6817,
+      "step": 4605
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2862960000000001e-05,
+      "loss": 1.3125,
+      "step": 4606
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2862691428571429e-05,
+      "loss": 1.6612,
+      "step": 4607
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2862422857142857e-05,
+      "loss": 1.4641,
+      "step": 4608
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2862154285714285e-05,
+      "loss": 1.8394,
+      "step": 4609
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2861885714285715e-05,
+      "loss": 1.307,
+      "step": 4610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2861617142857142e-05,
+      "loss": 1.8364,
+      "step": 4611
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2861348571428572e-05,
+      "loss": 1.2349,
+      "step": 4612
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2861080000000002e-05,
+      "loss": 1.1724,
+      "step": 4613
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.286081142857143e-05,
+      "loss": 1.8434,
+      "step": 4614
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2860542857142858e-05,
+      "loss": 1.519,
+      "step": 4615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2860274285714286e-05,
+      "loss": 1.2677,
+      "step": 4616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2860005714285716e-05,
+      "loss": 1.4236,
+      "step": 4617
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2859737142857142e-05,
+      "loss": 1.3612,
+      "step": 4618
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2859468571428572e-05,
+      "loss": 1.6792,
+      "step": 4619
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.28592e-05,
+      "loss": 1.5913,
+      "step": 4620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2858931428571429e-05,
+      "loss": 1.5203,
+      "step": 4621
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2858662857142857e-05,
+      "loss": 1.5943,
+      "step": 4622
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2858394285714287e-05,
+      "loss": 1.6179,
+      "step": 4623
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2858125714285713e-05,
+      "loss": 1.4708,
+      "step": 4624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2857857142857143e-05,
+      "loss": 1.2458,
+      "step": 4625
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2857588571428573e-05,
+      "loss": 1.3155,
+      "step": 4626
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2857320000000001e-05,
+      "loss": 1.5281,
+      "step": 4627
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.285705142857143e-05,
+      "loss": 1.0639,
+      "step": 4628
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2856782857142857e-05,
+      "loss": 1.219,
+      "step": 4629
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2856514285714287e-05,
+      "loss": 1.4925,
+      "step": 4630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2856245714285714e-05,
+      "loss": 1.7651,
+      "step": 4631
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2855977142857144e-05,
+      "loss": 1.5028,
+      "step": 4632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2855708571428572e-05,
+      "loss": 1.3629,
+      "step": 4633
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.285544e-05,
+      "loss": 1.1721,
+      "step": 4634
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2855171428571428e-05,
+      "loss": 1.357,
+      "step": 4635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2854902857142858e-05,
+      "loss": 1.5223,
+      "step": 4636
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2854634285714286e-05,
+      "loss": 1.384,
+      "step": 4637
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2854365714285714e-05,
+      "loss": 1.4534,
+      "step": 4638
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2854097142857144e-05,
+      "loss": 1.699,
+      "step": 4639
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2853828571428572e-05,
+      "loss": 1.4123,
+      "step": 4640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.285356e-05,
+      "loss": 1.6071,
+      "step": 4641
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2853291428571429e-05,
+      "loss": 1.2025,
+      "step": 4642
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2853022857142859e-05,
+      "loss": 1.5227,
+      "step": 4643
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2852754285714285e-05,
+      "loss": 1.2917,
+      "step": 4644
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2852485714285715e-05,
+      "loss": 1.5518,
+      "step": 4645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2852217142857143e-05,
+      "loss": 1.5651,
+      "step": 4646
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2851948571428571e-05,
+      "loss": 1.6911,
+      "step": 4647
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.285168e-05,
+      "loss": 1.5921,
+      "step": 4648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.285141142857143e-05,
+      "loss": 1.471,
+      "step": 4649
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2851142857142858e-05,
+      "loss": 1.3551,
+      "step": 4650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2850874285714286e-05,
+      "loss": 1.5089,
+      "step": 4651
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2850605714285716e-05,
+      "loss": 1.8837,
+      "step": 4652
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2850337142857144e-05,
+      "loss": 1.4216,
+      "step": 4653
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2850068571428572e-05,
+      "loss": 1.1563,
+      "step": 4654
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.28498e-05,
+      "loss": 1.5191,
+      "step": 4655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.284953142857143e-05,
+      "loss": 1.4743,
+      "step": 4656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2849262857142857e-05,
+      "loss": 1.4647,
+      "step": 4657
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2848994285714286e-05,
+      "loss": 1.4438,
+      "step": 4658
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2848725714285715e-05,
+      "loss": 1.3585,
+      "step": 4659
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2848457142857144e-05,
+      "loss": 1.121,
+      "step": 4660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2848188571428571e-05,
+      "loss": 1.6054,
+      "step": 4661
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.284792e-05,
+      "loss": 1.9108,
+      "step": 4662
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2847651428571429e-05,
+      "loss": 1.6041,
+      "step": 4663
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2847382857142857e-05,
+      "loss": 1.0648,
+      "step": 4664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2847114285714287e-05,
+      "loss": 1.6247,
+      "step": 4665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2846845714285715e-05,
+      "loss": 1.4199,
+      "step": 4666
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2846577142857143e-05,
+      "loss": 1.6028,
+      "step": 4667
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2846308571428572e-05,
+      "loss": 1.2365,
+      "step": 4668
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2846040000000001e-05,
+      "loss": 1.1768,
+      "step": 4669
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2845771428571428e-05,
+      "loss": 1.1749,
+      "step": 4670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2845502857142858e-05,
+      "loss": 1.2007,
+      "step": 4671
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2845234285714286e-05,
+      "loss": 1.4487,
+      "step": 4672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2844965714285716e-05,
+      "loss": 1.2879,
+      "step": 4673
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2844697142857142e-05,
+      "loss": 1.4184,
+      "step": 4674
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2844428571428572e-05,
+      "loss": 1.7775,
+      "step": 4675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.284416e-05,
+      "loss": 1.0422,
+      "step": 4676
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2843891428571429e-05,
+      "loss": 1.7205,
+      "step": 4677
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2843622857142857e-05,
+      "loss": 1.4093,
+      "step": 4678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2843354285714287e-05,
+      "loss": 1.4011,
+      "step": 4679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2843085714285715e-05,
+      "loss": 1.3526,
+      "step": 4680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2842817142857143e-05,
+      "loss": 1.2904,
+      "step": 4681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2842548571428573e-05,
+      "loss": 1.3359,
+      "step": 4682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2842280000000001e-05,
+      "loss": 1.3855,
+      "step": 4683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.284201142857143e-05,
+      "loss": 1.3133,
+      "step": 4684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2841742857142857e-05,
+      "loss": 1.4722,
+      "step": 4685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2841474285714287e-05,
+      "loss": 1.5443,
+      "step": 4686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2841205714285714e-05,
+      "loss": 1.2839,
+      "step": 4687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2840937142857144e-05,
+      "loss": 1.2416,
+      "step": 4688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2840668571428572e-05,
+      "loss": 1.5015,
+      "step": 4689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.28404e-05,
+      "loss": 1.2756,
+      "step": 4690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2840131428571428e-05,
+      "loss": 1.312,
+      "step": 4691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2839862857142858e-05,
+      "loss": 1.4845,
+      "step": 4692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2839594285714286e-05,
+      "loss": 1.3292,
+      "step": 4693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2839325714285714e-05,
+      "loss": 1.5686,
+      "step": 4694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2839057142857144e-05,
+      "loss": 1.7162,
+      "step": 4695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2838788571428572e-05,
+      "loss": 1.3873,
+      "step": 4696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283852e-05,
+      "loss": 1.5435,
+      "step": 4697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2838251428571429e-05,
+      "loss": 1.6025,
+      "step": 4698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2837982857142859e-05,
+      "loss": 1.2654,
+      "step": 4699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2837714285714285e-05,
+      "loss": 1.3421,
+      "step": 4700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2837445714285715e-05,
+      "loss": 1.4077,
+      "step": 4701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2837177142857143e-05,
+      "loss": 1.014,
+      "step": 4702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2836908571428571e-05,
+      "loss": 1.1316,
+      "step": 4703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283664e-05,
+      "loss": 0.9442,
+      "step": 4704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283637142857143e-05,
+      "loss": 1.5147,
+      "step": 4705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283610285714286e-05,
+      "loss": 1.325,
+      "step": 4706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2835834285714286e-05,
+      "loss": 1.3803,
+      "step": 4707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2835565714285716e-05,
+      "loss": 1.4763,
+      "step": 4708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2835297142857144e-05,
+      "loss": 1.3787,
+      "step": 4709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2835028571428572e-05,
+      "loss": 1.6404,
+      "step": 4710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283476e-05,
+      "loss": 1.403,
+      "step": 4711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283449142857143e-05,
+      "loss": 1.5944,
+      "step": 4712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2834222857142856e-05,
+      "loss": 1.0341,
+      "step": 4713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2833954285714286e-05,
+      "loss": 1.5626,
+      "step": 4714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2833685714285714e-05,
+      "loss": 1.3903,
+      "step": 4715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2833417142857143e-05,
+      "loss": 1.7113,
+      "step": 4716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283314857142857e-05,
+      "loss": 1.6068,
+      "step": 4717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283288e-05,
+      "loss": 1.8997,
+      "step": 4718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.283261142857143e-05,
+      "loss": 1.2777,
+      "step": 4719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2832342857142857e-05,
+      "loss": 1.308,
+      "step": 4720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2832074285714287e-05,
+      "loss": 1.5337,
+      "step": 4721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2831805714285715e-05,
+      "loss": 1.5742,
+      "step": 4722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2831537142857143e-05,
+      "loss": 1.2733,
+      "step": 4723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2831268571428571e-05,
+      "loss": 1.8662,
+      "step": 4724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2831000000000001e-05,
+      "loss": 1.8176,
+      "step": 4725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2830731428571428e-05,
+      "loss": 1.3025,
+      "step": 4726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2830462857142858e-05,
+      "loss": 1.0686,
+      "step": 4727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2830194285714286e-05,
+      "loss": 1.2079,
+      "step": 4728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2829925714285716e-05,
+      "loss": 1.5894,
+      "step": 4729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2829657142857142e-05,
+      "loss": 1.6631,
+      "step": 4730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2829388571428572e-05,
+      "loss": 1.3757,
+      "step": 4731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2829120000000002e-05,
+      "loss": 1.3431,
+      "step": 4732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2828851428571428e-05,
+      "loss": 1.3857,
+      "step": 4733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2828582857142858e-05,
+      "loss": 1.2399,
+      "step": 4734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2828314285714286e-05,
+      "loss": 1.4658,
+      "step": 4735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2828045714285715e-05,
+      "loss": 1.4675,
+      "step": 4736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2827777142857143e-05,
+      "loss": 1.0538,
+      "step": 4737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2827508571428573e-05,
+      "loss": 1.4677,
+      "step": 4738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.282724e-05,
+      "loss": 1.6385,
+      "step": 4739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2826971428571429e-05,
+      "loss": 1.9016,
+      "step": 4740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2826702857142857e-05,
+      "loss": 1.3125,
+      "step": 4741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2826434285714287e-05,
+      "loss": 1.5387,
+      "step": 4742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2826165714285714e-05,
+      "loss": 1.3136,
+      "step": 4743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2825897142857143e-05,
+      "loss": 1.4073,
+      "step": 4744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2825628571428573e-05,
+      "loss": 1.8872,
+      "step": 4745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.282536e-05,
+      "loss": 1.5231,
+      "step": 4746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.282509142857143e-05,
+      "loss": 1.2402,
+      "step": 4747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2824822857142858e-05,
+      "loss": 1.5169,
+      "step": 4748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2824554285714286e-05,
+      "loss": 1.4082,
+      "step": 4749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2824285714285714e-05,
+      "loss": 1.2041,
+      "step": 4750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2824017142857144e-05,
+      "loss": 1.2398,
+      "step": 4751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.282374857142857e-05,
+      "loss": 1.4877,
+      "step": 4752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.282348e-05,
+      "loss": 1.3561,
+      "step": 4753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2823211428571429e-05,
+      "loss": 1.0613,
+      "step": 4754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2822942857142858e-05,
+      "loss": 1.3121,
+      "step": 4755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2822674285714285e-05,
+      "loss": 1.5085,
+      "step": 4756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2822405714285715e-05,
+      "loss": 1.6702,
+      "step": 4757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2822137142857145e-05,
+      "loss": 1.5026,
+      "step": 4758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2821868571428571e-05,
+      "loss": 1.5273,
+      "step": 4759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2821600000000001e-05,
+      "loss": 1.4301,
+      "step": 4760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.282133142857143e-05,
+      "loss": 1.2372,
+      "step": 4761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2821062857142857e-05,
+      "loss": 1.1656,
+      "step": 4762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2820794285714286e-05,
+      "loss": 1.4169,
+      "step": 4763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2820525714285715e-05,
+      "loss": 1.5535,
+      "step": 4764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2820257142857144e-05,
+      "loss": 1.5984,
+      "step": 4765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2819988571428572e-05,
+      "loss": 1.479,
+      "step": 4766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.281972e-05,
+      "loss": 1.4158,
+      "step": 4767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.281945142857143e-05,
+      "loss": 1.384,
+      "step": 4768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2819182857142856e-05,
+      "loss": 1.3356,
+      "step": 4769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2818914285714286e-05,
+      "loss": 1.0827,
+      "step": 4770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2818645714285716e-05,
+      "loss": 1.3749,
+      "step": 4771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2818377142857143e-05,
+      "loss": 1.7785,
+      "step": 4772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2818108571428572e-05,
+      "loss": 1.595,
+      "step": 4773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.281784e-05,
+      "loss": 1.3519,
+      "step": 4774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2817571428571429e-05,
+      "loss": 1.1457,
+      "step": 4775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2817302857142857e-05,
+      "loss": 1.6402,
+      "step": 4776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2817034285714287e-05,
+      "loss": 1.4525,
+      "step": 4777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2816765714285715e-05,
+      "loss": 1.2469,
+      "step": 4778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2816497142857143e-05,
+      "loss": 1.4315,
+      "step": 4779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2816228571428571e-05,
+      "loss": 1.3496,
+      "step": 4780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2815960000000001e-05,
+      "loss": 1.1237,
+      "step": 4781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2815691428571428e-05,
+      "loss": 1.2178,
+      "step": 4782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2815422857142858e-05,
+      "loss": 1.3698,
+      "step": 4783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2815154285714286e-05,
+      "loss": 1.8066,
+      "step": 4784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2814885714285714e-05,
+      "loss": 1.7737,
+      "step": 4785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2814617142857144e-05,
+      "loss": 1.3903,
+      "step": 4786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2814348571428572e-05,
+      "loss": 1.1473,
+      "step": 4787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2814080000000002e-05,
+      "loss": 1.2666,
+      "step": 4788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2813811428571428e-05,
+      "loss": 1.294,
+      "step": 4789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2813542857142858e-05,
+      "loss": 1.1807,
+      "step": 4790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2813274285714286e-05,
+      "loss": 1.2905,
+      "step": 4791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2813005714285715e-05,
+      "loss": 1.5109,
+      "step": 4792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2812737142857143e-05,
+      "loss": 1.8141,
+      "step": 4793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2812468571428573e-05,
+      "loss": 1.6573,
+      "step": 4794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2812199999999999e-05,
+      "loss": 1.2288,
+      "step": 4795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2811931428571429e-05,
+      "loss": 1.7301,
+      "step": 4796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2811662857142857e-05,
+      "loss": 1.3091,
+      "step": 4797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2811394285714285e-05,
+      "loss": 1.2911,
+      "step": 4798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2811125714285715e-05,
+      "loss": 1.6548,
+      "step": 4799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2810857142857143e-05,
+      "loss": 1.5639,
+      "step": 4800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2810588571428573e-05,
+      "loss": 1.2964,
+      "step": 4801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.281032e-05,
+      "loss": 1.5358,
+      "step": 4802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.281005142857143e-05,
+      "loss": 1.4818,
+      "step": 4803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2809782857142858e-05,
+      "loss": 1.3975,
+      "step": 4804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2809514285714286e-05,
+      "loss": 1.3701,
+      "step": 4805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2809245714285714e-05,
+      "loss": 1.4829,
+      "step": 4806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2808977142857144e-05,
+      "loss": 1.4932,
+      "step": 4807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.280870857142857e-05,
+      "loss": 1.2442,
+      "step": 4808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.280844e-05,
+      "loss": 1.7158,
+      "step": 4809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2808171428571428e-05,
+      "loss": 0.9745,
+      "step": 4810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2807902857142858e-05,
+      "loss": 1.7092,
+      "step": 4811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.2807634285714287e-05,
+      "loss": 1.3632,
+      "step": 4812
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2807365714285715e-05,
+      "loss": 1.1503,
+      "step": 4813
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2807097142857145e-05,
+      "loss": 1.2065,
+      "step": 4814
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2806828571428571e-05,
+      "loss": 1.9193,
+      "step": 4815
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2806560000000001e-05,
+      "loss": 1.4523,
+      "step": 4816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2806291428571429e-05,
+      "loss": 1.3134,
+      "step": 4817
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2806022857142857e-05,
+      "loss": 1.7088,
+      "step": 4818
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2805754285714285e-05,
+      "loss": 1.4388,
+      "step": 4819
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2805485714285715e-05,
+      "loss": 1.2765,
+      "step": 4820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2805217142857142e-05,
+      "loss": 1.6056,
+      "step": 4821
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2804948571428572e-05,
+      "loss": 1.2447,
+      "step": 4822
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.280468e-05,
+      "loss": 1.8945,
+      "step": 4823
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.280441142857143e-05,
+      "loss": 1.5295,
+      "step": 4824
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2804142857142858e-05,
+      "loss": 1.7919,
+      "step": 4825
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2803874285714286e-05,
+      "loss": 1.3263,
+      "step": 4826
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2803605714285716e-05,
+      "loss": 1.6075,
+      "step": 4827
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2803337142857142e-05,
+      "loss": 1.5652,
+      "step": 4828
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2803068571428572e-05,
+      "loss": 1.4313,
+      "step": 4829
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.28028e-05,
+      "loss": 1.3123,
+      "step": 4830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2802531428571429e-05,
+      "loss": 1.322,
+      "step": 4831
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2802262857142857e-05,
+      "loss": 1.3698,
+      "step": 4832
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2801994285714287e-05,
+      "loss": 1.9567,
+      "step": 4833
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2801725714285715e-05,
+      "loss": 1.5864,
+      "step": 4834
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2801457142857143e-05,
+      "loss": 1.4438,
+      "step": 4835
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2801188571428571e-05,
+      "loss": 1.7798,
+      "step": 4836
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2800920000000001e-05,
+      "loss": 1.3868,
+      "step": 4837
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.280065142857143e-05,
+      "loss": 1.5157,
+      "step": 4838
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2800382857142857e-05,
+      "loss": 1.5821,
+      "step": 4839
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2800114285714287e-05,
+      "loss": 1.4923,
+      "step": 4840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2799845714285714e-05,
+      "loss": 1.5184,
+      "step": 4841
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2799577142857144e-05,
+      "loss": 1.5399,
+      "step": 4842
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2799308571428572e-05,
+      "loss": 1.6994,
+      "step": 4843
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.279904e-05,
+      "loss": 1.5423,
+      "step": 4844
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2798771428571428e-05,
+      "loss": 1.4434,
+      "step": 4845
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2798502857142858e-05,
+      "loss": 1.1813,
+      "step": 4846
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2798234285714286e-05,
+      "loss": 1.544,
+      "step": 4847
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2797965714285714e-05,
+      "loss": 1.2771,
+      "step": 4848
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2797697142857143e-05,
+      "loss": 1.6075,
+      "step": 4849
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2797428571428572e-05,
+      "loss": 1.351,
+      "step": 4850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.279716e-05,
+      "loss": 1.4886,
+      "step": 4851
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2796891428571429e-05,
+      "loss": 1.0181,
+      "step": 4852
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2796622857142859e-05,
+      "loss": 1.2453,
+      "step": 4853
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2796354285714285e-05,
+      "loss": 1.2512,
+      "step": 4854
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2796085714285715e-05,
+      "loss": 1.372,
+      "step": 4855
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2795817142857143e-05,
+      "loss": 1.7073,
+      "step": 4856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2795548571428573e-05,
+      "loss": 1.6189,
+      "step": 4857
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.279528e-05,
+      "loss": 1.9295,
+      "step": 4858
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.279501142857143e-05,
+      "loss": 1.3918,
+      "step": 4859
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2794742857142858e-05,
+      "loss": 1.5285,
+      "step": 4860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2794474285714286e-05,
+      "loss": 1.2506,
+      "step": 4861
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2794205714285714e-05,
+      "loss": 1.2533,
+      "step": 4862
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2793937142857144e-05,
+      "loss": 1.5579,
+      "step": 4863
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2793668571428572e-05,
+      "loss": 1.4049,
+      "step": 4864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.27934e-05,
+      "loss": 1.7141,
+      "step": 4865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.279313142857143e-05,
+      "loss": 1.756,
+      "step": 4866
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2792862857142857e-05,
+      "loss": 1.6292,
+      "step": 4867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2792594285714286e-05,
+      "loss": 1.27,
+      "step": 4868
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2792325714285715e-05,
+      "loss": 1.2941,
+      "step": 4869
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2792057142857144e-05,
+      "loss": 1.3633,
+      "step": 4870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2791788571428571e-05,
+      "loss": 1.3483,
+      "step": 4871
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.279152e-05,
+      "loss": 1.6794,
+      "step": 4872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2791251428571429e-05,
+      "loss": 1.6734,
+      "step": 4873
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2790982857142857e-05,
+      "loss": 1.5731,
+      "step": 4874
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2790714285714285e-05,
+      "loss": 1.4676,
+      "step": 4875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2790445714285715e-05,
+      "loss": 1.6348,
+      "step": 4876
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2790177142857143e-05,
+      "loss": 1.3744,
+      "step": 4877
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2789908571428572e-05,
+      "loss": 1.2715,
+      "step": 4878
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2789640000000001e-05,
+      "loss": 1.5435,
+      "step": 4879
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.278937142857143e-05,
+      "loss": 1.5501,
+      "step": 4880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2789102857142858e-05,
+      "loss": 1.4056,
+      "step": 4881
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2788834285714286e-05,
+      "loss": 1.5521,
+      "step": 4882
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2788565714285716e-05,
+      "loss": 1.4156,
+      "step": 4883
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2788297142857142e-05,
+      "loss": 1.5192,
+      "step": 4884
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2788028571428572e-05,
+      "loss": 1.2531,
+      "step": 4885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.278776e-05,
+      "loss": 1.0894,
+      "step": 4886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2787491428571429e-05,
+      "loss": 1.4136,
+      "step": 4887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2787222857142857e-05,
+      "loss": 1.5145,
+      "step": 4888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2786954285714287e-05,
+      "loss": 1.2899,
+      "step": 4889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2786685714285715e-05,
+      "loss": 1.6789,
+      "step": 4890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2786417142857143e-05,
+      "loss": 1.5445,
+      "step": 4891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2786148571428573e-05,
+      "loss": 1.7187,
+      "step": 4892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2785880000000001e-05,
+      "loss": 1.4477,
+      "step": 4893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.278561142857143e-05,
+      "loss": 1.4,
+      "step": 4894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2785342857142857e-05,
+      "loss": 1.1373,
+      "step": 4895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2785074285714287e-05,
+      "loss": 1.2587,
+      "step": 4896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2784805714285714e-05,
+      "loss": 1.7387,
+      "step": 4897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2784537142857144e-05,
+      "loss": 1.5343,
+      "step": 4898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2784268571428572e-05,
+      "loss": 1.5635,
+      "step": 4899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2784e-05,
+      "loss": 1.3687,
+      "step": 4900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2783731428571428e-05,
+      "loss": 1.3409,
+      "step": 4901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2783462857142858e-05,
+      "loss": 1.6706,
+      "step": 4902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2783194285714286e-05,
+      "loss": 1.4041,
+      "step": 4903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2782925714285714e-05,
+      "loss": 1.5795,
+      "step": 4904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2782657142857144e-05,
+      "loss": 1.471,
+      "step": 4905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2782388571428572e-05,
+      "loss": 1.2187,
+      "step": 4906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.278212e-05,
+      "loss": 1.4593,
+      "step": 4907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2781851428571429e-05,
+      "loss": 1.6636,
+      "step": 4908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2781582857142859e-05,
+      "loss": 1.582,
+      "step": 4909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2781314285714285e-05,
+      "loss": 1.5482,
+      "step": 4910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2781045714285715e-05,
+      "loss": 1.4943,
+      "step": 4911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2780777142857143e-05,
+      "loss": 1.204,
+      "step": 4912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2780508571428571e-05,
+      "loss": 1.4538,
+      "step": 4913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.278024e-05,
+      "loss": 1.6651,
+      "step": 4914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.277997142857143e-05,
+      "loss": 0.9574,
+      "step": 4915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2779702857142858e-05,
+      "loss": 1.369,
+      "step": 4916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2779434285714286e-05,
+      "loss": 1.3546,
+      "step": 4917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2779165714285716e-05,
+      "loss": 1.3009,
+      "step": 4918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2778897142857144e-05,
+      "loss": 1.4609,
+      "step": 4919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2778628571428572e-05,
+      "loss": 1.5422,
+      "step": 4920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.277836e-05,
+      "loss": 1.2845,
+      "step": 4921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.277809142857143e-05,
+      "loss": 1.4476,
+      "step": 4922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2777822857142856e-05,
+      "loss": 1.5708,
+      "step": 4923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2777554285714286e-05,
+      "loss": 1.9186,
+      "step": 4924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2777285714285714e-05,
+      "loss": 1.4421,
+      "step": 4925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2777017142857143e-05,
+      "loss": 1.5147,
+      "step": 4926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.277674857142857e-05,
+      "loss": 1.2007,
+      "step": 4927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.277648e-05,
+      "loss": 1.6801,
+      "step": 4928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2776211428571429e-05,
+      "loss": 1.3908,
+      "step": 4929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2775942857142857e-05,
+      "loss": 1.3251,
+      "step": 4930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2775674285714287e-05,
+      "loss": 1.5303,
+      "step": 4931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2775405714285715e-05,
+      "loss": 1.5611,
+      "step": 4932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2775137142857143e-05,
+      "loss": 1.6507,
+      "step": 4933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2774868571428571e-05,
+      "loss": 1.4077,
+      "step": 4934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2774600000000001e-05,
+      "loss": 1.6559,
+      "step": 4935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2774331428571428e-05,
+      "loss": 1.7459,
+      "step": 4936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2774062857142858e-05,
+      "loss": 1.8151,
+      "step": 4937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2773794285714286e-05,
+      "loss": 1.2476,
+      "step": 4938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2773525714285716e-05,
+      "loss": 1.0886,
+      "step": 4939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2773257142857142e-05,
+      "loss": 1.5479,
+      "step": 4940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2772988571428572e-05,
+      "loss": 1.2961,
+      "step": 4941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.277272e-05,
+      "loss": 1.2203,
+      "step": 4942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2772451428571428e-05,
+      "loss": 1.4329,
+      "step": 4943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2772182857142858e-05,
+      "loss": 1.4846,
+      "step": 4944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2771914285714286e-05,
+      "loss": 1.4108,
+      "step": 4945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2771645714285715e-05,
+      "loss": 1.5581,
+      "step": 4946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2771377142857143e-05,
+      "loss": 1.6427,
+      "step": 4947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2771108571428573e-05,
+      "loss": 1.8688,
+      "step": 4948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.277084e-05,
+      "loss": 1.0343,
+      "step": 4949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2770571428571429e-05,
+      "loss": 1.2046,
+      "step": 4950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2770302857142857e-05,
+      "loss": 1.4088,
+      "step": 4951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2770034285714287e-05,
+      "loss": 1.3342,
+      "step": 4952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2769765714285714e-05,
+      "loss": 1.3997,
+      "step": 4953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2769497142857143e-05,
+      "loss": 1.1668,
+      "step": 4954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2769228571428572e-05,
+      "loss": 1.266,
+      "step": 4955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276896e-05,
+      "loss": 1.709,
+      "step": 4956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276869142857143e-05,
+      "loss": 1.2697,
+      "step": 4957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2768422857142858e-05,
+      "loss": 1.3653,
+      "step": 4958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2768154285714286e-05,
+      "loss": 1.4633,
+      "step": 4959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2767885714285714e-05,
+      "loss": 1.3124,
+      "step": 4960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2767617142857144e-05,
+      "loss": 1.4533,
+      "step": 4961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2767348571428572e-05,
+      "loss": 1.5337,
+      "step": 4962
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276708e-05,
+      "loss": 1.5307,
+      "step": 4963
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2766811428571429e-05,
+      "loss": 1.15,
+      "step": 4964
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2766542857142858e-05,
+      "loss": 1.3812,
+      "step": 4965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2766274285714285e-05,
+      "loss": 1.206,
+      "step": 4966
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2766005714285715e-05,
+      "loss": 1.3159,
+      "step": 4967
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2765737142857143e-05,
+      "loss": 1.3593,
+      "step": 4968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2765468571428571e-05,
+      "loss": 1.6093,
+      "step": 4969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2765200000000001e-05,
+      "loss": 1.6096,
+      "step": 4970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276493142857143e-05,
+      "loss": 1.4457,
+      "step": 4971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2764662857142857e-05,
+      "loss": 1.5288,
+      "step": 4972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2764394285714286e-05,
+      "loss": 1.3753,
+      "step": 4973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2764125714285715e-05,
+      "loss": 1.8228,
+      "step": 4974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2763857142857144e-05,
+      "loss": 1.4697,
+      "step": 4975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2763588571428572e-05,
+      "loss": 1.5769,
+      "step": 4976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276332e-05,
+      "loss": 1.0743,
+      "step": 4977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276305142857143e-05,
+      "loss": 1.5557,
+      "step": 4978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2762782857142856e-05,
+      "loss": 1.5244,
+      "step": 4979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2762514285714286e-05,
+      "loss": 1.4702,
+      "step": 4980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2762245714285714e-05,
+      "loss": 1.2704,
+      "step": 4981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2761977142857143e-05,
+      "loss": 1.7436,
+      "step": 4982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2761708571428572e-05,
+      "loss": 1.5126,
+      "step": 4983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276144e-05,
+      "loss": 1.2738,
+      "step": 4984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.276117142857143e-05,
+      "loss": 1.3412,
+      "step": 4985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2760902857142857e-05,
+      "loss": 1.4877,
+      "step": 4986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.2760634285714287e-05,
+      "loss": 1.6624,
+      "step": 4987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2760365714285715e-05,
+      "loss": 1.5372,
+      "step": 4988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2760097142857143e-05,
+      "loss": 1.198,
+      "step": 4989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2759828571428571e-05,
+      "loss": 1.1666,
+      "step": 4990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2759560000000001e-05,
+      "loss": 1.5567,
+      "step": 4991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2759291428571428e-05,
+      "loss": 1.1636,
+      "step": 4992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2759022857142858e-05,
+      "loss": 1.588,
+      "step": 4993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2758754285714286e-05,
+      "loss": 1.7501,
+      "step": 4994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2758485714285714e-05,
+      "loss": 1.7781,
+      "step": 4995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2758217142857144e-05,
+      "loss": 1.4517,
+      "step": 4996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2757948571428572e-05,
+      "loss": 1.4996,
+      "step": 4997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2757680000000002e-05,
+      "loss": 1.5635,
+      "step": 4998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2757411428571428e-05,
+      "loss": 1.5233,
+      "step": 4999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2757142857142858e-05,
+      "loss": 1.4358,
+      "step": 5000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2756874285714286e-05,
+      "loss": 1.1897,
+      "step": 5001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2756605714285715e-05,
+      "loss": 1.6212,
+      "step": 5002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2756337142857143e-05,
+      "loss": 1.5188,
+      "step": 5003
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2756068571428573e-05,
+      "loss": 1.0,
+      "step": 5004
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2755799999999999e-05,
+      "loss": 1.9433,
+      "step": 5005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2755531428571429e-05,
+      "loss": 1.6572,
+      "step": 5006
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2755262857142857e-05,
+      "loss": 1.1885,
+      "step": 5007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2754994285714287e-05,
+      "loss": 1.3592,
+      "step": 5008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2754725714285715e-05,
+      "loss": 1.4212,
+      "step": 5009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2754457142857143e-05,
+      "loss": 1.3972,
+      "step": 5010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2754188571428573e-05,
+      "loss": 1.6725,
+      "step": 5011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.275392e-05,
+      "loss": 1.3772,
+      "step": 5012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.275365142857143e-05,
+      "loss": 1.4389,
+      "step": 5013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2753382857142858e-05,
+      "loss": 1.3795,
+      "step": 5014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2753114285714286e-05,
+      "loss": 1.4728,
+      "step": 5015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2752845714285714e-05,
+      "loss": 1.2176,
+      "step": 5016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2752577142857144e-05,
+      "loss": 1.3769,
+      "step": 5017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.275230857142857e-05,
+      "loss": 1.4496,
+      "step": 5018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.275204e-05,
+      "loss": 1.4296,
+      "step": 5019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2751771428571428e-05,
+      "loss": 1.562,
+      "step": 5020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2751502857142858e-05,
+      "loss": 1.4657,
+      "step": 5021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2751234285714287e-05,
+      "loss": 1.539,
+      "step": 5022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2750965714285715e-05,
+      "loss": 1.526,
+      "step": 5023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2750697142857145e-05,
+      "loss": 1.4328,
+      "step": 5024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2750428571428571e-05,
+      "loss": 1.5961,
+      "step": 5025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2750160000000001e-05,
+      "loss": 1.6812,
+      "step": 5026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2749891428571429e-05,
+      "loss": 1.2354,
+      "step": 5027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2749622857142857e-05,
+      "loss": 1.4492,
+      "step": 5028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2749354285714285e-05,
+      "loss": 1.3839,
+      "step": 5029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2749085714285715e-05,
+      "loss": 1.6514,
+      "step": 5030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2748817142857144e-05,
+      "loss": 1.5416,
+      "step": 5031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2748548571428572e-05,
+      "loss": 1.3745,
+      "step": 5032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.274828e-05,
+      "loss": 1.5406,
+      "step": 5033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.274801142857143e-05,
+      "loss": 1.3017,
+      "step": 5034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2747742857142858e-05,
+      "loss": 1.1107,
+      "step": 5035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2747474285714286e-05,
+      "loss": 1.3425,
+      "step": 5036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2747205714285716e-05,
+      "loss": 1.0991,
+      "step": 5037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2746937142857142e-05,
+      "loss": 1.1678,
+      "step": 5038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2746668571428572e-05,
+      "loss": 1.1777,
+      "step": 5039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.27464e-05,
+      "loss": 1.2347,
+      "step": 5040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2746131428571429e-05,
+      "loss": 1.5924,
+      "step": 5041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2745862857142857e-05,
+      "loss": 2.0016,
+      "step": 5042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2745594285714287e-05,
+      "loss": 1.4053,
+      "step": 5043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2745325714285715e-05,
+      "loss": 1.429,
+      "step": 5044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2745057142857143e-05,
+      "loss": 1.5201,
+      "step": 5045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2744788571428571e-05,
+      "loss": 1.6269,
+      "step": 5046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2744520000000001e-05,
+      "loss": 1.3164,
+      "step": 5047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.274425142857143e-05,
+      "loss": 1.4117,
+      "step": 5048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2743982857142857e-05,
+      "loss": 1.3998,
+      "step": 5049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2743714285714287e-05,
+      "loss": 1.5012,
+      "step": 5050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2743445714285714e-05,
+      "loss": 1.8338,
+      "step": 5051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2743177142857144e-05,
+      "loss": 1.5075,
+      "step": 5052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2742908571428572e-05,
+      "loss": 1.4865,
+      "step": 5053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2742640000000002e-05,
+      "loss": 1.333,
+      "step": 5054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2742371428571428e-05,
+      "loss": 1.5855,
+      "step": 5055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2742102857142858e-05,
+      "loss": 1.188,
+      "step": 5056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2741834285714286e-05,
+      "loss": 1.1512,
+      "step": 5057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2741565714285714e-05,
+      "loss": 1.3451,
+      "step": 5058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2741297142857143e-05,
+      "loss": 1.3295,
+      "step": 5059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2741028571428572e-05,
+      "loss": 1.4376,
+      "step": 5060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.274076e-05,
+      "loss": 1.5875,
+      "step": 5061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2740491428571429e-05,
+      "loss": 1.4743,
+      "step": 5062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2740222857142859e-05,
+      "loss": 1.189,
+      "step": 5063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2739954285714285e-05,
+      "loss": 1.3659,
+      "step": 5064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2739685714285715e-05,
+      "loss": 1.5416,
+      "step": 5065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2739417142857143e-05,
+      "loss": 1.8211,
+      "step": 5066
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2739148571428573e-05,
+      "loss": 1.545,
+      "step": 5067
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.273888e-05,
+      "loss": 1.4058,
+      "step": 5068
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.273861142857143e-05,
+      "loss": 1.7832,
+      "step": 5069
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2738342857142858e-05,
+      "loss": 1.4052,
+      "step": 5070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2738074285714286e-05,
+      "loss": 1.7147,
+      "step": 5071
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2737805714285714e-05,
+      "loss": 1.8046,
+      "step": 5072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2737537142857144e-05,
+      "loss": 1.4546,
+      "step": 5073
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2737268571428572e-05,
+      "loss": 1.3578,
+      "step": 5074
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2737e-05,
+      "loss": 1.1975,
+      "step": 5075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.273673142857143e-05,
+      "loss": 1.5389,
+      "step": 5076
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2736462857142857e-05,
+      "loss": 1.0546,
+      "step": 5077
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2736194285714286e-05,
+      "loss": 1.0077,
+      "step": 5078
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2735925714285715e-05,
+      "loss": 1.7588,
+      "step": 5079
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2735657142857144e-05,
+      "loss": 1.4269,
+      "step": 5080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2735388571428571e-05,
+      "loss": 1.5092,
+      "step": 5081
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.273512e-05,
+      "loss": 1.1924,
+      "step": 5082
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2734851428571429e-05,
+      "loss": 1.671,
+      "step": 5083
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2734582857142857e-05,
+      "loss": 1.1329,
+      "step": 5084
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2734314285714285e-05,
+      "loss": 1.8204,
+      "step": 5085
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2734045714285715e-05,
+      "loss": 1.49,
+      "step": 5086
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2733777142857143e-05,
+      "loss": 1.3374,
+      "step": 5087
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2733508571428572e-05,
+      "loss": 1.2996,
+      "step": 5088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2733240000000001e-05,
+      "loss": 1.5676,
+      "step": 5089
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.273297142857143e-05,
+      "loss": 1.6619,
+      "step": 5090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2732702857142858e-05,
+      "loss": 1.3984,
+      "step": 5091
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2732434285714286e-05,
+      "loss": 1.2419,
+      "step": 5092
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2732165714285716e-05,
+      "loss": 1.2179,
+      "step": 5093
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2731897142857142e-05,
+      "loss": 1.7697,
+      "step": 5094
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2731628571428572e-05,
+      "loss": 1.5243,
+      "step": 5095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.273136e-05,
+      "loss": 1.8016,
+      "step": 5096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2731091428571429e-05,
+      "loss": 1.5043,
+      "step": 5097
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2730822857142857e-05,
+      "loss": 1.1169,
+      "step": 5098
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2730554285714287e-05,
+      "loss": 1.1919,
+      "step": 5099
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2730285714285715e-05,
+      "loss": 1.607,
+      "step": 5100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2730017142857143e-05,
+      "loss": 1.467,
+      "step": 5101
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2729748571428573e-05,
+      "loss": 1.3891,
+      "step": 5102
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2729480000000001e-05,
+      "loss": 1.3468,
+      "step": 5103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272921142857143e-05,
+      "loss": 1.3471,
+      "step": 5104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2728942857142857e-05,
+      "loss": 1.157,
+      "step": 5105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2728674285714287e-05,
+      "loss": 1.5064,
+      "step": 5106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2728405714285714e-05,
+      "loss": 1.5066,
+      "step": 5107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2728137142857144e-05,
+      "loss": 1.5299,
+      "step": 5108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2727868571428572e-05,
+      "loss": 1.4362,
+      "step": 5109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.27276e-05,
+      "loss": 1.4653,
+      "step": 5110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2727331428571428e-05,
+      "loss": 1.223,
+      "step": 5111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2727062857142858e-05,
+      "loss": 1.5621,
+      "step": 5112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2726794285714286e-05,
+      "loss": 1.2041,
+      "step": 5113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2726525714285714e-05,
+      "loss": 1.4523,
+      "step": 5114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2726257142857144e-05,
+      "loss": 1.3115,
+      "step": 5115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2725988571428572e-05,
+      "loss": 1.5064,
+      "step": 5116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272572e-05,
+      "loss": 1.5855,
+      "step": 5117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2725451428571429e-05,
+      "loss": 1.344,
+      "step": 5118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2725182857142859e-05,
+      "loss": 1.5341,
+      "step": 5119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2724914285714285e-05,
+      "loss": 1.5523,
+      "step": 5120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2724645714285715e-05,
+      "loss": 1.6393,
+      "step": 5121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2724377142857143e-05,
+      "loss": 1.2831,
+      "step": 5122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2724108571428571e-05,
+      "loss": 1.3023,
+      "step": 5123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272384e-05,
+      "loss": 1.5989,
+      "step": 5124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272357142857143e-05,
+      "loss": 1.5853,
+      "step": 5125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2723302857142858e-05,
+      "loss": 1.2797,
+      "step": 5126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2723034285714286e-05,
+      "loss": 1.6129,
+      "step": 5127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2722765714285716e-05,
+      "loss": 1.5459,
+      "step": 5128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2722497142857144e-05,
+      "loss": 1.0199,
+      "step": 5129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2722228571428572e-05,
+      "loss": 0.9667,
+      "step": 5130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272196e-05,
+      "loss": 1.3724,
+      "step": 5131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272169142857143e-05,
+      "loss": 1.5914,
+      "step": 5132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2721422857142856e-05,
+      "loss": 1.0817,
+      "step": 5133
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2721154285714286e-05,
+      "loss": 1.3013,
+      "step": 5134
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2720885714285714e-05,
+      "loss": 1.4651,
+      "step": 5135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2720617142857144e-05,
+      "loss": 1.1628,
+      "step": 5136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272034857142857e-05,
+      "loss": 1.1453,
+      "step": 5137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.272008e-05,
+      "loss": 1.305,
+      "step": 5138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2719811428571429e-05,
+      "loss": 1.4406,
+      "step": 5139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2719542857142857e-05,
+      "loss": 1.9083,
+      "step": 5140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2719274285714287e-05,
+      "loss": 1.0538,
+      "step": 5141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2719005714285715e-05,
+      "loss": 1.7737,
+      "step": 5142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2718737142857143e-05,
+      "loss": 1.2705,
+      "step": 5143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2718468571428571e-05,
+      "loss": 1.3584,
+      "step": 5144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2718200000000001e-05,
+      "loss": 1.5251,
+      "step": 5145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2717931428571428e-05,
+      "loss": 1.3538,
+      "step": 5146
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2717662857142858e-05,
+      "loss": 1.4258,
+      "step": 5147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2717394285714286e-05,
+      "loss": 1.5979,
+      "step": 5148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2717125714285716e-05,
+      "loss": 1.6031,
+      "step": 5149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2716857142857142e-05,
+      "loss": 1.4035,
+      "step": 5150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2716588571428572e-05,
+      "loss": 1.5763,
+      "step": 5151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.271632e-05,
+      "loss": 1.4173,
+      "step": 5152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2716051428571428e-05,
+      "loss": 1.7987,
+      "step": 5153
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2715782857142858e-05,
+      "loss": 1.5172,
+      "step": 5154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2715514285714286e-05,
+      "loss": 1.5835,
+      "step": 5155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2715245714285715e-05,
+      "loss": 1.4944,
+      "step": 5156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2714977142857143e-05,
+      "loss": 0.9811,
+      "step": 5157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2714708571428573e-05,
+      "loss": 1.7086,
+      "step": 5158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2714440000000001e-05,
+      "loss": 1.3047,
+      "step": 5159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2714171428571429e-05,
+      "loss": 1.5143,
+      "step": 5160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2713902857142857e-05,
+      "loss": 1.2087,
+      "step": 5161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.2713634285714287e-05,
+      "loss": 1.4483,
+      "step": 5162
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2713365714285714e-05,
+      "loss": 1.4704,
+      "step": 5163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2713097142857143e-05,
+      "loss": 1.33,
+      "step": 5164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2712828571428572e-05,
+      "loss": 1.3233,
+      "step": 5165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.271256e-05,
+      "loss": 1.52,
+      "step": 5166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.271229142857143e-05,
+      "loss": 1.7088,
+      "step": 5167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2712022857142858e-05,
+      "loss": 1.524,
+      "step": 5168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2711754285714286e-05,
+      "loss": 1.5781,
+      "step": 5169
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2711485714285714e-05,
+      "loss": 1.5247,
+      "step": 5170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2711217142857144e-05,
+      "loss": 1.5646,
+      "step": 5171
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2710948571428572e-05,
+      "loss": 0.9971,
+      "step": 5172
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.271068e-05,
+      "loss": 1.0733,
+      "step": 5173
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2710411428571429e-05,
+      "loss": 1.5685,
+      "step": 5174
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2710142857142858e-05,
+      "loss": 1.1324,
+      "step": 5175
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2709874285714285e-05,
+      "loss": 1.2704,
+      "step": 5176
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2709605714285715e-05,
+      "loss": 1.3556,
+      "step": 5177
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2709337142857143e-05,
+      "loss": 1.3833,
+      "step": 5178
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2709068571428571e-05,
+      "loss": 1.1349,
+      "step": 5179
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2708800000000001e-05,
+      "loss": 1.5438,
+      "step": 5180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.270853142857143e-05,
+      "loss": 1.3873,
+      "step": 5181
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2708262857142859e-05,
+      "loss": 1.2346,
+      "step": 5182
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2707994285714286e-05,
+      "loss": 1.5007,
+      "step": 5183
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2707725714285715e-05,
+      "loss": 1.6307,
+      "step": 5184
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2707457142857144e-05,
+      "loss": 1.6845,
+      "step": 5185
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2707188571428572e-05,
+      "loss": 1.5544,
+      "step": 5186
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.270692e-05,
+      "loss": 1.2524,
+      "step": 5187
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.270665142857143e-05,
+      "loss": 1.3661,
+      "step": 5188
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2706382857142856e-05,
+      "loss": 1.5026,
+      "step": 5189
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2706114285714286e-05,
+      "loss": 1.3546,
+      "step": 5190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2705845714285714e-05,
+      "loss": 1.0449,
+      "step": 5191
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2705577142857143e-05,
+      "loss": 1.2828,
+      "step": 5192
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2705308571428572e-05,
+      "loss": 1.726,
+      "step": 5193
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.270504e-05,
+      "loss": 1.7724,
+      "step": 5194
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.270477142857143e-05,
+      "loss": 1.5404,
+      "step": 5195
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2704502857142857e-05,
+      "loss": 1.4399,
+      "step": 5196
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2704234285714287e-05,
+      "loss": 1.2014,
+      "step": 5197
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2703965714285715e-05,
+      "loss": 1.1595,
+      "step": 5198
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2703697142857143e-05,
+      "loss": 1.4235,
+      "step": 5199
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2703428571428571e-05,
+      "loss": 1.397,
+      "step": 5200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2703160000000001e-05,
+      "loss": 1.2507,
+      "step": 5201
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2702891428571428e-05,
+      "loss": 1.3959,
+      "step": 5202
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2702622857142858e-05,
+      "loss": 1.1459,
+      "step": 5203
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2702354285714286e-05,
+      "loss": 1.455,
+      "step": 5204
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2702085714285716e-05,
+      "loss": 1.3643,
+      "step": 5205
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2701817142857144e-05,
+      "loss": 1.518,
+      "step": 5206
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2701548571428572e-05,
+      "loss": 1.4475,
+      "step": 5207
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2701280000000002e-05,
+      "loss": 1.6064,
+      "step": 5208
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2701011428571428e-05,
+      "loss": 1.6147,
+      "step": 5209
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2700742857142858e-05,
+      "loss": 1.2597,
+      "step": 5210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2700474285714286e-05,
+      "loss": 1.5912,
+      "step": 5211
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2700205714285715e-05,
+      "loss": 1.121,
+      "step": 5212
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2699937142857143e-05,
+      "loss": 1.7088,
+      "step": 5213
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2699668571428573e-05,
+      "loss": 1.1325,
+      "step": 5214
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2699399999999999e-05,
+      "loss": 1.364,
+      "step": 5215
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2699131428571429e-05,
+      "loss": 1.5088,
+      "step": 5216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2698862857142857e-05,
+      "loss": 1.8429,
+      "step": 5217
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2698594285714287e-05,
+      "loss": 1.6581,
+      "step": 5218
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2698325714285715e-05,
+      "loss": 1.315,
+      "step": 5219
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2698057142857143e-05,
+      "loss": 1.5336,
+      "step": 5220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2697788571428573e-05,
+      "loss": 1.5707,
+      "step": 5221
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.269752e-05,
+      "loss": 1.4704,
+      "step": 5222
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.269725142857143e-05,
+      "loss": 1.1962,
+      "step": 5223
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2696982857142858e-05,
+      "loss": 1.2548,
+      "step": 5224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2696714285714286e-05,
+      "loss": 1.2821,
+      "step": 5225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2696445714285714e-05,
+      "loss": 1.7543,
+      "step": 5226
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2696177142857144e-05,
+      "loss": 1.1221,
+      "step": 5227
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2695908571428572e-05,
+      "loss": 1.3323,
+      "step": 5228
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.269564e-05,
+      "loss": 1.5549,
+      "step": 5229
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2695371428571428e-05,
+      "loss": 1.7316,
+      "step": 5230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2695102857142858e-05,
+      "loss": 1.476,
+      "step": 5231
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2694834285714287e-05,
+      "loss": 1.6335,
+      "step": 5232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2694565714285715e-05,
+      "loss": 1.1941,
+      "step": 5233
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2694297142857145e-05,
+      "loss": 1.6773,
+      "step": 5234
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2694028571428571e-05,
+      "loss": 1.4253,
+      "step": 5235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2693760000000001e-05,
+      "loss": 1.5632,
+      "step": 5236
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2693491428571429e-05,
+      "loss": 1.2421,
+      "step": 5237
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2693222857142857e-05,
+      "loss": 1.427,
+      "step": 5238
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2692954285714285e-05,
+      "loss": 1.5367,
+      "step": 5239
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2692685714285715e-05,
+      "loss": 1.3134,
+      "step": 5240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2692417142857144e-05,
+      "loss": 1.3149,
+      "step": 5241
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2692148571428572e-05,
+      "loss": 1.6607,
+      "step": 5242
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.269188e-05,
+      "loss": 1.5342,
+      "step": 5243
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.269161142857143e-05,
+      "loss": 1.4908,
+      "step": 5244
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2691342857142858e-05,
+      "loss": 1.2809,
+      "step": 5245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2691074285714286e-05,
+      "loss": 1.6648,
+      "step": 5246
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2690805714285716e-05,
+      "loss": 1.4171,
+      "step": 5247
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2690537142857142e-05,
+      "loss": 1.183,
+      "step": 5248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2690268571428572e-05,
+      "loss": 1.7287,
+      "step": 5249
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.269e-05,
+      "loss": 1.6836,
+      "step": 5250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2689731428571429e-05,
+      "loss": 1.2576,
+      "step": 5251
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2689462857142857e-05,
+      "loss": 1.1561,
+      "step": 5252
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2689194285714287e-05,
+      "loss": 1.2233,
+      "step": 5253
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2688925714285715e-05,
+      "loss": 1.1897,
+      "step": 5254
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2688657142857143e-05,
+      "loss": 1.6563,
+      "step": 5255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2688388571428571e-05,
+      "loss": 1.401,
+      "step": 5256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2688120000000001e-05,
+      "loss": 1.1902,
+      "step": 5257
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.268785142857143e-05,
+      "loss": 1.3773,
+      "step": 5258
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2687582857142857e-05,
+      "loss": 1.4867,
+      "step": 5259
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2687314285714287e-05,
+      "loss": 1.4231,
+      "step": 5260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2687045714285714e-05,
+      "loss": 1.2336,
+      "step": 5261
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2686777142857144e-05,
+      "loss": 1.7043,
+      "step": 5262
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2686508571428572e-05,
+      "loss": 1.6527,
+      "step": 5263
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2686240000000002e-05,
+      "loss": 1.5689,
+      "step": 5264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2685971428571428e-05,
+      "loss": 1.7273,
+      "step": 5265
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2685702857142858e-05,
+      "loss": 0.9387,
+      "step": 5266
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2685434285714286e-05,
+      "loss": 1.4199,
+      "step": 5267
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2685165714285714e-05,
+      "loss": 1.3761,
+      "step": 5268
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2684897142857143e-05,
+      "loss": 1.2363,
+      "step": 5269
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2684628571428572e-05,
+      "loss": 1.1899,
+      "step": 5270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.268436e-05,
+      "loss": 1.4303,
+      "step": 5271
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2684091428571429e-05,
+      "loss": 1.5671,
+      "step": 5272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2683822857142859e-05,
+      "loss": 1.1094,
+      "step": 5273
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2683554285714285e-05,
+      "loss": 1.1509,
+      "step": 5274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2683285714285715e-05,
+      "loss": 1.2858,
+      "step": 5275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2683017142857143e-05,
+      "loss": 1.5512,
+      "step": 5276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2682748571428573e-05,
+      "loss": 1.4123,
+      "step": 5277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.268248e-05,
+      "loss": 1.3775,
+      "step": 5278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.268221142857143e-05,
+      "loss": 1.6286,
+      "step": 5279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2681942857142858e-05,
+      "loss": 1.0844,
+      "step": 5280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2681674285714286e-05,
+      "loss": 1.2096,
+      "step": 5281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2681405714285714e-05,
+      "loss": 1.4379,
+      "step": 5282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2681137142857144e-05,
+      "loss": 1.0958,
+      "step": 5283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2680868571428572e-05,
+      "loss": 1.3009,
+      "step": 5284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.26806e-05,
+      "loss": 1.703,
+      "step": 5285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.268033142857143e-05,
+      "loss": 1.167,
+      "step": 5286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2680062857142858e-05,
+      "loss": 1.4476,
+      "step": 5287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2679794285714286e-05,
+      "loss": 1.2627,
+      "step": 5288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2679525714285715e-05,
+      "loss": 1.4099,
+      "step": 5289
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2679257142857144e-05,
+      "loss": 1.4046,
+      "step": 5290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2678988571428571e-05,
+      "loss": 1.3585,
+      "step": 5291
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.267872e-05,
+      "loss": 1.3765,
+      "step": 5292
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2678451428571429e-05,
+      "loss": 1.372,
+      "step": 5293
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2678182857142857e-05,
+      "loss": 1.4917,
+      "step": 5294
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2677914285714285e-05,
+      "loss": 1.4441,
+      "step": 5295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2677645714285715e-05,
+      "loss": 1.2904,
+      "step": 5296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2677377142857143e-05,
+      "loss": 1.6543,
+      "step": 5297
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2677108571428572e-05,
+      "loss": 1.4322,
+      "step": 5298
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2676840000000001e-05,
+      "loss": 1.8543,
+      "step": 5299
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.267657142857143e-05,
+      "loss": 1.2507,
+      "step": 5300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2676302857142858e-05,
+      "loss": 1.5358,
+      "step": 5301
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2676034285714286e-05,
+      "loss": 1.5548,
+      "step": 5302
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2675765714285716e-05,
+      "loss": 1.5728,
+      "step": 5303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2675497142857142e-05,
+      "loss": 1.2657,
+      "step": 5304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2675228571428572e-05,
+      "loss": 1.5194,
+      "step": 5305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.267496e-05,
+      "loss": 1.4757,
+      "step": 5306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2674691428571429e-05,
+      "loss": 1.4427,
+      "step": 5307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2674422857142857e-05,
+      "loss": 1.47,
+      "step": 5308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2674154285714287e-05,
+      "loss": 1.3879,
+      "step": 5309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2673885714285715e-05,
+      "loss": 1.3794,
+      "step": 5310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2673617142857143e-05,
+      "loss": 1.5488,
+      "step": 5311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2673348571428573e-05,
+      "loss": 1.5877,
+      "step": 5312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2673080000000001e-05,
+      "loss": 1.379,
+      "step": 5313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.267281142857143e-05,
+      "loss": 1.5946,
+      "step": 5314
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2672542857142857e-05,
+      "loss": 1.6504,
+      "step": 5315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2672274285714287e-05,
+      "loss": 1.2474,
+      "step": 5316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2672005714285714e-05,
+      "loss": 1.4615,
+      "step": 5317
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2671737142857144e-05,
+      "loss": 1.4122,
+      "step": 5318
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2671468571428572e-05,
+      "loss": 1.4674,
+      "step": 5319
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.26712e-05,
+      "loss": 1.5319,
+      "step": 5320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2670931428571428e-05,
+      "loss": 1.5873,
+      "step": 5321
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2670662857142858e-05,
+      "loss": 1.0639,
+      "step": 5322
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2670394285714286e-05,
+      "loss": 1.3554,
+      "step": 5323
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2670125714285714e-05,
+      "loss": 1.601,
+      "step": 5324
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2669857142857144e-05,
+      "loss": 1.3072,
+      "step": 5325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2669588571428572e-05,
+      "loss": 1.4563,
+      "step": 5326
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.266932e-05,
+      "loss": 1.6655,
+      "step": 5327
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2669051428571429e-05,
+      "loss": 1.3843,
+      "step": 5328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2668782857142859e-05,
+      "loss": 1.3063,
+      "step": 5329
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2668514285714285e-05,
+      "loss": 1.4689,
+      "step": 5330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2668245714285715e-05,
+      "loss": 1.8696,
+      "step": 5331
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2667977142857143e-05,
+      "loss": 1.1544,
+      "step": 5332
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2667708571428573e-05,
+      "loss": 1.2142,
+      "step": 5333
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.266744e-05,
+      "loss": 1.2801,
+      "step": 5334
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.266717142857143e-05,
+      "loss": 1.672,
+      "step": 5335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2666902857142858e-05,
+      "loss": 1.1602,
+      "step": 5336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.2666634285714286e-05,
+      "loss": 1.3029,
+      "step": 5337
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2666365714285716e-05,
+      "loss": 1.6648,
+      "step": 5338
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2666097142857144e-05,
+      "loss": 1.784,
+      "step": 5339
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2665828571428572e-05,
+      "loss": 1.3399,
+      "step": 5340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.266556e-05,
+      "loss": 1.3334,
+      "step": 5341
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.266529142857143e-05,
+      "loss": 1.6914,
+      "step": 5342
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2665022857142856e-05,
+      "loss": 1.5095,
+      "step": 5343
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2664754285714286e-05,
+      "loss": 1.2273,
+      "step": 5344
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2664485714285714e-05,
+      "loss": 1.2539,
+      "step": 5345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2664217142857144e-05,
+      "loss": 1.5075,
+      "step": 5346
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.266394857142857e-05,
+      "loss": 1.2379,
+      "step": 5347
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.266368e-05,
+      "loss": 1.4734,
+      "step": 5348
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2663411428571429e-05,
+      "loss": 1.5793,
+      "step": 5349
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2663142857142857e-05,
+      "loss": 1.5419,
+      "step": 5350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2662874285714287e-05,
+      "loss": 1.4316,
+      "step": 5351
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2662605714285715e-05,
+      "loss": 1.0519,
+      "step": 5352
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2662337142857143e-05,
+      "loss": 1.5489,
+      "step": 5353
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2662068571428571e-05,
+      "loss": 1.6247,
+      "step": 5354
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2661800000000001e-05,
+      "loss": 1.5831,
+      "step": 5355
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.266153142857143e-05,
+      "loss": 1.6247,
+      "step": 5356
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2661262857142858e-05,
+      "loss": 1.3238,
+      "step": 5357
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2660994285714286e-05,
+      "loss": 1.3539,
+      "step": 5358
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2660725714285716e-05,
+      "loss": 1.5463,
+      "step": 5359
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2660457142857142e-05,
+      "loss": 1.5884,
+      "step": 5360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2660188571428572e-05,
+      "loss": 1.2271,
+      "step": 5361
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.265992e-05,
+      "loss": 1.4379,
+      "step": 5362
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2659651428571428e-05,
+      "loss": 1.3964,
+      "step": 5363
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2659382857142858e-05,
+      "loss": 1.4052,
+      "step": 5364
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2659114285714286e-05,
+      "loss": 1.2556,
+      "step": 5365
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2658845714285715e-05,
+      "loss": 1.4537,
+      "step": 5366
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2658577142857143e-05,
+      "loss": 1.729,
+      "step": 5367
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2658308571428573e-05,
+      "loss": 1.5749,
+      "step": 5368
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2658040000000001e-05,
+      "loss": 1.4091,
+      "step": 5369
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2657771428571429e-05,
+      "loss": 1.2432,
+      "step": 5370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2657502857142857e-05,
+      "loss": 1.385,
+      "step": 5371
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2657234285714287e-05,
+      "loss": 1.2598,
+      "step": 5372
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2656965714285714e-05,
+      "loss": 1.7066,
+      "step": 5373
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2656697142857143e-05,
+      "loss": 1.3147,
+      "step": 5374
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2656428571428572e-05,
+      "loss": 1.6303,
+      "step": 5375
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.265616e-05,
+      "loss": 1.4585,
+      "step": 5376
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.265589142857143e-05,
+      "loss": 1.4223,
+      "step": 5377
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2655622857142858e-05,
+      "loss": 1.7013,
+      "step": 5378
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2655354285714288e-05,
+      "loss": 1.552,
+      "step": 5379
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2655085714285714e-05,
+      "loss": 1.2184,
+      "step": 5380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2654817142857144e-05,
+      "loss": 1.3965,
+      "step": 5381
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2654548571428572e-05,
+      "loss": 1.0025,
+      "step": 5382
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.265428e-05,
+      "loss": 1.3047,
+      "step": 5383
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2654011428571429e-05,
+      "loss": 1.6006,
+      "step": 5384
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2653742857142858e-05,
+      "loss": 1.0541,
+      "step": 5385
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2653474285714285e-05,
+      "loss": 1.1673,
+      "step": 5386
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2653205714285715e-05,
+      "loss": 1.3038,
+      "step": 5387
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2652937142857143e-05,
+      "loss": 1.6549,
+      "step": 5388
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2652668571428571e-05,
+      "loss": 1.4228,
+      "step": 5389
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2652400000000001e-05,
+      "loss": 1.5638,
+      "step": 5390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.265213142857143e-05,
+      "loss": 1.0984,
+      "step": 5391
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2651862857142859e-05,
+      "loss": 1.3116,
+      "step": 5392
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2651594285714286e-05,
+      "loss": 1.9634,
+      "step": 5393
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2651325714285715e-05,
+      "loss": 1.7758,
+      "step": 5394
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2651057142857144e-05,
+      "loss": 1.794,
+      "step": 5395
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2650788571428572e-05,
+      "loss": 1.7341,
+      "step": 5396
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.265052e-05,
+      "loss": 1.2327,
+      "step": 5397
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.265025142857143e-05,
+      "loss": 1.2916,
+      "step": 5398
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2649982857142856e-05,
+      "loss": 1.8607,
+      "step": 5399
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2649714285714286e-05,
+      "loss": 1.3038,
+      "step": 5400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2649445714285714e-05,
+      "loss": 1.5745,
+      "step": 5401
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2649177142857143e-05,
+      "loss": 1.5095,
+      "step": 5402
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2648908571428572e-05,
+      "loss": 1.0411,
+      "step": 5403
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.264864e-05,
+      "loss": 1.4624,
+      "step": 5404
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.264837142857143e-05,
+      "loss": 1.5305,
+      "step": 5405
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2648102857142857e-05,
+      "loss": 1.4092,
+      "step": 5406
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2647834285714287e-05,
+      "loss": 1.5177,
+      "step": 5407
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2647565714285715e-05,
+      "loss": 1.1882,
+      "step": 5408
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2647297142857143e-05,
+      "loss": 1.14,
+      "step": 5409
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2647028571428571e-05,
+      "loss": 1.4016,
+      "step": 5410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2646760000000001e-05,
+      "loss": 1.3325,
+      "step": 5411
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2646491428571428e-05,
+      "loss": 1.9369,
+      "step": 5412
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2646222857142858e-05,
+      "loss": 1.2401,
+      "step": 5413
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2645954285714286e-05,
+      "loss": 1.2044,
+      "step": 5414
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2645685714285716e-05,
+      "loss": 1.5558,
+      "step": 5415
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2645417142857144e-05,
+      "loss": 1.4526,
+      "step": 5416
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2645148571428572e-05,
+      "loss": 1.0392,
+      "step": 5417
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2644880000000002e-05,
+      "loss": 1.6857,
+      "step": 5418
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2644611428571428e-05,
+      "loss": 1.4846,
+      "step": 5419
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2644342857142858e-05,
+      "loss": 1.2967,
+      "step": 5420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2644074285714286e-05,
+      "loss": 1.1822,
+      "step": 5421
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2643805714285715e-05,
+      "loss": 1.3641,
+      "step": 5422
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2643537142857143e-05,
+      "loss": 1.4829,
+      "step": 5423
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2643268571428573e-05,
+      "loss": 1.1631,
+      "step": 5424
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2642999999999999e-05,
+      "loss": 1.2631,
+      "step": 5425
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2642731428571429e-05,
+      "loss": 1.4936,
+      "step": 5426
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2642462857142857e-05,
+      "loss": 1.66,
+      "step": 5427
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2642194285714287e-05,
+      "loss": 1.5335,
+      "step": 5428
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2641925714285715e-05,
+      "loss": 1.7015,
+      "step": 5429
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2641657142857143e-05,
+      "loss": 1.6028,
+      "step": 5430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2641388571428573e-05,
+      "loss": 1.9469,
+      "step": 5431
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.264112e-05,
+      "loss": 1.5934,
+      "step": 5432
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.264085142857143e-05,
+      "loss": 1.18,
+      "step": 5433
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2640582857142858e-05,
+      "loss": 1.6769,
+      "step": 5434
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2640314285714286e-05,
+      "loss": 1.8116,
+      "step": 5435
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2640045714285714e-05,
+      "loss": 1.2688,
+      "step": 5436
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2639777142857144e-05,
+      "loss": 1.388,
+      "step": 5437
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2639508571428572e-05,
+      "loss": 1.2423,
+      "step": 5438
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.263924e-05,
+      "loss": 1.2868,
+      "step": 5439
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2638971428571428e-05,
+      "loss": 1.3717,
+      "step": 5440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2638702857142858e-05,
+      "loss": 1.5656,
+      "step": 5441
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2638434285714287e-05,
+      "loss": 1.1563,
+      "step": 5442
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2638165714285715e-05,
+      "loss": 0.9962,
+      "step": 5443
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2637897142857145e-05,
+      "loss": 1.3356,
+      "step": 5444
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2637628571428571e-05,
+      "loss": 1.6589,
+      "step": 5445
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2637360000000001e-05,
+      "loss": 1.0785,
+      "step": 5446
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2637091428571429e-05,
+      "loss": 1.2741,
+      "step": 5447
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2636822857142857e-05,
+      "loss": 1.3523,
+      "step": 5448
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2636554285714285e-05,
+      "loss": 1.3428,
+      "step": 5449
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2636285714285715e-05,
+      "loss": 0.9544,
+      "step": 5450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2636017142857143e-05,
+      "loss": 1.5869,
+      "step": 5451
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2635748571428572e-05,
+      "loss": 1.3848,
+      "step": 5452
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.263548e-05,
+      "loss": 1.3857,
+      "step": 5453
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.263521142857143e-05,
+      "loss": 1.7338,
+      "step": 5454
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2634942857142858e-05,
+      "loss": 1.7111,
+      "step": 5455
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2634674285714286e-05,
+      "loss": 1.437,
+      "step": 5456
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2634405714285716e-05,
+      "loss": 1.1112,
+      "step": 5457
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2634137142857142e-05,
+      "loss": 1.6728,
+      "step": 5458
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2633868571428572e-05,
+      "loss": 1.2527,
+      "step": 5459
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.26336e-05,
+      "loss": 1.7572,
+      "step": 5460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.263333142857143e-05,
+      "loss": 1.6774,
+      "step": 5461
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2633062857142857e-05,
+      "loss": 1.3649,
+      "step": 5462
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2632794285714287e-05,
+      "loss": 1.4935,
+      "step": 5463
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2632525714285715e-05,
+      "loss": 1.1949,
+      "step": 5464
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2632257142857143e-05,
+      "loss": 1.4164,
+      "step": 5465
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2631988571428571e-05,
+      "loss": 1.5195,
+      "step": 5466
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2631720000000001e-05,
+      "loss": 1.4146,
+      "step": 5467
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.263145142857143e-05,
+      "loss": 1.4378,
+      "step": 5468
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2631182857142857e-05,
+      "loss": 1.2908,
+      "step": 5469
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2630914285714287e-05,
+      "loss": 1.0254,
+      "step": 5470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2630645714285714e-05,
+      "loss": 1.1752,
+      "step": 5471
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2630377142857144e-05,
+      "loss": 1.4879,
+      "step": 5472
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2630108571428572e-05,
+      "loss": 1.2729,
+      "step": 5473
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2629840000000002e-05,
+      "loss": 1.4116,
+      "step": 5474
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2629571428571428e-05,
+      "loss": 1.5035,
+      "step": 5475
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2629302857142858e-05,
+      "loss": 1.246,
+      "step": 5476
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2629034285714286e-05,
+      "loss": 1.3258,
+      "step": 5477
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2628765714285714e-05,
+      "loss": 1.2091,
+      "step": 5478
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2628497142857143e-05,
+      "loss": 1.2804,
+      "step": 5479
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2628228571428572e-05,
+      "loss": 1.5716,
+      "step": 5480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2627959999999999e-05,
+      "loss": 1.2355,
+      "step": 5481
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2627691428571429e-05,
+      "loss": 1.1528,
+      "step": 5482
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2627422857142859e-05,
+      "loss": 1.2087,
+      "step": 5483
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2627154285714287e-05,
+      "loss": 1.3507,
+      "step": 5484
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2626885714285715e-05,
+      "loss": 1.4186,
+      "step": 5485
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2626617142857143e-05,
+      "loss": 1.5322,
+      "step": 5486
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2626348571428573e-05,
+      "loss": 1.0869,
+      "step": 5487
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.262608e-05,
+      "loss": 1.2853,
+      "step": 5488
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.262581142857143e-05,
+      "loss": 2.0415,
+      "step": 5489
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2625542857142858e-05,
+      "loss": 1.2139,
+      "step": 5490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2625274285714286e-05,
+      "loss": 1.5581,
+      "step": 5491
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2625005714285714e-05,
+      "loss": 1.2686,
+      "step": 5492
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2624737142857144e-05,
+      "loss": 1.3417,
+      "step": 5493
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.262446857142857e-05,
+      "loss": 1.6577,
+      "step": 5494
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.26242e-05,
+      "loss": 1.1394,
+      "step": 5495
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.262393142857143e-05,
+      "loss": 1.3168,
+      "step": 5496
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2623662857142858e-05,
+      "loss": 1.4416,
+      "step": 5497
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2623394285714286e-05,
+      "loss": 1.1957,
+      "step": 5498
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2623125714285715e-05,
+      "loss": 1.328,
+      "step": 5499
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2622857142857144e-05,
+      "loss": 1.3396,
+      "step": 5500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2622588571428571e-05,
+      "loss": 1.7346,
+      "step": 5501
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.262232e-05,
+      "loss": 1.5913,
+      "step": 5502
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2622051428571429e-05,
+      "loss": 1.3308,
+      "step": 5503
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2621782857142857e-05,
+      "loss": 1.3565,
+      "step": 5504
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2621514285714285e-05,
+      "loss": 1.8093,
+      "step": 5505
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2621245714285715e-05,
+      "loss": 1.4323,
+      "step": 5506
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2620977142857143e-05,
+      "loss": 1.3305,
+      "step": 5507
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2620708571428572e-05,
+      "loss": 1.8265,
+      "step": 5508
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2620440000000001e-05,
+      "loss": 1.4045,
+      "step": 5509
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.262017142857143e-05,
+      "loss": 1.542,
+      "step": 5510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2619902857142858e-05,
+      "loss": 1.3824,
+      "step": 5511
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2619634285714286e-05,
+      "loss": 1.6025,
+      "step": 5512
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2619365714285716e-05,
+      "loss": 1.4066,
+      "step": 5513
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2619097142857142e-05,
+      "loss": 1.3762,
+      "step": 5514
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2618828571428572e-05,
+      "loss": 1.3944,
+      "step": 5515
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.261856e-05,
+      "loss": 1.2442,
+      "step": 5516
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2618291428571429e-05,
+      "loss": 1.5493,
+      "step": 5517
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2618022857142857e-05,
+      "loss": 1.5878,
+      "step": 5518
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2617754285714287e-05,
+      "loss": 1.2049,
+      "step": 5519
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2617485714285715e-05,
+      "loss": 1.1849,
+      "step": 5520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2617217142857143e-05,
+      "loss": 1.1046,
+      "step": 5521
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2616948571428573e-05,
+      "loss": 1.351,
+      "step": 5522
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2616680000000001e-05,
+      "loss": 1.3593,
+      "step": 5523
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.261641142857143e-05,
+      "loss": 1.4991,
+      "step": 5524
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2616142857142857e-05,
+      "loss": 1.4505,
+      "step": 5525
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2615874285714287e-05,
+      "loss": 1.3171,
+      "step": 5526
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2615605714285714e-05,
+      "loss": 1.1948,
+      "step": 5527
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2615337142857144e-05,
+      "loss": 1.6526,
+      "step": 5528
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2615068571428572e-05,
+      "loss": 1.4324,
+      "step": 5529
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2614800000000002e-05,
+      "loss": 1.622,
+      "step": 5530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2614531428571428e-05,
+      "loss": 1.4296,
+      "step": 5531
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2614262857142858e-05,
+      "loss": 1.1296,
+      "step": 5532
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2613994285714286e-05,
+      "loss": 1.2118,
+      "step": 5533
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2613725714285714e-05,
+      "loss": 1.5688,
+      "step": 5534
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2613457142857144e-05,
+      "loss": 1.5991,
+      "step": 5535
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2613188571428572e-05,
+      "loss": 1.7762,
+      "step": 5536
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.261292e-05,
+      "loss": 2.0991,
+      "step": 5537
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2612651428571429e-05,
+      "loss": 1.1015,
+      "step": 5538
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2612382857142859e-05,
+      "loss": 1.6526,
+      "step": 5539
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2612114285714285e-05,
+      "loss": 1.7432,
+      "step": 5540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2611845714285715e-05,
+      "loss": 1.4555,
+      "step": 5541
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2611577142857143e-05,
+      "loss": 1.2771,
+      "step": 5542
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2611308571428573e-05,
+      "loss": 1.3463,
+      "step": 5543
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.261104e-05,
+      "loss": 1.3675,
+      "step": 5544
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.261077142857143e-05,
+      "loss": 1.3896,
+      "step": 5545
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2610502857142858e-05,
+      "loss": 1.3405,
+      "step": 5546
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2610234285714286e-05,
+      "loss": 1.7542,
+      "step": 5547
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2609965714285716e-05,
+      "loss": 1.6302,
+      "step": 5548
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2609697142857144e-05,
+      "loss": 1.2469,
+      "step": 5549
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2609428571428572e-05,
+      "loss": 1.3619,
+      "step": 5550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.260916e-05,
+      "loss": 1.8318,
+      "step": 5551
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.260889142857143e-05,
+      "loss": 1.4788,
+      "step": 5552
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2608622857142858e-05,
+      "loss": 1.1522,
+      "step": 5553
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2608354285714286e-05,
+      "loss": 1.7106,
+      "step": 5554
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2608085714285714e-05,
+      "loss": 1.2536,
+      "step": 5555
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2607817142857144e-05,
+      "loss": 1.5078,
+      "step": 5556
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.260754857142857e-05,
+      "loss": 1.4455,
+      "step": 5557
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.260728e-05,
+      "loss": 1.1586,
+      "step": 5558
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2607011428571429e-05,
+      "loss": 1.0895,
+      "step": 5559
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2606742857142857e-05,
+      "loss": 0.6984,
+      "step": 5560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2606474285714287e-05,
+      "loss": 1.269,
+      "step": 5561
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2606205714285715e-05,
+      "loss": 1.245,
+      "step": 5562
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2605937142857143e-05,
+      "loss": 1.369,
+      "step": 5563
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2605668571428571e-05,
+      "loss": 1.3442,
+      "step": 5564
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2605400000000001e-05,
+      "loss": 1.1965,
+      "step": 5565
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.260513142857143e-05,
+      "loss": 1.147,
+      "step": 5566
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2604862857142858e-05,
+      "loss": 1.4671,
+      "step": 5567
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2604594285714286e-05,
+      "loss": 1.5022,
+      "step": 5568
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2604325714285716e-05,
+      "loss": 1.6881,
+      "step": 5569
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2604057142857142e-05,
+      "loss": 1.4706,
+      "step": 5570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2603788571428572e-05,
+      "loss": 1.1161,
+      "step": 5571
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.260352e-05,
+      "loss": 1.3125,
+      "step": 5572
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2603251428571428e-05,
+      "loss": 1.4855,
+      "step": 5573
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2602982857142858e-05,
+      "loss": 1.5977,
+      "step": 5574
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2602714285714286e-05,
+      "loss": 1.1519,
+      "step": 5575
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2602445714285715e-05,
+      "loss": 1.5257,
+      "step": 5576
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2602177142857143e-05,
+      "loss": 1.2735,
+      "step": 5577
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2601908571428573e-05,
+      "loss": 1.391,
+      "step": 5578
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2601640000000001e-05,
+      "loss": 1.4696,
+      "step": 5579
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2601371428571429e-05,
+      "loss": 1.2664,
+      "step": 5580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2601102857142857e-05,
+      "loss": 1.4488,
+      "step": 5581
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2600834285714287e-05,
+      "loss": 1.7308,
+      "step": 5582
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2600565714285714e-05,
+      "loss": 1.4793,
+      "step": 5583
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2600297142857143e-05,
+      "loss": 1.3446,
+      "step": 5584
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2600028571428572e-05,
+      "loss": 1.5236,
+      "step": 5585
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259976e-05,
+      "loss": 1.291,
+      "step": 5586
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2599491428571428e-05,
+      "loss": 1.4579,
+      "step": 5587
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2599222857142858e-05,
+      "loss": 1.3976,
+      "step": 5588
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2598954285714288e-05,
+      "loss": 1.77,
+      "step": 5589
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2598685714285714e-05,
+      "loss": 1.3282,
+      "step": 5590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2598417142857144e-05,
+      "loss": 1.5146,
+      "step": 5591
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2598148571428572e-05,
+      "loss": 1.6442,
+      "step": 5592
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259788e-05,
+      "loss": 1.0705,
+      "step": 5593
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2597611428571429e-05,
+      "loss": 0.8213,
+      "step": 5594
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2597342857142858e-05,
+      "loss": 1.1564,
+      "step": 5595
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2597074285714285e-05,
+      "loss": 1.6729,
+      "step": 5596
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2596805714285715e-05,
+      "loss": 1.5087,
+      "step": 5597
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2596537142857143e-05,
+      "loss": 1.2677,
+      "step": 5598
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2596268571428571e-05,
+      "loss": 1.2559,
+      "step": 5599
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2596e-05,
+      "loss": 1.5266,
+      "step": 5600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259573142857143e-05,
+      "loss": 1.0807,
+      "step": 5601
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2595462857142859e-05,
+      "loss": 1.5064,
+      "step": 5602
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2595194285714286e-05,
+      "loss": 1.4259,
+      "step": 5603
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2594925714285715e-05,
+      "loss": 1.1891,
+      "step": 5604
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2594657142857144e-05,
+      "loss": 1.0655,
+      "step": 5605
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2594388571428572e-05,
+      "loss": 1.3033,
+      "step": 5606
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259412e-05,
+      "loss": 1.4385,
+      "step": 5607
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259385142857143e-05,
+      "loss": 1.4745,
+      "step": 5608
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2593582857142856e-05,
+      "loss": 1.337,
+      "step": 5609
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2593314285714286e-05,
+      "loss": 1.1783,
+      "step": 5610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2593045714285714e-05,
+      "loss": 1.334,
+      "step": 5611
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2592777142857144e-05,
+      "loss": 1.7106,
+      "step": 5612
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259250857142857e-05,
+      "loss": 1.1971,
+      "step": 5613
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259224e-05,
+      "loss": 1.158,
+      "step": 5614
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.259197142857143e-05,
+      "loss": 1.2666,
+      "step": 5615
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2591702857142857e-05,
+      "loss": 1.3676,
+      "step": 5616
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2591434285714287e-05,
+      "loss": 1.364,
+      "step": 5617
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2591165714285715e-05,
+      "loss": 1.2444,
+      "step": 5618
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2590897142857143e-05,
+      "loss": 1.5995,
+      "step": 5619
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2590628571428571e-05,
+      "loss": 1.3059,
+      "step": 5620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2590360000000001e-05,
+      "loss": 1.3695,
+      "step": 5621
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2590091428571428e-05,
+      "loss": 1.1961,
+      "step": 5622
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2589822857142858e-05,
+      "loss": 1.5626,
+      "step": 5623
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2589554285714286e-05,
+      "loss": 1.3692,
+      "step": 5624
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2589285714285716e-05,
+      "loss": 1.4022,
+      "step": 5625
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2589017142857142e-05,
+      "loss": 1.3724,
+      "step": 5626
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2588748571428572e-05,
+      "loss": 1.1234,
+      "step": 5627
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2588480000000002e-05,
+      "loss": 1.1604,
+      "step": 5628
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2588211428571428e-05,
+      "loss": 1.6703,
+      "step": 5629
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2587942857142858e-05,
+      "loss": 1.7115,
+      "step": 5630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2587674285714286e-05,
+      "loss": 1.408,
+      "step": 5631
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2587405714285715e-05,
+      "loss": 1.4859,
+      "step": 5632
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2587137142857143e-05,
+      "loss": 1.7061,
+      "step": 5633
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2586868571428573e-05,
+      "loss": 1.4704,
+      "step": 5634
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.25866e-05,
+      "loss": 1.3258,
+      "step": 5635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2586331428571429e-05,
+      "loss": 1.395,
+      "step": 5636
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2586062857142857e-05,
+      "loss": 1.4198,
+      "step": 5637
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2585794285714287e-05,
+      "loss": 1.6323,
+      "step": 5638
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2585525714285713e-05,
+      "loss": 1.3559,
+      "step": 5639
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2585257142857143e-05,
+      "loss": 1.6275,
+      "step": 5640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2584988571428573e-05,
+      "loss": 1.5457,
+      "step": 5641
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.258472e-05,
+      "loss": 1.6195,
+      "step": 5642
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.258445142857143e-05,
+      "loss": 1.5775,
+      "step": 5643
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2584182857142858e-05,
+      "loss": 1.5018,
+      "step": 5644
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2583914285714286e-05,
+      "loss": 1.1348,
+      "step": 5645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2583645714285714e-05,
+      "loss": 1.352,
+      "step": 5646
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2583377142857144e-05,
+      "loss": 1.1101,
+      "step": 5647
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2583108571428572e-05,
+      "loss": 1.3945,
+      "step": 5648
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.258284e-05,
+      "loss": 1.7045,
+      "step": 5649
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2582571428571428e-05,
+      "loss": 1.3024,
+      "step": 5650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2582302857142858e-05,
+      "loss": 1.21,
+      "step": 5651
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2582034285714285e-05,
+      "loss": 1.1664,
+      "step": 5652
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2581765714285715e-05,
+      "loss": 1.1353,
+      "step": 5653
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2581497142857145e-05,
+      "loss": 1.3108,
+      "step": 5654
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2581228571428571e-05,
+      "loss": 1.7682,
+      "step": 5655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2580960000000001e-05,
+      "loss": 1.3863,
+      "step": 5656
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2580691428571429e-05,
+      "loss": 1.4497,
+      "step": 5657
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2580422857142859e-05,
+      "loss": 1.5879,
+      "step": 5658
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2580154285714285e-05,
+      "loss": 1.3825,
+      "step": 5659
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2579885714285715e-05,
+      "loss": 1.2889,
+      "step": 5660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2579617142857143e-05,
+      "loss": 1.5614,
+      "step": 5661
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2579348571428572e-05,
+      "loss": 1.6423,
+      "step": 5662
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.257908e-05,
+      "loss": 1.15,
+      "step": 5663
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.257881142857143e-05,
+      "loss": 1.1388,
+      "step": 5664
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2578542857142856e-05,
+      "loss": 1.587,
+      "step": 5665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2578274285714286e-05,
+      "loss": 1.2283,
+      "step": 5666
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2578005714285716e-05,
+      "loss": 1.8792,
+      "step": 5667
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2577737142857142e-05,
+      "loss": 1.4678,
+      "step": 5668
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2577468571428572e-05,
+      "loss": 1.1854,
+      "step": 5669
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.25772e-05,
+      "loss": 1.418,
+      "step": 5670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.257693142857143e-05,
+      "loss": 1.3634,
+      "step": 5671
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2576662857142857e-05,
+      "loss": 1.4586,
+      "step": 5672
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2576394285714287e-05,
+      "loss": 0.7986,
+      "step": 5673
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2576125714285715e-05,
+      "loss": 1.4959,
+      "step": 5674
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2575857142857143e-05,
+      "loss": 1.5031,
+      "step": 5675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2575588571428571e-05,
+      "loss": 1.7603,
+      "step": 5676
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2575320000000001e-05,
+      "loss": 1.568,
+      "step": 5677
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2575051428571428e-05,
+      "loss": 0.9871,
+      "step": 5678
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2574782857142857e-05,
+      "loss": 1.0006,
+      "step": 5679
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2574514285714287e-05,
+      "loss": 1.4712,
+      "step": 5680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2574245714285715e-05,
+      "loss": 1.9237,
+      "step": 5681
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2573977142857144e-05,
+      "loss": 1.0749,
+      "step": 5682
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2573708571428572e-05,
+      "loss": 1.071,
+      "step": 5683
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2573440000000002e-05,
+      "loss": 1.4747,
+      "step": 5684
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2573171428571428e-05,
+      "loss": 1.2787,
+      "step": 5685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2572902857142858e-05,
+      "loss": 1.4403,
+      "step": 5686
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2572634285714286e-05,
+      "loss": 1.0455,
+      "step": 5687
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2572365714285714e-05,
+      "loss": 1.3726,
+      "step": 5688
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2572097142857143e-05,
+      "loss": 1.5402,
+      "step": 5689
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2571828571428572e-05,
+      "loss": 1.409,
+      "step": 5690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2571559999999999e-05,
+      "loss": 1.2215,
+      "step": 5691
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2571291428571429e-05,
+      "loss": 1.1997,
+      "step": 5692
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2571022857142857e-05,
+      "loss": 1.4102,
+      "step": 5693
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2570754285714287e-05,
+      "loss": 1.3576,
+      "step": 5694
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2570485714285715e-05,
+      "loss": 1.4649,
+      "step": 5695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2570217142857143e-05,
+      "loss": 1.4107,
+      "step": 5696
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2569948571428573e-05,
+      "loss": 1.2211,
+      "step": 5697
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256968e-05,
+      "loss": 1.8331,
+      "step": 5698
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256941142857143e-05,
+      "loss": 1.8062,
+      "step": 5699
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2569142857142858e-05,
+      "loss": 1.7702,
+      "step": 5700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2568874285714286e-05,
+      "loss": 1.029,
+      "step": 5701
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2568605714285714e-05,
+      "loss": 1.6272,
+      "step": 5702
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2568337142857144e-05,
+      "loss": 1.4676,
+      "step": 5703
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2568068571428572e-05,
+      "loss": 1.3996,
+      "step": 5704
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.25678e-05,
+      "loss": 1.7525,
+      "step": 5705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2567531428571428e-05,
+      "loss": 1.2885,
+      "step": 5706
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2567262857142858e-05,
+      "loss": 1.2767,
+      "step": 5707
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2566994285714286e-05,
+      "loss": 1.3899,
+      "step": 5708
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2566725714285715e-05,
+      "loss": 1.4415,
+      "step": 5709
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2566457142857144e-05,
+      "loss": 1.46,
+      "step": 5710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2566188571428571e-05,
+      "loss": 1.325,
+      "step": 5711
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256592e-05,
+      "loss": 1.4745,
+      "step": 5712
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2565651428571429e-05,
+      "loss": 0.8271,
+      "step": 5713
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2565382857142857e-05,
+      "loss": 1.4199,
+      "step": 5714
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2565114285714285e-05,
+      "loss": 1.4232,
+      "step": 5715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2564845714285715e-05,
+      "loss": 1.6303,
+      "step": 5716
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2564577142857143e-05,
+      "loss": 1.309,
+      "step": 5717
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2564308571428572e-05,
+      "loss": 1.3933,
+      "step": 5718
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256404e-05,
+      "loss": 1.2624,
+      "step": 5719
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256377142857143e-05,
+      "loss": 1.2712,
+      "step": 5720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2563502857142858e-05,
+      "loss": 1.9398,
+      "step": 5721
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2563234285714286e-05,
+      "loss": 1.0453,
+      "step": 5722
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2562965714285716e-05,
+      "loss": 1.2239,
+      "step": 5723
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2562697142857142e-05,
+      "loss": 1.5638,
+      "step": 5724
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2562428571428572e-05,
+      "loss": 1.4128,
+      "step": 5725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256216e-05,
+      "loss": 1.3258,
+      "step": 5726
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256189142857143e-05,
+      "loss": 1.4041,
+      "step": 5727
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2561622857142857e-05,
+      "loss": 1.3625,
+      "step": 5728
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2561354285714287e-05,
+      "loss": 1.1613,
+      "step": 5729
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2561085714285715e-05,
+      "loss": 1.7302,
+      "step": 5730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2560817142857143e-05,
+      "loss": 1.6911,
+      "step": 5731
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2560548571428571e-05,
+      "loss": 1.4901,
+      "step": 5732
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2560280000000001e-05,
+      "loss": 1.9164,
+      "step": 5733
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.256001142857143e-05,
+      "loss": 1.2291,
+      "step": 5734
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2559742857142857e-05,
+      "loss": 1.273,
+      "step": 5735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2559474285714287e-05,
+      "loss": 1.286,
+      "step": 5736
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2559205714285714e-05,
+      "loss": 1.5919,
+      "step": 5737
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2558937142857144e-05,
+      "loss": 1.2069,
+      "step": 5738
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2558668571428572e-05,
+      "loss": 1.3502,
+      "step": 5739
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2558400000000002e-05,
+      "loss": 1.2043,
+      "step": 5740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2558131428571428e-05,
+      "loss": 1.1488,
+      "step": 5741
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2557862857142858e-05,
+      "loss": 1.4405,
+      "step": 5742
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2557594285714286e-05,
+      "loss": 1.2835,
+      "step": 5743
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2557325714285714e-05,
+      "loss": 1.4832,
+      "step": 5744
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2557057142857142e-05,
+      "loss": 1.6955,
+      "step": 5745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2556788571428572e-05,
+      "loss": 1.1759,
+      "step": 5746
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.255652e-05,
+      "loss": 1.5588,
+      "step": 5747
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2556251428571429e-05,
+      "loss": 1.5745,
+      "step": 5748
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2555982857142859e-05,
+      "loss": 1.3336,
+      "step": 5749
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2555714285714285e-05,
+      "loss": 1.1813,
+      "step": 5750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2555445714285715e-05,
+      "loss": 1.7719,
+      "step": 5751
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2555177142857143e-05,
+      "loss": 1.3492,
+      "step": 5752
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2554908571428573e-05,
+      "loss": 1.5614,
+      "step": 5753
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.255464e-05,
+      "loss": 1.7105,
+      "step": 5754
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.255437142857143e-05,
+      "loss": 1.2076,
+      "step": 5755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2554102857142857e-05,
+      "loss": 1.0447,
+      "step": 5756
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2553834285714286e-05,
+      "loss": 1.6756,
+      "step": 5757
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2553565714285714e-05,
+      "loss": 1.4491,
+      "step": 5758
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2553297142857144e-05,
+      "loss": 1.3622,
+      "step": 5759
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2553028571428572e-05,
+      "loss": 1.1021,
+      "step": 5760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.255276e-05,
+      "loss": 1.3354,
+      "step": 5761
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.255249142857143e-05,
+      "loss": 1.629,
+      "step": 5762
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2552222857142858e-05,
+      "loss": 1.2913,
+      "step": 5763
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2551954285714286e-05,
+      "loss": 1.0554,
+      "step": 5764
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2551685714285714e-05,
+      "loss": 1.7835,
+      "step": 5765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2551417142857144e-05,
+      "loss": 1.3701,
+      "step": 5766
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.255114857142857e-05,
+      "loss": 1.4263,
+      "step": 5767
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.255088e-05,
+      "loss": 1.1856,
+      "step": 5768
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2550611428571429e-05,
+      "loss": 1.2616,
+      "step": 5769
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2550342857142857e-05,
+      "loss": 1.1944,
+      "step": 5770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2550074285714285e-05,
+      "loss": 1.2523,
+      "step": 5771
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2549805714285715e-05,
+      "loss": 1.4205,
+      "step": 5772
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2549537142857143e-05,
+      "loss": 1.4186,
+      "step": 5773
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2549268571428571e-05,
+      "loss": 1.1806,
+      "step": 5774
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2549000000000001e-05,
+      "loss": 1.3489,
+      "step": 5775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.254873142857143e-05,
+      "loss": 1.5402,
+      "step": 5776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2548462857142858e-05,
+      "loss": 1.2915,
+      "step": 5777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2548194285714286e-05,
+      "loss": 1.2609,
+      "step": 5778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2547925714285716e-05,
+      "loss": 1.7294,
+      "step": 5779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2547657142857142e-05,
+      "loss": 0.9405,
+      "step": 5780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2547388571428572e-05,
+      "loss": 1.2657,
+      "step": 5781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.254712e-05,
+      "loss": 1.6121,
+      "step": 5782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2546851428571428e-05,
+      "loss": 1.4704,
+      "step": 5783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2546582857142857e-05,
+      "loss": 1.2473,
+      "step": 5784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2546314285714286e-05,
+      "loss": 1.5209,
+      "step": 5785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2546045714285716e-05,
+      "loss": 1.346,
+      "step": 5786
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2545777142857143e-05,
+      "loss": 1.5951,
+      "step": 5787
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2545508571428573e-05,
+      "loss": 1.7022,
+      "step": 5788
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2545240000000001e-05,
+      "loss": 1.4123,
+      "step": 5789
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2544971428571429e-05,
+      "loss": 1.3813,
+      "step": 5790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2544702857142857e-05,
+      "loss": 1.9587,
+      "step": 5791
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2544434285714287e-05,
+      "loss": 1.4005,
+      "step": 5792
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2544165714285714e-05,
+      "loss": 1.4704,
+      "step": 5793
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2543897142857143e-05,
+      "loss": 1.6951,
+      "step": 5794
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2543628571428572e-05,
+      "loss": 1.3639,
+      "step": 5795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.254336e-05,
+      "loss": 1.4878,
+      "step": 5796
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2543091428571428e-05,
+      "loss": 1.413,
+      "step": 5797
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2542822857142858e-05,
+      "loss": 1.2282,
+      "step": 5798
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2542554285714286e-05,
+      "loss": 1.8064,
+      "step": 5799
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2542285714285714e-05,
+      "loss": 1.5836,
+      "step": 5800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2542017142857144e-05,
+      "loss": 1.3468,
+      "step": 5801
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2541748571428572e-05,
+      "loss": 1.276,
+      "step": 5802
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.254148e-05,
+      "loss": 1.4518,
+      "step": 5803
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2541211428571429e-05,
+      "loss": 1.5925,
+      "step": 5804
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2540942857142858e-05,
+      "loss": 1.3379,
+      "step": 5805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2540674285714285e-05,
+      "loss": 1.0061,
+      "step": 5806
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2540405714285715e-05,
+      "loss": 1.4901,
+      "step": 5807
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2540137142857143e-05,
+      "loss": 1.5965,
+      "step": 5808
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2539868571428573e-05,
+      "loss": 1.2115,
+      "step": 5809
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.25396e-05,
+      "loss": 1.0403,
+      "step": 5810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.253933142857143e-05,
+      "loss": 1.5228,
+      "step": 5811
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2539062857142857e-05,
+      "loss": 1.5572,
+      "step": 5812
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2538794285714286e-05,
+      "loss": 1.1862,
+      "step": 5813
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2538525714285715e-05,
+      "loss": 1.5135,
+      "step": 5814
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2538257142857144e-05,
+      "loss": 1.4688,
+      "step": 5815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2537988571428572e-05,
+      "loss": 1.1133,
+      "step": 5816
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.253772e-05,
+      "loss": 1.5157,
+      "step": 5817
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.253745142857143e-05,
+      "loss": 1.4751,
+      "step": 5818
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2537182857142856e-05,
+      "loss": 1.5335,
+      "step": 5819
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2536914285714286e-05,
+      "loss": 1.254,
+      "step": 5820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2536645714285714e-05,
+      "loss": 1.6966,
+      "step": 5821
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2536377142857144e-05,
+      "loss": 1.9787,
+      "step": 5822
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.253610857142857e-05,
+      "loss": 1.0644,
+      "step": 5823
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.253584e-05,
+      "loss": 0.9996,
+      "step": 5824
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2535571428571429e-05,
+      "loss": 1.265,
+      "step": 5825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2535302857142857e-05,
+      "loss": 1.4211,
+      "step": 5826
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2535034285714287e-05,
+      "loss": 1.6106,
+      "step": 5827
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2534765714285715e-05,
+      "loss": 1.3355,
+      "step": 5828
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2534497142857143e-05,
+      "loss": 1.4366,
+      "step": 5829
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2534228571428571e-05,
+      "loss": 1.3028,
+      "step": 5830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2533960000000001e-05,
+      "loss": 1.4102,
+      "step": 5831
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.253369142857143e-05,
+      "loss": 1.5429,
+      "step": 5832
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2533422857142858e-05,
+      "loss": 1.2529,
+      "step": 5833
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2533154285714286e-05,
+      "loss": 1.4249,
+      "step": 5834
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2532885714285716e-05,
+      "loss": 1.4058,
+      "step": 5835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2532617142857142e-05,
+      "loss": 1.2285,
+      "step": 5836
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2532348571428572e-05,
+      "loss": 1.4095,
+      "step": 5837
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.253208e-05,
+      "loss": 1.5333,
+      "step": 5838
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2531811428571428e-05,
+      "loss": 1.3675,
+      "step": 5839
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2531542857142858e-05,
+      "loss": 1.7141,
+      "step": 5840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2531274285714286e-05,
+      "loss": 1.6505,
+      "step": 5841
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2531005714285715e-05,
+      "loss": 1.4986,
+      "step": 5842
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2530737142857143e-05,
+      "loss": 1.5455,
+      "step": 5843
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2530468571428573e-05,
+      "loss": 1.5792,
+      "step": 5844
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.25302e-05,
+      "loss": 1.4098,
+      "step": 5845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2529931428571429e-05,
+      "loss": 1.7991,
+      "step": 5846
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2529662857142857e-05,
+      "loss": 1.0353,
+      "step": 5847
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2529394285714287e-05,
+      "loss": 1.3712,
+      "step": 5848
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2529125714285713e-05,
+      "loss": 1.5855,
+      "step": 5849
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2528857142857143e-05,
+      "loss": 1.6997,
+      "step": 5850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2528588571428572e-05,
+      "loss": 1.2438,
+      "step": 5851
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.252832e-05,
+      "loss": 1.4052,
+      "step": 5852
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.252805142857143e-05,
+      "loss": 1.2433,
+      "step": 5853
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2527782857142858e-05,
+      "loss": 1.5594,
+      "step": 5854
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2527514285714288e-05,
+      "loss": 1.4525,
+      "step": 5855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2527245714285714e-05,
+      "loss": 1.1783,
+      "step": 5856
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2526977142857144e-05,
+      "loss": 1.5205,
+      "step": 5857
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2526708571428572e-05,
+      "loss": 1.3285,
+      "step": 5858
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.252644e-05,
+      "loss": 1.6985,
+      "step": 5859
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2526171428571428e-05,
+      "loss": 1.2052,
+      "step": 5860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2525902857142858e-05,
+      "loss": 1.6443,
+      "step": 5861
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.2525634285714285e-05,
+      "loss": 1.2154,
+      "step": 5862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2525365714285715e-05,
+      "loss": 1.3287,
+      "step": 5863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2525097142857143e-05,
+      "loss": 1.3032,
+      "step": 5864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2524828571428571e-05,
+      "loss": 1.5481,
+      "step": 5865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2524560000000001e-05,
+      "loss": 1.4414,
+      "step": 5866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2524291428571429e-05,
+      "loss": 1.597,
+      "step": 5867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2524022857142859e-05,
+      "loss": 1.1458,
+      "step": 5868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2523754285714285e-05,
+      "loss": 1.4641,
+      "step": 5869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2523485714285715e-05,
+      "loss": 1.4227,
+      "step": 5870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2523217142857143e-05,
+      "loss": 1.0395,
+      "step": 5871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2522948571428572e-05,
+      "loss": 1.4536,
+      "step": 5872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.252268e-05,
+      "loss": 1.1457,
+      "step": 5873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.252241142857143e-05,
+      "loss": 1.5798,
+      "step": 5874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2522142857142856e-05,
+      "loss": 1.1378,
+      "step": 5875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2521874285714286e-05,
+      "loss": 1.7984,
+      "step": 5876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2521605714285714e-05,
+      "loss": 1.6762,
+      "step": 5877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2521337142857144e-05,
+      "loss": 1.5803,
+      "step": 5878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2521068571428572e-05,
+      "loss": 1.5248,
+      "step": 5879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.25208e-05,
+      "loss": 1.1874,
+      "step": 5880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.252053142857143e-05,
+      "loss": 1.3776,
+      "step": 5881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2520262857142857e-05,
+      "loss": 1.3732,
+      "step": 5882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2519994285714287e-05,
+      "loss": 1.1565,
+      "step": 5883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2519725714285715e-05,
+      "loss": 1.5578,
+      "step": 5884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2519457142857143e-05,
+      "loss": 1.0215,
+      "step": 5885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2519188571428571e-05,
+      "loss": 1.4287,
+      "step": 5886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2518920000000001e-05,
+      "loss": 1.4667,
+      "step": 5887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2518651428571428e-05,
+      "loss": 1.4687,
+      "step": 5888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2518382857142857e-05,
+      "loss": 1.5151,
+      "step": 5889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2518114285714286e-05,
+      "loss": 1.3911,
+      "step": 5890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2517845714285715e-05,
+      "loss": 1.4859,
+      "step": 5891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2517577142857144e-05,
+      "loss": 1.3834,
+      "step": 5892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2517308571428572e-05,
+      "loss": 1.6897,
+      "step": 5893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2517040000000002e-05,
+      "loss": 1.1899,
+      "step": 5894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2516771428571428e-05,
+      "loss": 1.3606,
+      "step": 5895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2516502857142858e-05,
+      "loss": 1.3131,
+      "step": 5896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2516234285714286e-05,
+      "loss": 1.0727,
+      "step": 5897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2515965714285714e-05,
+      "loss": 1.5259,
+      "step": 5898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2515697142857143e-05,
+      "loss": 1.8549,
+      "step": 5899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2515428571428572e-05,
+      "loss": 1.5309,
+      "step": 5900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2515159999999999e-05,
+      "loss": 1.3349,
+      "step": 5901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2514891428571429e-05,
+      "loss": 1.6138,
+      "step": 5902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2514622857142857e-05,
+      "loss": 1.5292,
+      "step": 5903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2514354285714287e-05,
+      "loss": 1.3828,
+      "step": 5904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2514085714285715e-05,
+      "loss": 1.2855,
+      "step": 5905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2513817142857143e-05,
+      "loss": 1.5436,
+      "step": 5906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2513548571428573e-05,
+      "loss": 1.3916,
+      "step": 5907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.251328e-05,
+      "loss": 1.3164,
+      "step": 5908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.251301142857143e-05,
+      "loss": 1.675,
+      "step": 5909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2512742857142858e-05,
+      "loss": 1.4919,
+      "step": 5910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2512474285714286e-05,
+      "loss": 1.2268,
+      "step": 5911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2512205714285714e-05,
+      "loss": 1.8937,
+      "step": 5912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2511937142857144e-05,
+      "loss": 1.2616,
+      "step": 5913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2511668571428572e-05,
+      "loss": 1.2873,
+      "step": 5914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.25114e-05,
+      "loss": 1.4279,
+      "step": 5915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2511131428571428e-05,
+      "loss": 1.2974,
+      "step": 5916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2510862857142858e-05,
+      "loss": 1.0522,
+      "step": 5917
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2510594285714286e-05,
+      "loss": 1.3936,
+      "step": 5918
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2510325714285715e-05,
+      "loss": 1.5116,
+      "step": 5919
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2510057142857144e-05,
+      "loss": 1.3439,
+      "step": 5920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2509788571428571e-05,
+      "loss": 1.2635,
+      "step": 5921
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.250952e-05,
+      "loss": 1.2641,
+      "step": 5922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2509251428571429e-05,
+      "loss": 1.1679,
+      "step": 5923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2508982857142857e-05,
+      "loss": 1.4732,
+      "step": 5924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2508714285714285e-05,
+      "loss": 1.6232,
+      "step": 5925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2508445714285715e-05,
+      "loss": 1.3957,
+      "step": 5926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2508177142857143e-05,
+      "loss": 1.5069,
+      "step": 5927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2507908571428572e-05,
+      "loss": 1.1848,
+      "step": 5928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.250764e-05,
+      "loss": 0.9641,
+      "step": 5929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.250737142857143e-05,
+      "loss": 1.738,
+      "step": 5930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2507102857142858e-05,
+      "loss": 1.6902,
+      "step": 5931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2506834285714286e-05,
+      "loss": 1.726,
+      "step": 5932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2506565714285716e-05,
+      "loss": 1.4241,
+      "step": 5933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2506297142857142e-05,
+      "loss": 1.3485,
+      "step": 5934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2506028571428572e-05,
+      "loss": 1.619,
+      "step": 5935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.250576e-05,
+      "loss": 1.3121,
+      "step": 5936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.250549142857143e-05,
+      "loss": 1.2085,
+      "step": 5937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2505222857142857e-05,
+      "loss": 1.5353,
+      "step": 5938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2504954285714287e-05,
+      "loss": 1.6126,
+      "step": 5939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2504685714285715e-05,
+      "loss": 1.4046,
+      "step": 5940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2504417142857143e-05,
+      "loss": 1.3451,
+      "step": 5941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2504148571428571e-05,
+      "loss": 1.4291,
+      "step": 5942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2503880000000001e-05,
+      "loss": 1.2373,
+      "step": 5943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2503611428571429e-05,
+      "loss": 1.494,
+      "step": 5944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2503342857142857e-05,
+      "loss": 1.436,
+      "step": 5945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2503074285714287e-05,
+      "loss": 1.0847,
+      "step": 5946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2502805714285714e-05,
+      "loss": 1.2521,
+      "step": 5947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2502537142857144e-05,
+      "loss": 1.3336,
+      "step": 5948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2502268571428572e-05,
+      "loss": 1.3258,
+      "step": 5949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2502000000000002e-05,
+      "loss": 1.471,
+      "step": 5950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2501731428571428e-05,
+      "loss": 1.2917,
+      "step": 5951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2501462857142858e-05,
+      "loss": 1.2752,
+      "step": 5952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2501194285714286e-05,
+      "loss": 1.4394,
+      "step": 5953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2500925714285714e-05,
+      "loss": 1.4778,
+      "step": 5954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2500657142857142e-05,
+      "loss": 1.3274,
+      "step": 5955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2500388571428572e-05,
+      "loss": 1.3665,
+      "step": 5956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.250012e-05,
+      "loss": 1.5728,
+      "step": 5957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2499851428571429e-05,
+      "loss": 1.5306,
+      "step": 5958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2499582857142859e-05,
+      "loss": 1.1452,
+      "step": 5959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2499314285714287e-05,
+      "loss": 1.5776,
+      "step": 5960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2499045714285715e-05,
+      "loss": 1.5874,
+      "step": 5961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2498777142857143e-05,
+      "loss": 1.2019,
+      "step": 5962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2498508571428573e-05,
+      "loss": 1.4284,
+      "step": 5963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249824e-05,
+      "loss": 1.3412,
+      "step": 5964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249797142857143e-05,
+      "loss": 1.2364,
+      "step": 5965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2497702857142857e-05,
+      "loss": 1.0747,
+      "step": 5966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2497434285714286e-05,
+      "loss": 1.4086,
+      "step": 5967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2497165714285714e-05,
+      "loss": 1.7097,
+      "step": 5968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2496897142857144e-05,
+      "loss": 1.1395,
+      "step": 5969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2496628571428572e-05,
+      "loss": 1.5291,
+      "step": 5970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249636e-05,
+      "loss": 1.4379,
+      "step": 5971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249609142857143e-05,
+      "loss": 1.4653,
+      "step": 5972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2495822857142858e-05,
+      "loss": 1.5347,
+      "step": 5973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2495554285714286e-05,
+      "loss": 1.3779,
+      "step": 5974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2495285714285714e-05,
+      "loss": 1.7324,
+      "step": 5975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2495017142857144e-05,
+      "loss": 1.8665,
+      "step": 5976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249474857142857e-05,
+      "loss": 1.2529,
+      "step": 5977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249448e-05,
+      "loss": 1.3023,
+      "step": 5978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2494211428571429e-05,
+      "loss": 1.6893,
+      "step": 5979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2493942857142857e-05,
+      "loss": 1.636,
+      "step": 5980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2493674285714285e-05,
+      "loss": 1.0515,
+      "step": 5981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2493405714285715e-05,
+      "loss": 1.0253,
+      "step": 5982
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2493137142857143e-05,
+      "loss": 1.4837,
+      "step": 5983
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2492868571428571e-05,
+      "loss": 0.9898,
+      "step": 5984
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2492600000000001e-05,
+      "loss": 1.2747,
+      "step": 5985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249233142857143e-05,
+      "loss": 1.3505,
+      "step": 5986
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2492062857142858e-05,
+      "loss": 1.2708,
+      "step": 5987
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2491794285714286e-05,
+      "loss": 1.1298,
+      "step": 5988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2491525714285716e-05,
+      "loss": 1.4226,
+      "step": 5989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2491257142857142e-05,
+      "loss": 1.6307,
+      "step": 5990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2490988571428572e-05,
+      "loss": 1.5385,
+      "step": 5991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.249072e-05,
+      "loss": 1.214,
+      "step": 5992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2490451428571428e-05,
+      "loss": 1.6485,
+      "step": 5993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2490182857142857e-05,
+      "loss": 1.3364,
+      "step": 5994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2489914285714286e-05,
+      "loss": 1.2152,
+      "step": 5995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2489645714285715e-05,
+      "loss": 1.6615,
+      "step": 5996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2489377142857143e-05,
+      "loss": 1.482,
+      "step": 5997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2489108571428573e-05,
+      "loss": 1.5716,
+      "step": 5998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2488840000000001e-05,
+      "loss": 1.8868,
+      "step": 5999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2488571428571429e-05,
+      "loss": 1.5776,
+      "step": 6000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2488302857142857e-05,
+      "loss": 1.2454,
+      "step": 6001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2488034285714287e-05,
+      "loss": 1.2978,
+      "step": 6002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2487765714285714e-05,
+      "loss": 1.3467,
+      "step": 6003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2487497142857143e-05,
+      "loss": 1.4438,
+      "step": 6004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2487228571428572e-05,
+      "loss": 1.414,
+      "step": 6005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2486960000000001e-05,
+      "loss": 1.8916,
+      "step": 6006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2486691428571428e-05,
+      "loss": 1.4627,
+      "step": 6007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2486422857142858e-05,
+      "loss": 1.8742,
+      "step": 6008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2486154285714286e-05,
+      "loss": 1.1049,
+      "step": 6009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2485885714285714e-05,
+      "loss": 1.1722,
+      "step": 6010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2485617142857144e-05,
+      "loss": 1.4848,
+      "step": 6011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2485348571428572e-05,
+      "loss": 1.2909,
+      "step": 6012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.248508e-05,
+      "loss": 1.7376,
+      "step": 6013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2484811428571429e-05,
+      "loss": 1.2317,
+      "step": 6014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2484542857142858e-05,
+      "loss": 1.891,
+      "step": 6015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2484274285714285e-05,
+      "loss": 1.5122,
+      "step": 6016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2484005714285715e-05,
+      "loss": 1.6097,
+      "step": 6017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2483737142857143e-05,
+      "loss": 1.6006,
+      "step": 6018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2483468571428573e-05,
+      "loss": 1.6646,
+      "step": 6019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.24832e-05,
+      "loss": 0.8643,
+      "step": 6020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.248293142857143e-05,
+      "loss": 1.3755,
+      "step": 6021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2482662857142857e-05,
+      "loss": 1.5232,
+      "step": 6022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2482394285714286e-05,
+      "loss": 1.4476,
+      "step": 6023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2482125714285715e-05,
+      "loss": 1.5388,
+      "step": 6024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2481857142857144e-05,
+      "loss": 1.1459,
+      "step": 6025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2481588571428572e-05,
+      "loss": 1.937,
+      "step": 6026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.248132e-05,
+      "loss": 1.6247,
+      "step": 6027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.248105142857143e-05,
+      "loss": 1.6401,
+      "step": 6028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2480782857142858e-05,
+      "loss": 1.434,
+      "step": 6029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2480514285714286e-05,
+      "loss": 1.5793,
+      "step": 6030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2480245714285714e-05,
+      "loss": 1.156,
+      "step": 6031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2479977142857144e-05,
+      "loss": 1.5101,
+      "step": 6032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.247970857142857e-05,
+      "loss": 1.4809,
+      "step": 6033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.247944e-05,
+      "loss": 1.4484,
+      "step": 6034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2479171428571429e-05,
+      "loss": 1.4329,
+      "step": 6035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2478902857142857e-05,
+      "loss": 1.1966,
+      "step": 6036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2478634285714287e-05,
+      "loss": 1.229,
+      "step": 6037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2478365714285715e-05,
+      "loss": 1.1159,
+      "step": 6038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2478097142857143e-05,
+      "loss": 1.2421,
+      "step": 6039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2477828571428571e-05,
+      "loss": 1.445,
+      "step": 6040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2477560000000001e-05,
+      "loss": 1.4923,
+      "step": 6041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.247729142857143e-05,
+      "loss": 1.831,
+      "step": 6042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2477022857142858e-05,
+      "loss": 1.1307,
+      "step": 6043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2476754285714286e-05,
+      "loss": 1.7527,
+      "step": 6044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2476485714285716e-05,
+      "loss": 1.3554,
+      "step": 6045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2476217142857142e-05,
+      "loss": 1.0759,
+      "step": 6046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2475948571428572e-05,
+      "loss": 1.4564,
+      "step": 6047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.247568e-05,
+      "loss": 1.3942,
+      "step": 6048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2475411428571428e-05,
+      "loss": 1.4072,
+      "step": 6049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2475142857142858e-05,
+      "loss": 1.4136,
+      "step": 6050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2474874285714286e-05,
+      "loss": 1.3179,
+      "step": 6051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2474605714285716e-05,
+      "loss": 1.4868,
+      "step": 6052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2474337142857143e-05,
+      "loss": 1.5969,
+      "step": 6053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2474068571428573e-05,
+      "loss": 1.5161,
+      "step": 6054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.24738e-05,
+      "loss": 1.4008,
+      "step": 6055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2473531428571429e-05,
+      "loss": 1.1568,
+      "step": 6056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2473262857142857e-05,
+      "loss": 1.4253,
+      "step": 6057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2472994285714287e-05,
+      "loss": 1.5403,
+      "step": 6058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2472725714285713e-05,
+      "loss": 1.2716,
+      "step": 6059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2472457142857143e-05,
+      "loss": 1.2641,
+      "step": 6060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2472188571428571e-05,
+      "loss": 1.5528,
+      "step": 6061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.247192e-05,
+      "loss": 1.651,
+      "step": 6062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.247165142857143e-05,
+      "loss": 1.676,
+      "step": 6063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2471382857142858e-05,
+      "loss": 1.4685,
+      "step": 6064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2471114285714288e-05,
+      "loss": 1.4471,
+      "step": 6065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2470845714285714e-05,
+      "loss": 1.0849,
+      "step": 6066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2470577142857144e-05,
+      "loss": 1.0734,
+      "step": 6067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2470308571428572e-05,
+      "loss": 1.2055,
+      "step": 6068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.247004e-05,
+      "loss": 1.0359,
+      "step": 6069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2469771428571428e-05,
+      "loss": 1.324,
+      "step": 6070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2469502857142858e-05,
+      "loss": 1.6813,
+      "step": 6071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2469234285714285e-05,
+      "loss": 1.4786,
+      "step": 6072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2468965714285715e-05,
+      "loss": 1.7529,
+      "step": 6073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2468697142857143e-05,
+      "loss": 1.3824,
+      "step": 6074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2468428571428571e-05,
+      "loss": 1.4456,
+      "step": 6075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2468160000000001e-05,
+      "loss": 1.4183,
+      "step": 6076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2467891428571429e-05,
+      "loss": 1.2361,
+      "step": 6077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2467622857142859e-05,
+      "loss": 1.4758,
+      "step": 6078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2467354285714285e-05,
+      "loss": 1.5359,
+      "step": 6079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2467085714285715e-05,
+      "loss": 1.4975,
+      "step": 6080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2466817142857143e-05,
+      "loss": 1.2081,
+      "step": 6081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2466548571428572e-05,
+      "loss": 1.7107,
+      "step": 6082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.246628e-05,
+      "loss": 1.3203,
+      "step": 6083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.246601142857143e-05,
+      "loss": 1.5473,
+      "step": 6084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2465742857142856e-05,
+      "loss": 1.461,
+      "step": 6085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2465474285714286e-05,
+      "loss": 1.2074,
+      "step": 6086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2465205714285714e-05,
+      "loss": 1.6296,
+      "step": 6087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2464937142857144e-05,
+      "loss": 1.5802,
+      "step": 6088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2464668571428572e-05,
+      "loss": 1.25,
+      "step": 6089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.24644e-05,
+      "loss": 1.7593,
+      "step": 6090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.246413142857143e-05,
+      "loss": 1.124,
+      "step": 6091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2463862857142857e-05,
+      "loss": 1.4869,
+      "step": 6092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2463594285714287e-05,
+      "loss": 1.5847,
+      "step": 6093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2463325714285715e-05,
+      "loss": 1.6009,
+      "step": 6094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2463057142857143e-05,
+      "loss": 1.6403,
+      "step": 6095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2462788571428571e-05,
+      "loss": 1.2499,
+      "step": 6096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2462520000000001e-05,
+      "loss": 1.32,
+      "step": 6097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2462251428571428e-05,
+      "loss": 1.3634,
+      "step": 6098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2461982857142857e-05,
+      "loss": 1.1578,
+      "step": 6099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2461714285714286e-05,
+      "loss": 1.1029,
+      "step": 6100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2461445714285715e-05,
+      "loss": 1.3779,
+      "step": 6101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2461177142857144e-05,
+      "loss": 1.4093,
+      "step": 6102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2460908571428572e-05,
+      "loss": 2.1084,
+      "step": 6103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2460640000000002e-05,
+      "loss": 1.6542,
+      "step": 6104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2460371428571428e-05,
+      "loss": 1.5004,
+      "step": 6105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2460102857142858e-05,
+      "loss": 1.4906,
+      "step": 6106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2459834285714286e-05,
+      "loss": 1.0606,
+      "step": 6107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2459565714285714e-05,
+      "loss": 1.3759,
+      "step": 6108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2459297142857143e-05,
+      "loss": 1.8838,
+      "step": 6109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2459028571428572e-05,
+      "loss": 1.169,
+      "step": 6110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.245876e-05,
+      "loss": 1.1895,
+      "step": 6111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2458491428571429e-05,
+      "loss": 1.6474,
+      "step": 6112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2458222857142857e-05,
+      "loss": 1.3583,
+      "step": 6113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2457954285714287e-05,
+      "loss": 1.3652,
+      "step": 6114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2457685714285715e-05,
+      "loss": 1.3095,
+      "step": 6115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2457417142857143e-05,
+      "loss": 1.2524,
+      "step": 6116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2457148571428573e-05,
+      "loss": 1.3661,
+      "step": 6117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.245688e-05,
+      "loss": 1.8329,
+      "step": 6118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.245661142857143e-05,
+      "loss": 1.4426,
+      "step": 6119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2456342857142858e-05,
+      "loss": 1.4419,
+      "step": 6120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2456074285714286e-05,
+      "loss": 1.3257,
+      "step": 6121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2455805714285714e-05,
+      "loss": 1.1966,
+      "step": 6122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2455537142857144e-05,
+      "loss": 1.1688,
+      "step": 6123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2455268571428572e-05,
+      "loss": 1.6407,
+      "step": 6124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2455e-05,
+      "loss": 1.1667,
+      "step": 6125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2454731428571428e-05,
+      "loss": 1.5986,
+      "step": 6126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2454462857142858e-05,
+      "loss": 1.7105,
+      "step": 6127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2454194285714286e-05,
+      "loss": 1.4375,
+      "step": 6128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2453925714285715e-05,
+      "loss": 1.2924,
+      "step": 6129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2453657142857144e-05,
+      "loss": 1.2387,
+      "step": 6130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2453388571428571e-05,
+      "loss": 1.3885,
+      "step": 6131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.245312e-05,
+      "loss": 1.2004,
+      "step": 6132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2452851428571429e-05,
+      "loss": 1.2913,
+      "step": 6133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2452582857142859e-05,
+      "loss": 1.1614,
+      "step": 6134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2452314285714285e-05,
+      "loss": 1.0867,
+      "step": 6135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2452045714285715e-05,
+      "loss": 1.2134,
+      "step": 6136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2451777142857143e-05,
+      "loss": 1.0307,
+      "step": 6137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2451508571428572e-05,
+      "loss": 1.3176,
+      "step": 6138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.245124e-05,
+      "loss": 1.2437,
+      "step": 6139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.245097142857143e-05,
+      "loss": 0.9149,
+      "step": 6140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2450702857142858e-05,
+      "loss": 1.1901,
+      "step": 6141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2450434285714286e-05,
+      "loss": 1.4975,
+      "step": 6142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2450165714285716e-05,
+      "loss": 1.5602,
+      "step": 6143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2449897142857142e-05,
+      "loss": 1.6319,
+      "step": 6144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2449628571428572e-05,
+      "loss": 1.2395,
+      "step": 6145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.244936e-05,
+      "loss": 0.9415,
+      "step": 6146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.244909142857143e-05,
+      "loss": 1.2948,
+      "step": 6147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2448822857142857e-05,
+      "loss": 1.7381,
+      "step": 6148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2448554285714287e-05,
+      "loss": 1.4758,
+      "step": 6149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2448285714285715e-05,
+      "loss": 1.3501,
+      "step": 6150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2448017142857143e-05,
+      "loss": 1.0612,
+      "step": 6151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2447748571428571e-05,
+      "loss": 1.3329,
+      "step": 6152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2447480000000001e-05,
+      "loss": 1.2433,
+      "step": 6153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2447211428571429e-05,
+      "loss": 1.068,
+      "step": 6154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2446942857142857e-05,
+      "loss": 1.1771,
+      "step": 6155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2446674285714287e-05,
+      "loss": 1.5574,
+      "step": 6156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2446405714285715e-05,
+      "loss": 1.4227,
+      "step": 6157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2446137142857144e-05,
+      "loss": 1.2335,
+      "step": 6158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2445868571428572e-05,
+      "loss": 1.6208,
+      "step": 6159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2445600000000002e-05,
+      "loss": 1.5164,
+      "step": 6160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2445331428571428e-05,
+      "loss": 1.6722,
+      "step": 6161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2445062857142858e-05,
+      "loss": 1.6363,
+      "step": 6162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2444794285714286e-05,
+      "loss": 1.4345,
+      "step": 6163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2444525714285714e-05,
+      "loss": 1.5702,
+      "step": 6164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2444257142857142e-05,
+      "loss": 1.5831,
+      "step": 6165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2443988571428572e-05,
+      "loss": 1.318,
+      "step": 6166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.244372e-05,
+      "loss": 1.8047,
+      "step": 6167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2443451428571429e-05,
+      "loss": 1.3096,
+      "step": 6168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2443182857142859e-05,
+      "loss": 1.6153,
+      "step": 6169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2442914285714287e-05,
+      "loss": 1.439,
+      "step": 6170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2442645714285715e-05,
+      "loss": 1.4791,
+      "step": 6171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2442377142857143e-05,
+      "loss": 1.6861,
+      "step": 6172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2442108571428573e-05,
+      "loss": 1.7054,
+      "step": 6173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.244184e-05,
+      "loss": 1.188,
+      "step": 6174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.244157142857143e-05,
+      "loss": 1.4315,
+      "step": 6175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2441302857142857e-05,
+      "loss": 1.3877,
+      "step": 6176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2441034285714286e-05,
+      "loss": 1.5234,
+      "step": 6177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2440765714285714e-05,
+      "loss": 1.2752,
+      "step": 6178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2440497142857144e-05,
+      "loss": 1.1926,
+      "step": 6179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2440228571428572e-05,
+      "loss": 1.9179,
+      "step": 6180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.243996e-05,
+      "loss": 1.3422,
+      "step": 6181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.243969142857143e-05,
+      "loss": 1.3307,
+      "step": 6182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2439422857142858e-05,
+      "loss": 1.3398,
+      "step": 6183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2439154285714286e-05,
+      "loss": 1.4084,
+      "step": 6184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2438885714285714e-05,
+      "loss": 1.5809,
+      "step": 6185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2438617142857144e-05,
+      "loss": 1.2771,
+      "step": 6186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.243834857142857e-05,
+      "loss": 1.6239,
+      "step": 6187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.243808e-05,
+      "loss": 1.4502,
+      "step": 6188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2437811428571429e-05,
+      "loss": 1.3328,
+      "step": 6189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2437542857142857e-05,
+      "loss": 1.255,
+      "step": 6190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2437274285714285e-05,
+      "loss": 1.6219,
+      "step": 6191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2437005714285715e-05,
+      "loss": 1.1306,
+      "step": 6192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2436737142857143e-05,
+      "loss": 1.3847,
+      "step": 6193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2436468571428571e-05,
+      "loss": 1.6137,
+      "step": 6194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2436200000000001e-05,
+      "loss": 1.6609,
+      "step": 6195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.243593142857143e-05,
+      "loss": 1.459,
+      "step": 6196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2435662857142858e-05,
+      "loss": 1.5804,
+      "step": 6197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2435394285714286e-05,
+      "loss": 1.8747,
+      "step": 6198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2435125714285716e-05,
+      "loss": 1.8616,
+      "step": 6199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2434857142857142e-05,
+      "loss": 1.524,
+      "step": 6200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2434588571428572e-05,
+      "loss": 1.7889,
+      "step": 6201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.243432e-05,
+      "loss": 1.1848,
+      "step": 6202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.243405142857143e-05,
+      "loss": 1.0157,
+      "step": 6203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2433782857142857e-05,
+      "loss": 1.158,
+      "step": 6204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2433514285714286e-05,
+      "loss": 1.3286,
+      "step": 6205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2433245714285715e-05,
+      "loss": 1.4024,
+      "step": 6206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2432977142857143e-05,
+      "loss": 1.5713,
+      "step": 6207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2432708571428573e-05,
+      "loss": 1.3047,
+      "step": 6208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2432440000000001e-05,
+      "loss": 1.8825,
+      "step": 6209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2432171428571429e-05,
+      "loss": 1.5496,
+      "step": 6210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2431902857142857e-05,
+      "loss": 1.5576,
+      "step": 6211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.2431634285714287e-05,
+      "loss": 1.3144,
+      "step": 6212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2431365714285714e-05,
+      "loss": 1.8143,
+      "step": 6213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2431097142857143e-05,
+      "loss": 1.5785,
+      "step": 6214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2430828571428572e-05,
+      "loss": 1.4265,
+      "step": 6215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2430560000000001e-05,
+      "loss": 1.4095,
+      "step": 6216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2430291428571428e-05,
+      "loss": 1.4818,
+      "step": 6217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2430022857142858e-05,
+      "loss": 1.6805,
+      "step": 6218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2429754285714286e-05,
+      "loss": 1.7717,
+      "step": 6219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2429485714285714e-05,
+      "loss": 1.8611,
+      "step": 6220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2429217142857144e-05,
+      "loss": 1.6009,
+      "step": 6221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2428948571428572e-05,
+      "loss": 1.3516,
+      "step": 6222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.242868e-05,
+      "loss": 1.2554,
+      "step": 6223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2428411428571429e-05,
+      "loss": 1.577,
+      "step": 6224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2428142857142858e-05,
+      "loss": 0.9971,
+      "step": 6225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2427874285714287e-05,
+      "loss": 1.619,
+      "step": 6226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2427605714285715e-05,
+      "loss": 1.421,
+      "step": 6227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2427337142857143e-05,
+      "loss": 1.4248,
+      "step": 6228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2427068571428573e-05,
+      "loss": 1.0226,
+      "step": 6229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.24268e-05,
+      "loss": 1.7772,
+      "step": 6230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.242653142857143e-05,
+      "loss": 1.2017,
+      "step": 6231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2426262857142857e-05,
+      "loss": 1.1444,
+      "step": 6232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2425994285714286e-05,
+      "loss": 1.0601,
+      "step": 6233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2425725714285715e-05,
+      "loss": 1.7219,
+      "step": 6234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2425457142857144e-05,
+      "loss": 1.5583,
+      "step": 6235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2425188571428572e-05,
+      "loss": 1.4838,
+      "step": 6236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.242492e-05,
+      "loss": 1.7126,
+      "step": 6237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.242465142857143e-05,
+      "loss": 1.3049,
+      "step": 6238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2424382857142858e-05,
+      "loss": 1.6374,
+      "step": 6239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2424114285714286e-05,
+      "loss": 1.0837,
+      "step": 6240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2423845714285714e-05,
+      "loss": 1.3337,
+      "step": 6241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2423577142857144e-05,
+      "loss": 1.1148,
+      "step": 6242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.242330857142857e-05,
+      "loss": 1.0351,
+      "step": 6243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.242304e-05,
+      "loss": 1.219,
+      "step": 6244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2422771428571429e-05,
+      "loss": 1.5208,
+      "step": 6245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2422502857142857e-05,
+      "loss": 1.307,
+      "step": 6246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2422234285714287e-05,
+      "loss": 1.67,
+      "step": 6247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2421965714285715e-05,
+      "loss": 1.2154,
+      "step": 6248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2421697142857143e-05,
+      "loss": 1.4312,
+      "step": 6249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2421428571428571e-05,
+      "loss": 1.0547,
+      "step": 6250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2421160000000001e-05,
+      "loss": 1.4318,
+      "step": 6251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.242089142857143e-05,
+      "loss": 1.6576,
+      "step": 6252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2420622857142858e-05,
+      "loss": 1.1942,
+      "step": 6253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2420354285714286e-05,
+      "loss": 1.3799,
+      "step": 6254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2420085714285716e-05,
+      "loss": 1.159,
+      "step": 6255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2419817142857142e-05,
+      "loss": 1.2442,
+      "step": 6256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2419548571428572e-05,
+      "loss": 1.5199,
+      "step": 6257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.241928e-05,
+      "loss": 1.5233,
+      "step": 6258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2419011428571428e-05,
+      "loss": 1.5267,
+      "step": 6259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2418742857142858e-05,
+      "loss": 1.4612,
+      "step": 6260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2418474285714286e-05,
+      "loss": 1.549,
+      "step": 6261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2418205714285716e-05,
+      "loss": 1.5018,
+      "step": 6262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2417937142857143e-05,
+      "loss": 1.1517,
+      "step": 6263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2417668571428573e-05,
+      "loss": 1.5986,
+      "step": 6264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.24174e-05,
+      "loss": 1.6228,
+      "step": 6265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2417131428571429e-05,
+      "loss": 1.7219,
+      "step": 6266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2416862857142857e-05,
+      "loss": 1.5867,
+      "step": 6267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2416594285714287e-05,
+      "loss": 1.2273,
+      "step": 6268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2416325714285713e-05,
+      "loss": 1.7207,
+      "step": 6269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2416057142857143e-05,
+      "loss": 1.0775,
+      "step": 6270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2415788571428571e-05,
+      "loss": 1.2966,
+      "step": 6271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.241552e-05,
+      "loss": 1.461,
+      "step": 6272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.241525142857143e-05,
+      "loss": 1.4365,
+      "step": 6273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2414982857142858e-05,
+      "loss": 1.3756,
+      "step": 6274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2414714285714288e-05,
+      "loss": 1.5385,
+      "step": 6275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2414445714285714e-05,
+      "loss": 1.4677,
+      "step": 6276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2414177142857144e-05,
+      "loss": 1.1731,
+      "step": 6277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2413908571428572e-05,
+      "loss": 1.6433,
+      "step": 6278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.241364e-05,
+      "loss": 1.2546,
+      "step": 6279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2413371428571428e-05,
+      "loss": 1.7641,
+      "step": 6280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2413102857142858e-05,
+      "loss": 1.5465,
+      "step": 6281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2412834285714285e-05,
+      "loss": 1.2888,
+      "step": 6282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2412565714285715e-05,
+      "loss": 1.3858,
+      "step": 6283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2412297142857143e-05,
+      "loss": 1.3157,
+      "step": 6284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2412028571428573e-05,
+      "loss": 1.1376,
+      "step": 6285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2411760000000001e-05,
+      "loss": 1.5425,
+      "step": 6286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2411491428571429e-05,
+      "loss": 1.901,
+      "step": 6287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2411222857142859e-05,
+      "loss": 1.3447,
+      "step": 6288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2410954285714285e-05,
+      "loss": 1.6919,
+      "step": 6289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2410685714285715e-05,
+      "loss": 1.2868,
+      "step": 6290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2410417142857143e-05,
+      "loss": 1.2489,
+      "step": 6291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2410148571428572e-05,
+      "loss": 1.3173,
+      "step": 6292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.240988e-05,
+      "loss": 1.3958,
+      "step": 6293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.240961142857143e-05,
+      "loss": 1.4695,
+      "step": 6294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2409342857142856e-05,
+      "loss": 1.0463,
+      "step": 6295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2409074285714286e-05,
+      "loss": 1.6172,
+      "step": 6296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2408805714285714e-05,
+      "loss": 1.144,
+      "step": 6297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2408537142857144e-05,
+      "loss": 0.8036,
+      "step": 6298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2408268571428572e-05,
+      "loss": 1.4716,
+      "step": 6299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2408e-05,
+      "loss": 1.6811,
+      "step": 6300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.240773142857143e-05,
+      "loss": 1.4283,
+      "step": 6301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2407462857142857e-05,
+      "loss": 1.2545,
+      "step": 6302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2407194285714287e-05,
+      "loss": 1.637,
+      "step": 6303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2406925714285715e-05,
+      "loss": 1.1745,
+      "step": 6304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2406657142857143e-05,
+      "loss": 1.248,
+      "step": 6305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2406388571428571e-05,
+      "loss": 1.5653,
+      "step": 6306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2406120000000001e-05,
+      "loss": 1.6259,
+      "step": 6307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.240585142857143e-05,
+      "loss": 1.5228,
+      "step": 6308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2405582857142857e-05,
+      "loss": 1.3353,
+      "step": 6309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2405314285714286e-05,
+      "loss": 1.0908,
+      "step": 6310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2405045714285715e-05,
+      "loss": 1.4061,
+      "step": 6311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2404777142857144e-05,
+      "loss": 1.6746,
+      "step": 6312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2404508571428572e-05,
+      "loss": 1.3582,
+      "step": 6313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2404240000000002e-05,
+      "loss": 1.2191,
+      "step": 6314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2403971428571428e-05,
+      "loss": 1.5611,
+      "step": 6315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2403702857142858e-05,
+      "loss": 1.5695,
+      "step": 6316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2403434285714286e-05,
+      "loss": 1.4767,
+      "step": 6317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2403165714285714e-05,
+      "loss": 1.3827,
+      "step": 6318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2402897142857143e-05,
+      "loss": 1.1725,
+      "step": 6319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2402628571428572e-05,
+      "loss": 1.0237,
+      "step": 6320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.240236e-05,
+      "loss": 1.6225,
+      "step": 6321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2402091428571429e-05,
+      "loss": 1.5573,
+      "step": 6322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2401822857142857e-05,
+      "loss": 1.6942,
+      "step": 6323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2401554285714287e-05,
+      "loss": 1.2042,
+      "step": 6324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2401285714285715e-05,
+      "loss": 1.4061,
+      "step": 6325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2401017142857143e-05,
+      "loss": 0.9976,
+      "step": 6326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2400748571428573e-05,
+      "loss": 1.3097,
+      "step": 6327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.240048e-05,
+      "loss": 1.2363,
+      "step": 6328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.240021142857143e-05,
+      "loss": 1.7868,
+      "step": 6329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2399942857142858e-05,
+      "loss": 1.4819,
+      "step": 6330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2399674285714287e-05,
+      "loss": 1.2116,
+      "step": 6331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2399405714285714e-05,
+      "loss": 1.2225,
+      "step": 6332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2399137142857144e-05,
+      "loss": 1.0418,
+      "step": 6333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2398868571428572e-05,
+      "loss": 1.4989,
+      "step": 6334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.23986e-05,
+      "loss": 1.1629,
+      "step": 6335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2398331428571428e-05,
+      "loss": 1.2083,
+      "step": 6336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2398062857142858e-05,
+      "loss": 1.4195,
+      "step": 6337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2397794285714286e-05,
+      "loss": 0.991,
+      "step": 6338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2397525714285715e-05,
+      "loss": 1.0386,
+      "step": 6339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2397257142857144e-05,
+      "loss": 1.2926,
+      "step": 6340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2396988571428571e-05,
+      "loss": 1.3087,
+      "step": 6341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.239672e-05,
+      "loss": 1.3372,
+      "step": 6342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2396451428571429e-05,
+      "loss": 1.4515,
+      "step": 6343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2396182857142859e-05,
+      "loss": 1.3708,
+      "step": 6344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2395914285714285e-05,
+      "loss": 1.2298,
+      "step": 6345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2395645714285715e-05,
+      "loss": 1.4041,
+      "step": 6346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2395377142857143e-05,
+      "loss": 1.807,
+      "step": 6347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2395108571428572e-05,
+      "loss": 1.6855,
+      "step": 6348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.239484e-05,
+      "loss": 1.5585,
+      "step": 6349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.239457142857143e-05,
+      "loss": 1.5286,
+      "step": 6350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2394302857142858e-05,
+      "loss": 1.1554,
+      "step": 6351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2394034285714286e-05,
+      "loss": 1.2843,
+      "step": 6352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2393765714285716e-05,
+      "loss": 1.3886,
+      "step": 6353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2393497142857144e-05,
+      "loss": 1.3989,
+      "step": 6354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2393228571428572e-05,
+      "loss": 1.2356,
+      "step": 6355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.239296e-05,
+      "loss": 1.4777,
+      "step": 6356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.239269142857143e-05,
+      "loss": 1.1231,
+      "step": 6357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2392422857142857e-05,
+      "loss": 1.4943,
+      "step": 6358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2392154285714287e-05,
+      "loss": 1.3787,
+      "step": 6359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2391885714285715e-05,
+      "loss": 1.3488,
+      "step": 6360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2391617142857143e-05,
+      "loss": 1.7393,
+      "step": 6361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2391348571428571e-05,
+      "loss": 1.2893,
+      "step": 6362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2391080000000001e-05,
+      "loss": 1.345,
+      "step": 6363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2390811428571429e-05,
+      "loss": 1.3764,
+      "step": 6364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2390542857142857e-05,
+      "loss": 1.6921,
+      "step": 6365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2390274285714287e-05,
+      "loss": 1.3246,
+      "step": 6366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2390005714285715e-05,
+      "loss": 1.7,
+      "step": 6367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2389737142857144e-05,
+      "loss": 0.9896,
+      "step": 6368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2389468571428572e-05,
+      "loss": 1.2718,
+      "step": 6369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2389200000000002e-05,
+      "loss": 1.2055,
+      "step": 6370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2388931428571428e-05,
+      "loss": 1.1371,
+      "step": 6371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2388662857142858e-05,
+      "loss": 1.5051,
+      "step": 6372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2388394285714286e-05,
+      "loss": 1.3359,
+      "step": 6373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2388125714285714e-05,
+      "loss": 1.545,
+      "step": 6374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2387857142857142e-05,
+      "loss": 1.3238,
+      "step": 6375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2387588571428572e-05,
+      "loss": 1.0951,
+      "step": 6376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.238732e-05,
+      "loss": 1.5319,
+      "step": 6377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2387051428571429e-05,
+      "loss": 1.255,
+      "step": 6378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2386782857142859e-05,
+      "loss": 1.477,
+      "step": 6379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2386514285714287e-05,
+      "loss": 1.3369,
+      "step": 6380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2386245714285715e-05,
+      "loss": 1.2821,
+      "step": 6381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2385977142857143e-05,
+      "loss": 0.9847,
+      "step": 6382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2385708571428573e-05,
+      "loss": 1.258,
+      "step": 6383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.238544e-05,
+      "loss": 1.4606,
+      "step": 6384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.238517142857143e-05,
+      "loss": 1.1882,
+      "step": 6385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2384902857142857e-05,
+      "loss": 1.1954,
+      "step": 6386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2384634285714286e-05,
+      "loss": 1.2655,
+      "step": 6387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2384365714285714e-05,
+      "loss": 1.1997,
+      "step": 6388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2384097142857144e-05,
+      "loss": 1.4103,
+      "step": 6389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2383828571428572e-05,
+      "loss": 1.4515,
+      "step": 6390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.238356e-05,
+      "loss": 1.5845,
+      "step": 6391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.238329142857143e-05,
+      "loss": 1.157,
+      "step": 6392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2383022857142858e-05,
+      "loss": 1.3289,
+      "step": 6393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2382754285714286e-05,
+      "loss": 1.7096,
+      "step": 6394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2382485714285714e-05,
+      "loss": 1.1658,
+      "step": 6395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2382217142857144e-05,
+      "loss": 1.1856,
+      "step": 6396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.238194857142857e-05,
+      "loss": 1.4247,
+      "step": 6397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.238168e-05,
+      "loss": 1.4599,
+      "step": 6398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2381411428571429e-05,
+      "loss": 1.4792,
+      "step": 6399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2381142857142857e-05,
+      "loss": 1.4595,
+      "step": 6400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2380874285714285e-05,
+      "loss": 1.2641,
+      "step": 6401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2380605714285715e-05,
+      "loss": 1.4553,
+      "step": 6402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2380337142857143e-05,
+      "loss": 1.5854,
+      "step": 6403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2380068571428571e-05,
+      "loss": 1.5377,
+      "step": 6404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2379800000000001e-05,
+      "loss": 1.4882,
+      "step": 6405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.237953142857143e-05,
+      "loss": 1.3001,
+      "step": 6406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2379262857142858e-05,
+      "loss": 1.6439,
+      "step": 6407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2378994285714286e-05,
+      "loss": 1.5384,
+      "step": 6408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2378725714285716e-05,
+      "loss": 1.2539,
+      "step": 6409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2378457142857142e-05,
+      "loss": 1.5475,
+      "step": 6410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2378188571428572e-05,
+      "loss": 1.7968,
+      "step": 6411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.237792e-05,
+      "loss": 1.265,
+      "step": 6412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.237765142857143e-05,
+      "loss": 1.3073,
+      "step": 6413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2377382857142857e-05,
+      "loss": 1.4232,
+      "step": 6414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2377114285714286e-05,
+      "loss": 1.3441,
+      "step": 6415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2376845714285715e-05,
+      "loss": 1.3738,
+      "step": 6416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2376577142857143e-05,
+      "loss": 1.8599,
+      "step": 6417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2376308571428573e-05,
+      "loss": 1.4174,
+      "step": 6418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2376040000000001e-05,
+      "loss": 1.0832,
+      "step": 6419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2375771428571429e-05,
+      "loss": 0.9889,
+      "step": 6420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2375502857142857e-05,
+      "loss": 1.231,
+      "step": 6421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2375234285714287e-05,
+      "loss": 1.44,
+      "step": 6422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2374965714285714e-05,
+      "loss": 1.0372,
+      "step": 6423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2374697142857143e-05,
+      "loss": 0.968,
+      "step": 6424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2374428571428572e-05,
+      "loss": 1.3987,
+      "step": 6425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2374160000000001e-05,
+      "loss": 1.2703,
+      "step": 6426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2373891428571428e-05,
+      "loss": 1.7275,
+      "step": 6427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2373622857142858e-05,
+      "loss": 1.4524,
+      "step": 6428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2373354285714286e-05,
+      "loss": 1.5213,
+      "step": 6429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2373085714285714e-05,
+      "loss": 1.9171,
+      "step": 6430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2372817142857144e-05,
+      "loss": 1.3746,
+      "step": 6431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2372548571428572e-05,
+      "loss": 1.5105,
+      "step": 6432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.237228e-05,
+      "loss": 1.2094,
+      "step": 6433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2372011428571429e-05,
+      "loss": 1.6584,
+      "step": 6434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2371742857142858e-05,
+      "loss": 1.3062,
+      "step": 6435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2371474285714287e-05,
+      "loss": 1.5196,
+      "step": 6436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2371205714285715e-05,
+      "loss": 1.1769,
+      "step": 6437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2370937142857143e-05,
+      "loss": 1.5589,
+      "step": 6438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2370668571428573e-05,
+      "loss": 1.4118,
+      "step": 6439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.23704e-05,
+      "loss": 1.3054,
+      "step": 6440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.237013142857143e-05,
+      "loss": 1.4172,
+      "step": 6441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2369862857142857e-05,
+      "loss": 1.4493,
+      "step": 6442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2369594285714286e-05,
+      "loss": 1.2705,
+      "step": 6443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2369325714285715e-05,
+      "loss": 1.2498,
+      "step": 6444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2369057142857144e-05,
+      "loss": 1.6674,
+      "step": 6445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2368788571428572e-05,
+      "loss": 1.3255,
+      "step": 6446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.236852e-05,
+      "loss": 1.0961,
+      "step": 6447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.236825142857143e-05,
+      "loss": 1.3364,
+      "step": 6448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2367982857142858e-05,
+      "loss": 1.5071,
+      "step": 6449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2367714285714286e-05,
+      "loss": 1.3318,
+      "step": 6450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2367445714285714e-05,
+      "loss": 1.4859,
+      "step": 6451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2367177142857144e-05,
+      "loss": 1.5667,
+      "step": 6452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.236690857142857e-05,
+      "loss": 1.5642,
+      "step": 6453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.236664e-05,
+      "loss": 1.1652,
+      "step": 6454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2366371428571429e-05,
+      "loss": 1.6301,
+      "step": 6455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2366102857142857e-05,
+      "loss": 1.3088,
+      "step": 6456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2365834285714287e-05,
+      "loss": 1.4979,
+      "step": 6457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2365565714285715e-05,
+      "loss": 1.4321,
+      "step": 6458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2365297142857145e-05,
+      "loss": 1.1025,
+      "step": 6459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2365028571428571e-05,
+      "loss": 1.3804,
+      "step": 6460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2364760000000001e-05,
+      "loss": 1.0734,
+      "step": 6461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.236449142857143e-05,
+      "loss": 1.2236,
+      "step": 6462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2364222857142858e-05,
+      "loss": 0.9927,
+      "step": 6463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2363954285714286e-05,
+      "loss": 1.2029,
+      "step": 6464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2363685714285716e-05,
+      "loss": 1.7744,
+      "step": 6465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2363417142857142e-05,
+      "loss": 1.0235,
+      "step": 6466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2363148571428572e-05,
+      "loss": 0.9233,
+      "step": 6467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.236288e-05,
+      "loss": 1.3937,
+      "step": 6468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2362611428571428e-05,
+      "loss": 1.3657,
+      "step": 6469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2362342857142858e-05,
+      "loss": 1.7088,
+      "step": 6470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2362074285714286e-05,
+      "loss": 1.2229,
+      "step": 6471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2361805714285716e-05,
+      "loss": 1.2912,
+      "step": 6472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2361537142857143e-05,
+      "loss": 1.1328,
+      "step": 6473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2361268571428573e-05,
+      "loss": 1.7027,
+      "step": 6474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2361e-05,
+      "loss": 0.9731,
+      "step": 6475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2360731428571429e-05,
+      "loss": 1.096,
+      "step": 6476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2360462857142857e-05,
+      "loss": 1.4088,
+      "step": 6477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2360194285714287e-05,
+      "loss": 1.4823,
+      "step": 6478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2359925714285713e-05,
+      "loss": 1.6572,
+      "step": 6479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2359657142857143e-05,
+      "loss": 1.5558,
+      "step": 6480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2359388571428571e-05,
+      "loss": 1.582,
+      "step": 6481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2359120000000001e-05,
+      "loss": 1.5959,
+      "step": 6482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2358851428571428e-05,
+      "loss": 1.2382,
+      "step": 6483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2358582857142858e-05,
+      "loss": 1.1534,
+      "step": 6484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2358314285714288e-05,
+      "loss": 1.472,
+      "step": 6485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2358045714285714e-05,
+      "loss": 1.2223,
+      "step": 6486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2357777142857144e-05,
+      "loss": 1.5512,
+      "step": 6487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2357508571428572e-05,
+      "loss": 1.5486,
+      "step": 6488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.235724e-05,
+      "loss": 1.5544,
+      "step": 6489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2356971428571428e-05,
+      "loss": 1.7132,
+      "step": 6490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2356702857142858e-05,
+      "loss": 1.4682,
+      "step": 6491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2356434285714285e-05,
+      "loss": 1.256,
+      "step": 6492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2356165714285715e-05,
+      "loss": 1.2431,
+      "step": 6493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2355897142857143e-05,
+      "loss": 1.3262,
+      "step": 6494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2355628571428573e-05,
+      "loss": 1.3062,
+      "step": 6495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.235536e-05,
+      "loss": 0.8021,
+      "step": 6496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2355091428571429e-05,
+      "loss": 1.4686,
+      "step": 6497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2354822857142859e-05,
+      "loss": 1.2597,
+      "step": 6498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2354554285714285e-05,
+      "loss": 1.3483,
+      "step": 6499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2354285714285715e-05,
+      "loss": 1.309,
+      "step": 6500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2354017142857143e-05,
+      "loss": 1.328,
+      "step": 6501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2353748571428572e-05,
+      "loss": 1.2379,
+      "step": 6502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.235348e-05,
+      "loss": 1.3177,
+      "step": 6503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.235321142857143e-05,
+      "loss": 1.1975,
+      "step": 6504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2352942857142858e-05,
+      "loss": 1.1926,
+      "step": 6505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2352674285714286e-05,
+      "loss": 1.2928,
+      "step": 6506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2352405714285714e-05,
+      "loss": 1.1945,
+      "step": 6507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2352137142857144e-05,
+      "loss": 0.9634,
+      "step": 6508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.235186857142857e-05,
+      "loss": 1.6884,
+      "step": 6509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.23516e-05,
+      "loss": 1.6395,
+      "step": 6510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.235133142857143e-05,
+      "loss": 1.1671,
+      "step": 6511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2351062857142857e-05,
+      "loss": 1.5209,
+      "step": 6512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2350794285714287e-05,
+      "loss": 1.3309,
+      "step": 6513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2350525714285715e-05,
+      "loss": 1.2822,
+      "step": 6514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2350257142857143e-05,
+      "loss": 1.3835,
+      "step": 6515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2349988571428571e-05,
+      "loss": 1.4295,
+      "step": 6516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2349720000000001e-05,
+      "loss": 1.0922,
+      "step": 6517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.234945142857143e-05,
+      "loss": 1.1873,
+      "step": 6518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2349182857142857e-05,
+      "loss": 1.4409,
+      "step": 6519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2348914285714286e-05,
+      "loss": 1.176,
+      "step": 6520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2348645714285715e-05,
+      "loss": 1.6714,
+      "step": 6521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2348377142857142e-05,
+      "loss": 1.5707,
+      "step": 6522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2348108571428572e-05,
+      "loss": 1.3146,
+      "step": 6523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2347840000000002e-05,
+      "loss": 1.2322,
+      "step": 6524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2347571428571428e-05,
+      "loss": 1.2703,
+      "step": 6525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2347302857142858e-05,
+      "loss": 1.1741,
+      "step": 6526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2347034285714286e-05,
+      "loss": 1.3023,
+      "step": 6527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2346765714285716e-05,
+      "loss": 1.1436,
+      "step": 6528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2346497142857143e-05,
+      "loss": 1.6129,
+      "step": 6529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2346228571428572e-05,
+      "loss": 1.1568,
+      "step": 6530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.234596e-05,
+      "loss": 1.222,
+      "step": 6531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2345691428571429e-05,
+      "loss": 1.6373,
+      "step": 6532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2345422857142857e-05,
+      "loss": 1.2795,
+      "step": 6533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2345154285714287e-05,
+      "loss": 1.2165,
+      "step": 6534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2344885714285713e-05,
+      "loss": 1.6799,
+      "step": 6535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2344617142857143e-05,
+      "loss": 1.2286,
+      "step": 6536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2344348571428573e-05,
+      "loss": 1.5987,
+      "step": 6537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.234408e-05,
+      "loss": 1.0641,
+      "step": 6538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.234381142857143e-05,
+      "loss": 1.3687,
+      "step": 6539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2343542857142858e-05,
+      "loss": 1.2958,
+      "step": 6540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2343274285714287e-05,
+      "loss": 1.361,
+      "step": 6541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2343005714285714e-05,
+      "loss": 1.3208,
+      "step": 6542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2342737142857144e-05,
+      "loss": 1.2566,
+      "step": 6543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2342468571428572e-05,
+      "loss": 1.403,
+      "step": 6544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.23422e-05,
+      "loss": 1.6853,
+      "step": 6545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2341931428571428e-05,
+      "loss": 1.553,
+      "step": 6546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2341662857142858e-05,
+      "loss": 1.2485,
+      "step": 6547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2341394285714285e-05,
+      "loss": 1.3821,
+      "step": 6548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2341125714285715e-05,
+      "loss": 1.4134,
+      "step": 6549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2340857142857144e-05,
+      "loss": 1.3621,
+      "step": 6550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2340588571428573e-05,
+      "loss": 1.4298,
+      "step": 6551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.234032e-05,
+      "loss": 1.1185,
+      "step": 6552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2340051428571429e-05,
+      "loss": 0.8356,
+      "step": 6553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2339782857142859e-05,
+      "loss": 1.878,
+      "step": 6554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2339514285714285e-05,
+      "loss": 1.2693,
+      "step": 6555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2339245714285715e-05,
+      "loss": 1.4851,
+      "step": 6556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2338977142857143e-05,
+      "loss": 1.1798,
+      "step": 6557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2338708571428572e-05,
+      "loss": 1.2265,
+      "step": 6558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.233844e-05,
+      "loss": 1.5054,
+      "step": 6559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.233817142857143e-05,
+      "loss": 1.0104,
+      "step": 6560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2337902857142856e-05,
+      "loss": 1.4373,
+      "step": 6561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2337634285714286e-05,
+      "loss": 1.3672,
+      "step": 6562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2337365714285716e-05,
+      "loss": 1.4178,
+      "step": 6563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2337097142857144e-05,
+      "loss": 1.4718,
+      "step": 6564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2336828571428572e-05,
+      "loss": 1.2316,
+      "step": 6565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.233656e-05,
+      "loss": 1.4646,
+      "step": 6566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.233629142857143e-05,
+      "loss": 1.4452,
+      "step": 6567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2336022857142857e-05,
+      "loss": 1.6852,
+      "step": 6568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2335754285714287e-05,
+      "loss": 1.4603,
+      "step": 6569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2335485714285715e-05,
+      "loss": 1.2594,
+      "step": 6570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2335217142857143e-05,
+      "loss": 1.3322,
+      "step": 6571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2334948571428571e-05,
+      "loss": 1.3029,
+      "step": 6572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2334680000000001e-05,
+      "loss": 1.2968,
+      "step": 6573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2334411428571427e-05,
+      "loss": 1.2884,
+      "step": 6574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2334142857142857e-05,
+      "loss": 1.5102,
+      "step": 6575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2333874285714287e-05,
+      "loss": 1.7119,
+      "step": 6576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2333605714285715e-05,
+      "loss": 1.5225,
+      "step": 6577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2333337142857144e-05,
+      "loss": 1.1455,
+      "step": 6578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2333068571428572e-05,
+      "loss": 1.2778,
+      "step": 6579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2332800000000002e-05,
+      "loss": 1.3938,
+      "step": 6580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2332531428571428e-05,
+      "loss": 1.2679,
+      "step": 6581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2332262857142858e-05,
+      "loss": 1.5383,
+      "step": 6582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2331994285714286e-05,
+      "loss": 1.3826,
+      "step": 6583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2331725714285714e-05,
+      "loss": 1.1786,
+      "step": 6584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2331457142857142e-05,
+      "loss": 1.5807,
+      "step": 6585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2331188571428572e-05,
+      "loss": 1.5485,
+      "step": 6586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.233092e-05,
+      "loss": 1.3089,
+      "step": 6587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2330651428571429e-05,
+      "loss": 1.517,
+      "step": 6588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2330382857142859e-05,
+      "loss": 1.2478,
+      "step": 6589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2330114285714287e-05,
+      "loss": 1.2887,
+      "step": 6590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2329845714285715e-05,
+      "loss": 1.321,
+      "step": 6591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2329577142857143e-05,
+      "loss": 1.4948,
+      "step": 6592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2329308571428573e-05,
+      "loss": 1.1225,
+      "step": 6593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232904e-05,
+      "loss": 1.3233,
+      "step": 6594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232877142857143e-05,
+      "loss": 1.3738,
+      "step": 6595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2328502857142857e-05,
+      "loss": 1.2288,
+      "step": 6596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2328234285714286e-05,
+      "loss": 1.017,
+      "step": 6597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2327965714285714e-05,
+      "loss": 1.492,
+      "step": 6598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2327697142857144e-05,
+      "loss": 1.1753,
+      "step": 6599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2327428571428572e-05,
+      "loss": 1.5825,
+      "step": 6600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232716e-05,
+      "loss": 1.4303,
+      "step": 6601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2326891428571428e-05,
+      "loss": 1.1886,
+      "step": 6602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2326622857142858e-05,
+      "loss": 1.5732,
+      "step": 6603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2326354285714286e-05,
+      "loss": 0.9388,
+      "step": 6604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2326085714285714e-05,
+      "loss": 1.4717,
+      "step": 6605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2325817142857144e-05,
+      "loss": 1.2429,
+      "step": 6606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232554857142857e-05,
+      "loss": 1.1251,
+      "step": 6607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232528e-05,
+      "loss": 1.4347,
+      "step": 6608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2325011428571429e-05,
+      "loss": 0.9944,
+      "step": 6609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2324742857142859e-05,
+      "loss": 1.3501,
+      "step": 6610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2324474285714285e-05,
+      "loss": 1.1604,
+      "step": 6611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2324205714285715e-05,
+      "loss": 1.2053,
+      "step": 6612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2323937142857143e-05,
+      "loss": 1.3633,
+      "step": 6613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2323668571428571e-05,
+      "loss": 1.1049,
+      "step": 6614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.23234e-05,
+      "loss": 1.5895,
+      "step": 6615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232313142857143e-05,
+      "loss": 1.5687,
+      "step": 6616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2322862857142858e-05,
+      "loss": 1.128,
+      "step": 6617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2322594285714286e-05,
+      "loss": 1.5716,
+      "step": 6618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2322325714285716e-05,
+      "loss": 1.2883,
+      "step": 6619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2322057142857142e-05,
+      "loss": 1.297,
+      "step": 6620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2321788571428572e-05,
+      "loss": 1.5044,
+      "step": 6621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232152e-05,
+      "loss": 1.0313,
+      "step": 6622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.232125142857143e-05,
+      "loss": 1.3969,
+      "step": 6623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2320982857142857e-05,
+      "loss": 1.6102,
+      "step": 6624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2320714285714286e-05,
+      "loss": 1.2307,
+      "step": 6625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2320445714285715e-05,
+      "loss": 1.524,
+      "step": 6626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2320177142857143e-05,
+      "loss": 1.1779,
+      "step": 6627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2319908571428571e-05,
+      "loss": 1.1148,
+      "step": 6628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2319640000000001e-05,
+      "loss": 1.085,
+      "step": 6629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2319371428571429e-05,
+      "loss": 1.5595,
+      "step": 6630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2319102857142857e-05,
+      "loss": 1.4048,
+      "step": 6631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2318834285714287e-05,
+      "loss": 1.3787,
+      "step": 6632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2318565714285715e-05,
+      "loss": 1.0976,
+      "step": 6633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2318297142857143e-05,
+      "loss": 1.2795,
+      "step": 6634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2318028571428572e-05,
+      "loss": 1.7567,
+      "step": 6635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2317760000000001e-05,
+      "loss": 1.6657,
+      "step": 6636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2317491428571428e-05,
+      "loss": 1.8819,
+      "step": 6637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2317222857142858e-05,
+      "loss": 1.481,
+      "step": 6638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2316954285714286e-05,
+      "loss": 1.3302,
+      "step": 6639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2316685714285714e-05,
+      "loss": 1.4047,
+      "step": 6640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2316417142857142e-05,
+      "loss": 1.3396,
+      "step": 6641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2316148571428572e-05,
+      "loss": 1.6387,
+      "step": 6642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.231588e-05,
+      "loss": 1.4295,
+      "step": 6643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2315611428571429e-05,
+      "loss": 1.3505,
+      "step": 6644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2315342857142858e-05,
+      "loss": 1.445,
+      "step": 6645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2315074285714287e-05,
+      "loss": 1.261,
+      "step": 6646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2314805714285715e-05,
+      "loss": 1.5859,
+      "step": 6647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2314537142857143e-05,
+      "loss": 1.2673,
+      "step": 6648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2314268571428573e-05,
+      "loss": 1.2838,
+      "step": 6649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2314e-05,
+      "loss": 1.0321,
+      "step": 6650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.231373142857143e-05,
+      "loss": 0.9925,
+      "step": 6651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2313462857142857e-05,
+      "loss": 1.2309,
+      "step": 6652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2313194285714286e-05,
+      "loss": 1.4059,
+      "step": 6653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2312925714285714e-05,
+      "loss": 1.3403,
+      "step": 6654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2312657142857144e-05,
+      "loss": 1.297,
+      "step": 6655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2312388571428573e-05,
+      "loss": 1.5121,
+      "step": 6656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.231212e-05,
+      "loss": 1.2315,
+      "step": 6657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.231185142857143e-05,
+      "loss": 1.6593,
+      "step": 6658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2311582857142858e-05,
+      "loss": 1.3797,
+      "step": 6659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2311314285714286e-05,
+      "loss": 1.1143,
+      "step": 6660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2311045714285714e-05,
+      "loss": 1.1451,
+      "step": 6661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2310777142857144e-05,
+      "loss": 1.5051,
+      "step": 6662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.231050857142857e-05,
+      "loss": 1.2806,
+      "step": 6663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.231024e-05,
+      "loss": 1.465,
+      "step": 6664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2309971428571429e-05,
+      "loss": 1.2811,
+      "step": 6665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2309702857142857e-05,
+      "loss": 1.2367,
+      "step": 6666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2309434285714285e-05,
+      "loss": 1.8352,
+      "step": 6667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2309165714285715e-05,
+      "loss": 1.4858,
+      "step": 6668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2308897142857145e-05,
+      "loss": 1.4517,
+      "step": 6669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2308628571428571e-05,
+      "loss": 1.82,
+      "step": 6670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2308360000000001e-05,
+      "loss": 1.4231,
+      "step": 6671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.230809142857143e-05,
+      "loss": 1.5688,
+      "step": 6672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2307822857142858e-05,
+      "loss": 1.6162,
+      "step": 6673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2307554285714286e-05,
+      "loss": 1.5401,
+      "step": 6674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2307285714285716e-05,
+      "loss": 1.4454,
+      "step": 6675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2307017142857142e-05,
+      "loss": 1.3979,
+      "step": 6676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2306748571428572e-05,
+      "loss": 1.123,
+      "step": 6677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.230648e-05,
+      "loss": 1.2543,
+      "step": 6678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.230621142857143e-05,
+      "loss": 1.0845,
+      "step": 6679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2305942857142856e-05,
+      "loss": 1.4809,
+      "step": 6680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2305674285714286e-05,
+      "loss": 1.7127,
+      "step": 6681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2305405714285716e-05,
+      "loss": 1.3386,
+      "step": 6682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2305137142857143e-05,
+      "loss": 1.5736,
+      "step": 6683
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2304868571428573e-05,
+      "loss": 1.4516,
+      "step": 6684
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.23046e-05,
+      "loss": 2.1856,
+      "step": 6685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2304331428571429e-05,
+      "loss": 1.6304,
+      "step": 6686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2304062857142857e-05,
+      "loss": 1.2512,
+      "step": 6687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2303794285714287e-05,
+      "loss": 0.9937,
+      "step": 6688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2303525714285713e-05,
+      "loss": 1.7573,
+      "step": 6689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2303257142857143e-05,
+      "loss": 1.3275,
+      "step": 6690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2302988571428571e-05,
+      "loss": 1.0784,
+      "step": 6691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2302720000000001e-05,
+      "loss": 1.753,
+      "step": 6692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2302451428571428e-05,
+      "loss": 1.2138,
+      "step": 6693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2302182857142858e-05,
+      "loss": 1.1985,
+      "step": 6694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2301914285714288e-05,
+      "loss": 1.3098,
+      "step": 6695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2301645714285714e-05,
+      "loss": 1.2463,
+      "step": 6696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2301377142857144e-05,
+      "loss": 1.3811,
+      "step": 6697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2301108571428572e-05,
+      "loss": 1.2177,
+      "step": 6698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.230084e-05,
+      "loss": 1.8278,
+      "step": 6699
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2300571428571428e-05,
+      "loss": 1.0701,
+      "step": 6700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2300302857142858e-05,
+      "loss": 1.2628,
+      "step": 6701
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2300034285714287e-05,
+      "loss": 1.0744,
+      "step": 6702
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2299765714285715e-05,
+      "loss": 1.1798,
+      "step": 6703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2299497142857143e-05,
+      "loss": 1.3888,
+      "step": 6704
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2299228571428573e-05,
+      "loss": 1.6357,
+      "step": 6705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.229896e-05,
+      "loss": 1.308,
+      "step": 6706
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2298691428571429e-05,
+      "loss": 1.177,
+      "step": 6707
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2298422857142857e-05,
+      "loss": 1.319,
+      "step": 6708
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2298154285714285e-05,
+      "loss": 1.7598,
+      "step": 6709
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2297885714285715e-05,
+      "loss": 1.4957,
+      "step": 6710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2297617142857143e-05,
+      "loss": 1.116,
+      "step": 6711
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2297348571428572e-05,
+      "loss": 1.3471,
+      "step": 6712
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.229708e-05,
+      "loss": 1.2056,
+      "step": 6713
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.229681142857143e-05,
+      "loss": 1.7354,
+      "step": 6714
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2296542857142858e-05,
+      "loss": 1.2704,
+      "step": 6715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2296274285714286e-05,
+      "loss": 1.1315,
+      "step": 6716
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2296005714285714e-05,
+      "loss": 1.6883,
+      "step": 6717
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2295737142857144e-05,
+      "loss": 1.0811,
+      "step": 6718
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.229546857142857e-05,
+      "loss": 1.4016,
+      "step": 6719
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.22952e-05,
+      "loss": 2.033,
+      "step": 6720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2294931428571429e-05,
+      "loss": 1.569,
+      "step": 6721
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2294662857142857e-05,
+      "loss": 1.0996,
+      "step": 6722
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2294394285714287e-05,
+      "loss": 1.7271,
+      "step": 6723
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2294125714285715e-05,
+      "loss": 1.2773,
+      "step": 6724
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2293857142857145e-05,
+      "loss": 1.349,
+      "step": 6725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2293588571428571e-05,
+      "loss": 1.6649,
+      "step": 6726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2293320000000001e-05,
+      "loss": 1.2402,
+      "step": 6727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.229305142857143e-05,
+      "loss": 1.2821,
+      "step": 6728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2292782857142857e-05,
+      "loss": 1.2233,
+      "step": 6729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2292514285714286e-05,
+      "loss": 1.507,
+      "step": 6730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2292245714285715e-05,
+      "loss": 1.4746,
+      "step": 6731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2291977142857142e-05,
+      "loss": 1.6517,
+      "step": 6732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2291708571428572e-05,
+      "loss": 1.3331,
+      "step": 6733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.229144e-05,
+      "loss": 1.2849,
+      "step": 6734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2291171428571428e-05,
+      "loss": 1.4576,
+      "step": 6735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2290902857142858e-05,
+      "loss": 1.3486,
+      "step": 6736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2290634285714286e-05,
+      "loss": 0.9924,
+      "step": 6737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2290365714285716e-05,
+      "loss": 1.6322,
+      "step": 6738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2290097142857143e-05,
+      "loss": 1.2929,
+      "step": 6739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2289828571428572e-05,
+      "loss": 1.44,
+      "step": 6740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.228956e-05,
+      "loss": 1.4102,
+      "step": 6741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2289291428571429e-05,
+      "loss": 1.4359,
+      "step": 6742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2289022857142857e-05,
+      "loss": 1.3512,
+      "step": 6743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2288754285714287e-05,
+      "loss": 1.4002,
+      "step": 6744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2288485714285713e-05,
+      "loss": 1.145,
+      "step": 6745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2288217142857143e-05,
+      "loss": 1.3206,
+      "step": 6746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2287948571428571e-05,
+      "loss": 0.9,
+      "step": 6747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.228768e-05,
+      "loss": 1.4366,
+      "step": 6748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.228741142857143e-05,
+      "loss": 1.0389,
+      "step": 6749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2287142857142858e-05,
+      "loss": 1.3136,
+      "step": 6750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2286874285714287e-05,
+      "loss": 1.5385,
+      "step": 6751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2286605714285714e-05,
+      "loss": 1.4096,
+      "step": 6752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2286337142857144e-05,
+      "loss": 1.6401,
+      "step": 6753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2286068571428572e-05,
+      "loss": 1.2059,
+      "step": 6754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.22858e-05,
+      "loss": 1.2998,
+      "step": 6755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2285531428571428e-05,
+      "loss": 1.4606,
+      "step": 6756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2285262857142858e-05,
+      "loss": 1.3807,
+      "step": 6757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2284994285714285e-05,
+      "loss": 1.3758,
+      "step": 6758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2284725714285715e-05,
+      "loss": 1.3757,
+      "step": 6759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2284457142857143e-05,
+      "loss": 1.131,
+      "step": 6760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2284188571428573e-05,
+      "loss": 1.2542,
+      "step": 6761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.228392e-05,
+      "loss": 1.4602,
+      "step": 6762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2283651428571429e-05,
+      "loss": 1.4834,
+      "step": 6763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2283382857142859e-05,
+      "loss": 1.5198,
+      "step": 6764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2283114285714285e-05,
+      "loss": 1.287,
+      "step": 6765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2282845714285715e-05,
+      "loss": 1.3938,
+      "step": 6766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2282577142857143e-05,
+      "loss": 1.4323,
+      "step": 6767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2282308571428572e-05,
+      "loss": 1.3464,
+      "step": 6768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.228204e-05,
+      "loss": 1.6412,
+      "step": 6769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.228177142857143e-05,
+      "loss": 1.0981,
+      "step": 6770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2281502857142856e-05,
+      "loss": 1.2367,
+      "step": 6771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2281234285714286e-05,
+      "loss": 1.6985,
+      "step": 6772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2280965714285714e-05,
+      "loss": 1.5731,
+      "step": 6773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2280697142857144e-05,
+      "loss": 1.2224,
+      "step": 6774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2280428571428572e-05,
+      "loss": 1.5716,
+      "step": 6775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.228016e-05,
+      "loss": 1.2113,
+      "step": 6776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.227989142857143e-05,
+      "loss": 1.6953,
+      "step": 6777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2279622857142857e-05,
+      "loss": 1.614,
+      "step": 6778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2279354285714287e-05,
+      "loss": 1.3058,
+      "step": 6779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2279085714285715e-05,
+      "loss": 1.166,
+      "step": 6780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2278817142857143e-05,
+      "loss": 1.1698,
+      "step": 6781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2278548571428571e-05,
+      "loss": 1.5422,
+      "step": 6782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2278280000000001e-05,
+      "loss": 0.9124,
+      "step": 6783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2278011428571429e-05,
+      "loss": 1.4117,
+      "step": 6784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2277742857142857e-05,
+      "loss": 1.5712,
+      "step": 6785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2277474285714285e-05,
+      "loss": 0.8358,
+      "step": 6786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2277205714285715e-05,
+      "loss": 1.0799,
+      "step": 6787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2276937142857144e-05,
+      "loss": 1.5906,
+      "step": 6788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2276668571428572e-05,
+      "loss": 1.3266,
+      "step": 6789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2276400000000002e-05,
+      "loss": 1.7284,
+      "step": 6790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2276131428571428e-05,
+      "loss": 1.1242,
+      "step": 6791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2275862857142858e-05,
+      "loss": 1.2594,
+      "step": 6792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2275594285714286e-05,
+      "loss": 1.0983,
+      "step": 6793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2275325714285714e-05,
+      "loss": 1.5098,
+      "step": 6794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2275057142857142e-05,
+      "loss": 1.4397,
+      "step": 6795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2274788571428572e-05,
+      "loss": 1.1988,
+      "step": 6796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.227452e-05,
+      "loss": 1.4789,
+      "step": 6797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2274251428571429e-05,
+      "loss": 1.07,
+      "step": 6798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2273982857142857e-05,
+      "loss": 1.7093,
+      "step": 6799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2273714285714287e-05,
+      "loss": 1.7033,
+      "step": 6800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2273445714285715e-05,
+      "loss": 1.6353,
+      "step": 6801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2273177142857143e-05,
+      "loss": 1.5066,
+      "step": 6802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2272908571428573e-05,
+      "loss": 1.5536,
+      "step": 6803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.227264e-05,
+      "loss": 1.3834,
+      "step": 6804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.227237142857143e-05,
+      "loss": 1.3792,
+      "step": 6805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2272102857142857e-05,
+      "loss": 1.5629,
+      "step": 6806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2271834285714287e-05,
+      "loss": 1.1419,
+      "step": 6807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2271565714285714e-05,
+      "loss": 1.4582,
+      "step": 6808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2271297142857144e-05,
+      "loss": 1.123,
+      "step": 6809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2271028571428572e-05,
+      "loss": 1.3751,
+      "step": 6810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.227076e-05,
+      "loss": 1.3174,
+      "step": 6811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2270491428571428e-05,
+      "loss": 1.6954,
+      "step": 6812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2270222857142858e-05,
+      "loss": 0.9967,
+      "step": 6813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2269954285714286e-05,
+      "loss": 1.2914,
+      "step": 6814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2269685714285714e-05,
+      "loss": 1.5874,
+      "step": 6815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2269417142857144e-05,
+      "loss": 1.1614,
+      "step": 6816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.226914857142857e-05,
+      "loss": 1.7018,
+      "step": 6817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.226888e-05,
+      "loss": 1.4253,
+      "step": 6818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2268611428571429e-05,
+      "loss": 1.6058,
+      "step": 6819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2268342857142859e-05,
+      "loss": 1.3166,
+      "step": 6820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2268074285714285e-05,
+      "loss": 0.9842,
+      "step": 6821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2267805714285715e-05,
+      "loss": 1.4373,
+      "step": 6822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2267537142857143e-05,
+      "loss": 1.4869,
+      "step": 6823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2267268571428571e-05,
+      "loss": 1.4439,
+      "step": 6824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2267e-05,
+      "loss": 1.1486,
+      "step": 6825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.226673142857143e-05,
+      "loss": 1.6147,
+      "step": 6826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2266462857142858e-05,
+      "loss": 1.323,
+      "step": 6827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2266194285714286e-05,
+      "loss": 1.1507,
+      "step": 6828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2265925714285716e-05,
+      "loss": 1.1865,
+      "step": 6829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2265657142857144e-05,
+      "loss": 1.07,
+      "step": 6830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2265388571428572e-05,
+      "loss": 1.4927,
+      "step": 6831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.226512e-05,
+      "loss": 1.3841,
+      "step": 6832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.226485142857143e-05,
+      "loss": 1.429,
+      "step": 6833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2264582857142857e-05,
+      "loss": 1.5979,
+      "step": 6834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2264314285714286e-05,
+      "loss": 1.3301,
+      "step": 6835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2264045714285715e-05,
+      "loss": 1.1669,
+      "step": 6836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2263777142857143e-05,
+      "loss": 1.4822,
+      "step": 6837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2263508571428571e-05,
+      "loss": 1.9571,
+      "step": 6838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2263240000000001e-05,
+      "loss": 1.5452,
+      "step": 6839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2262971428571429e-05,
+      "loss": 1.5154,
+      "step": 6840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2262702857142857e-05,
+      "loss": 1.3057,
+      "step": 6841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2262434285714287e-05,
+      "loss": 1.6051,
+      "step": 6842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2262165714285715e-05,
+      "loss": 1.4563,
+      "step": 6843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2261897142857143e-05,
+      "loss": 1.758,
+      "step": 6844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2261628571428572e-05,
+      "loss": 1.2475,
+      "step": 6845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2261360000000001e-05,
+      "loss": 1.4171,
+      "step": 6846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2261091428571428e-05,
+      "loss": 1.6529,
+      "step": 6847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2260822857142858e-05,
+      "loss": 1.3248,
+      "step": 6848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2260554285714286e-05,
+      "loss": 1.6465,
+      "step": 6849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2260285714285714e-05,
+      "loss": 1.2937,
+      "step": 6850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2260017142857142e-05,
+      "loss": 1.7299,
+      "step": 6851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2259748571428572e-05,
+      "loss": 1.6263,
+      "step": 6852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225948e-05,
+      "loss": 1.2496,
+      "step": 6853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2259211428571429e-05,
+      "loss": 1.3155,
+      "step": 6854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2258942857142858e-05,
+      "loss": 1.7111,
+      "step": 6855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2258674285714287e-05,
+      "loss": 1.7467,
+      "step": 6856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2258405714285715e-05,
+      "loss": 1.0638,
+      "step": 6857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2258137142857143e-05,
+      "loss": 1.6997,
+      "step": 6858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2257868571428573e-05,
+      "loss": 1.2833,
+      "step": 6859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.22576e-05,
+      "loss": 1.2416,
+      "step": 6860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225733142857143e-05,
+      "loss": 1.1682,
+      "step": 6861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2257062857142857e-05,
+      "loss": 1.6843,
+      "step": 6862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2256794285714286e-05,
+      "loss": 1.2725,
+      "step": 6863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2256525714285714e-05,
+      "loss": 1.3407,
+      "step": 6864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2256257142857144e-05,
+      "loss": 1.2962,
+      "step": 6865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2255988571428572e-05,
+      "loss": 1.5034,
+      "step": 6866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225572e-05,
+      "loss": 1.1538,
+      "step": 6867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225545142857143e-05,
+      "loss": 1.1986,
+      "step": 6868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2255182857142858e-05,
+      "loss": 1.2971,
+      "step": 6869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2254914285714286e-05,
+      "loss": 1.5588,
+      "step": 6870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2254645714285714e-05,
+      "loss": 1.3945,
+      "step": 6871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2254377142857144e-05,
+      "loss": 1.3968,
+      "step": 6872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225410857142857e-05,
+      "loss": 1.0924,
+      "step": 6873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225384e-05,
+      "loss": 0.9867,
+      "step": 6874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2253571428571429e-05,
+      "loss": 1.1431,
+      "step": 6875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2253302857142859e-05,
+      "loss": 1.458,
+      "step": 6876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2253034285714285e-05,
+      "loss": 1.1132,
+      "step": 6877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2252765714285715e-05,
+      "loss": 1.3235,
+      "step": 6878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2252497142857143e-05,
+      "loss": 1.1384,
+      "step": 6879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2252228571428571e-05,
+      "loss": 1.3154,
+      "step": 6880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2251960000000001e-05,
+      "loss": 1.447,
+      "step": 6881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225169142857143e-05,
+      "loss": 1.2952,
+      "step": 6882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2251422857142858e-05,
+      "loss": 1.3551,
+      "step": 6883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2251154285714286e-05,
+      "loss": 1.9301,
+      "step": 6884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2250885714285716e-05,
+      "loss": 1.2985,
+      "step": 6885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2250617142857142e-05,
+      "loss": 1.1524,
+      "step": 6886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2250348571428572e-05,
+      "loss": 1.274,
+      "step": 6887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.225008e-05,
+      "loss": 1.6698,
+      "step": 6888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.224981142857143e-05,
+      "loss": 1.3602,
+      "step": 6889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2249542857142856e-05,
+      "loss": 1.4082,
+      "step": 6890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2249274285714286e-05,
+      "loss": 1.3161,
+      "step": 6891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2249005714285715e-05,
+      "loss": 1.309,
+      "step": 6892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2248737142857143e-05,
+      "loss": 1.4225,
+      "step": 6893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2248468571428573e-05,
+      "loss": 1.3899,
+      "step": 6894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.22482e-05,
+      "loss": 1.2209,
+      "step": 6895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2247931428571429e-05,
+      "loss": 1.3421,
+      "step": 6896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2247662857142857e-05,
+      "loss": 1.4914,
+      "step": 6897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2247394285714287e-05,
+      "loss": 1.3089,
+      "step": 6898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2247125714285713e-05,
+      "loss": 1.0093,
+      "step": 6899
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2246857142857143e-05,
+      "loss": 1.3981,
+      "step": 6900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2246588571428571e-05,
+      "loss": 1.5015,
+      "step": 6901
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2246320000000001e-05,
+      "loss": 1.3374,
+      "step": 6902
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2246051428571428e-05,
+      "loss": 1.5486,
+      "step": 6903
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2245782857142858e-05,
+      "loss": 1.3469,
+      "step": 6904
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2245514285714286e-05,
+      "loss": 1.8303,
+      "step": 6905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2245245714285714e-05,
+      "loss": 1.1973,
+      "step": 6906
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2244977142857144e-05,
+      "loss": 1.1338,
+      "step": 6907
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2244708571428572e-05,
+      "loss": 1.3469,
+      "step": 6908
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.224444e-05,
+      "loss": 1.2034,
+      "step": 6909
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2244171428571428e-05,
+      "loss": 1.0202,
+      "step": 6910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2243902857142858e-05,
+      "loss": 1.3926,
+      "step": 6911
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.2243634285714287e-05,
+      "loss": 1.3639,
+      "step": 6912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2243365714285715e-05,
+      "loss": 1.492,
+      "step": 6913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2243097142857143e-05,
+      "loss": 1.3238,
+      "step": 6914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2242828571428573e-05,
+      "loss": 1.2813,
+      "step": 6915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.224256e-05,
+      "loss": 1.3506,
+      "step": 6916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2242291428571429e-05,
+      "loss": 1.5156,
+      "step": 6917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2242022857142857e-05,
+      "loss": 1.2471,
+      "step": 6918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2241754285714285e-05,
+      "loss": 1.7522,
+      "step": 6919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2241485714285715e-05,
+      "loss": 1.6668,
+      "step": 6920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2241217142857143e-05,
+      "loss": 1.7928,
+      "step": 6921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2240948571428572e-05,
+      "loss": 1.3265,
+      "step": 6922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.224068e-05,
+      "loss": 1.4857,
+      "step": 6923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.224041142857143e-05,
+      "loss": 1.4394,
+      "step": 6924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2240142857142858e-05,
+      "loss": 1.2528,
+      "step": 6925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2239874285714286e-05,
+      "loss": 1.5697,
+      "step": 6926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2239605714285714e-05,
+      "loss": 1.3242,
+      "step": 6927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2239337142857144e-05,
+      "loss": 1.1927,
+      "step": 6928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.223906857142857e-05,
+      "loss": 1.3023,
+      "step": 6929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.22388e-05,
+      "loss": 1.3905,
+      "step": 6930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2238531428571429e-05,
+      "loss": 1.5181,
+      "step": 6931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2238262857142857e-05,
+      "loss": 1.2587,
+      "step": 6932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2237994285714287e-05,
+      "loss": 1.3997,
+      "step": 6933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2237725714285715e-05,
+      "loss": 1.4438,
+      "step": 6934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2237457142857145e-05,
+      "loss": 1.1235,
+      "step": 6935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2237188571428571e-05,
+      "loss": 1.6345,
+      "step": 6936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2236920000000001e-05,
+      "loss": 1.1511,
+      "step": 6937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.223665142857143e-05,
+      "loss": 1.4942,
+      "step": 6938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2236382857142857e-05,
+      "loss": 1.2001,
+      "step": 6939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2236114285714286e-05,
+      "loss": 1.644,
+      "step": 6940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2235845714285715e-05,
+      "loss": 1.5942,
+      "step": 6941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2235577142857142e-05,
+      "loss": 1.4154,
+      "step": 6942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2235308571428572e-05,
+      "loss": 1.1311,
+      "step": 6943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.223504e-05,
+      "loss": 1.4637,
+      "step": 6944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2234771428571428e-05,
+      "loss": 1.2824,
+      "step": 6945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2234502857142858e-05,
+      "loss": 1.4251,
+      "step": 6946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2234234285714286e-05,
+      "loss": 1.1302,
+      "step": 6947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2233965714285716e-05,
+      "loss": 1.2462,
+      "step": 6948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2233697142857143e-05,
+      "loss": 1.2989,
+      "step": 6949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2233428571428572e-05,
+      "loss": 1.5223,
+      "step": 6950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.223316e-05,
+      "loss": 1.2991,
+      "step": 6951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2232891428571429e-05,
+      "loss": 1.0518,
+      "step": 6952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2232622857142857e-05,
+      "loss": 1.4971,
+      "step": 6953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2232354285714287e-05,
+      "loss": 1.6661,
+      "step": 6954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2232085714285713e-05,
+      "loss": 1.176,
+      "step": 6955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2231817142857143e-05,
+      "loss": 1.5532,
+      "step": 6956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2231548571428571e-05,
+      "loss": 1.4857,
+      "step": 6957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2231280000000001e-05,
+      "loss": 1.7431,
+      "step": 6958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.223101142857143e-05,
+      "loss": 1.6304,
+      "step": 6959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2230742857142858e-05,
+      "loss": 1.4795,
+      "step": 6960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2230474285714287e-05,
+      "loss": 1.0573,
+      "step": 6961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2230205714285714e-05,
+      "loss": 1.3391,
+      "step": 6962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2229937142857144e-05,
+      "loss": 1.1916,
+      "step": 6963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2229668571428572e-05,
+      "loss": 1.1633,
+      "step": 6964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.22294e-05,
+      "loss": 1.7419,
+      "step": 6965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2229131428571428e-05,
+      "loss": 1.6594,
+      "step": 6966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2228862857142858e-05,
+      "loss": 1.2344,
+      "step": 6967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2228594285714285e-05,
+      "loss": 1.5401,
+      "step": 6968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2228325714285715e-05,
+      "loss": 1.2479,
+      "step": 6969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2228057142857143e-05,
+      "loss": 1.3247,
+      "step": 6970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2227788571428573e-05,
+      "loss": 1.2539,
+      "step": 6971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.222752e-05,
+      "loss": 1.1483,
+      "step": 6972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2227251428571429e-05,
+      "loss": 1.2382,
+      "step": 6973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2226982857142859e-05,
+      "loss": 1.392,
+      "step": 6974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2226714285714285e-05,
+      "loss": 0.8589,
+      "step": 6975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2226445714285715e-05,
+      "loss": 1.5096,
+      "step": 6976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2226177142857143e-05,
+      "loss": 1.395,
+      "step": 6977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2225908571428572e-05,
+      "loss": 1.0196,
+      "step": 6978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.222564e-05,
+      "loss": 1.2836,
+      "step": 6979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.222537142857143e-05,
+      "loss": 1.3499,
+      "step": 6980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2225102857142858e-05,
+      "loss": 1.3043,
+      "step": 6981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2224834285714286e-05,
+      "loss": 1.0468,
+      "step": 6982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2224565714285714e-05,
+      "loss": 1.4407,
+      "step": 6983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2224297142857144e-05,
+      "loss": 1.0756,
+      "step": 6984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2224028571428572e-05,
+      "loss": 1.2352,
+      "step": 6985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.222376e-05,
+      "loss": 1.586,
+      "step": 6986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.222349142857143e-05,
+      "loss": 1.2642,
+      "step": 6987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2223222857142857e-05,
+      "loss": 1.7148,
+      "step": 6988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2222954285714287e-05,
+      "loss": 1.5905,
+      "step": 6989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2222685714285715e-05,
+      "loss": 1.2772,
+      "step": 6990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2222417142857143e-05,
+      "loss": 1.1989,
+      "step": 6991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2222148571428571e-05,
+      "loss": 1.101,
+      "step": 6992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2221880000000001e-05,
+      "loss": 1.1093,
+      "step": 6993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2221611428571429e-05,
+      "loss": 1.2444,
+      "step": 6994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2221342857142857e-05,
+      "loss": 1.1904,
+      "step": 6995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2221074285714285e-05,
+      "loss": 1.3557,
+      "step": 6996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2220805714285715e-05,
+      "loss": 1.2358,
+      "step": 6997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2220537142857144e-05,
+      "loss": 1.9097,
+      "step": 6998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2220268571428572e-05,
+      "loss": 1.1524,
+      "step": 6999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2220000000000002e-05,
+      "loss": 1.1137,
+      "step": 7000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2219731428571428e-05,
+      "loss": 1.3113,
+      "step": 7001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2219462857142858e-05,
+      "loss": 1.2502,
+      "step": 7002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2219194285714286e-05,
+      "loss": 1.2291,
+      "step": 7003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2218925714285716e-05,
+      "loss": 1.4535,
+      "step": 7004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2218657142857142e-05,
+      "loss": 1.2384,
+      "step": 7005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2218388571428572e-05,
+      "loss": 1.0735,
+      "step": 7006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.221812e-05,
+      "loss": 1.4933,
+      "step": 7007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2217851428571429e-05,
+      "loss": 1.4322,
+      "step": 7008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2217582857142857e-05,
+      "loss": 1.6724,
+      "step": 7009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2217314285714287e-05,
+      "loss": 1.4205,
+      "step": 7010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2217045714285715e-05,
+      "loss": 1.3772,
+      "step": 7011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2216777142857143e-05,
+      "loss": 1.3794,
+      "step": 7012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2216508571428573e-05,
+      "loss": 1.1303,
+      "step": 7013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.221624e-05,
+      "loss": 1.1872,
+      "step": 7014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.221597142857143e-05,
+      "loss": 1.5522,
+      "step": 7015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2215702857142857e-05,
+      "loss": 1.0239,
+      "step": 7016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2215434285714287e-05,
+      "loss": 1.3674,
+      "step": 7017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2215165714285714e-05,
+      "loss": 1.4564,
+      "step": 7018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2214897142857144e-05,
+      "loss": 1.658,
+      "step": 7019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2214628571428572e-05,
+      "loss": 1.0339,
+      "step": 7020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.221436e-05,
+      "loss": 1.4256,
+      "step": 7021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2214091428571428e-05,
+      "loss": 1.4295,
+      "step": 7022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2213822857142858e-05,
+      "loss": 1.5148,
+      "step": 7023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2213554285714286e-05,
+      "loss": 1.4358,
+      "step": 7024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2213285714285714e-05,
+      "loss": 0.9887,
+      "step": 7025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2213017142857144e-05,
+      "loss": 1.4207,
+      "step": 7026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2212748571428573e-05,
+      "loss": 1.3754,
+      "step": 7027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.221248e-05,
+      "loss": 1.7328,
+      "step": 7028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2212211428571429e-05,
+      "loss": 1.1861,
+      "step": 7029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2211942857142859e-05,
+      "loss": 0.788,
+      "step": 7030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2211674285714285e-05,
+      "loss": 1.1464,
+      "step": 7031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2211405714285715e-05,
+      "loss": 1.418,
+      "step": 7032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2211137142857143e-05,
+      "loss": 1.7806,
+      "step": 7033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2210868571428571e-05,
+      "loss": 1.1137,
+      "step": 7034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.22106e-05,
+      "loss": 1.4111,
+      "step": 7035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.221033142857143e-05,
+      "loss": 1.4546,
+      "step": 7036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2210062857142858e-05,
+      "loss": 1.3116,
+      "step": 7037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2209794285714286e-05,
+      "loss": 1.0044,
+      "step": 7038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2209525714285716e-05,
+      "loss": 1.0247,
+      "step": 7039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2209257142857144e-05,
+      "loss": 1.1389,
+      "step": 7040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2208988571428572e-05,
+      "loss": 1.2968,
+      "step": 7041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.220872e-05,
+      "loss": 1.5232,
+      "step": 7042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.220845142857143e-05,
+      "loss": 1.4029,
+      "step": 7043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2208182857142857e-05,
+      "loss": 1.2187,
+      "step": 7044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2207914285714286e-05,
+      "loss": 1.2115,
+      "step": 7045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2207645714285715e-05,
+      "loss": 1.4513,
+      "step": 7046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2207377142857143e-05,
+      "loss": 1.4553,
+      "step": 7047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2207108571428571e-05,
+      "loss": 1.2877,
+      "step": 7048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2206840000000001e-05,
+      "loss": 1.5818,
+      "step": 7049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2206571428571429e-05,
+      "loss": 1.4505,
+      "step": 7050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2206302857142857e-05,
+      "loss": 1.5266,
+      "step": 7051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2206034285714287e-05,
+      "loss": 2.2196,
+      "step": 7052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2205765714285715e-05,
+      "loss": 1.303,
+      "step": 7053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2205497142857143e-05,
+      "loss": 1.3015,
+      "step": 7054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2205228571428572e-05,
+      "loss": 1.1033,
+      "step": 7055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2204960000000001e-05,
+      "loss": 1.3955,
+      "step": 7056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2204691428571428e-05,
+      "loss": 1.4803,
+      "step": 7057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2204422857142858e-05,
+      "loss": 1.3818,
+      "step": 7058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2204154285714286e-05,
+      "loss": 1.8392,
+      "step": 7059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2203885714285714e-05,
+      "loss": 1.4941,
+      "step": 7060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2203617142857142e-05,
+      "loss": 1.7538,
+      "step": 7061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2203348571428572e-05,
+      "loss": 1.1797,
+      "step": 7062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.220308e-05,
+      "loss": 1.1442,
+      "step": 7063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2202811428571429e-05,
+      "loss": 1.493,
+      "step": 7064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2202542857142858e-05,
+      "loss": 1.8822,
+      "step": 7065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2202274285714287e-05,
+      "loss": 0.9874,
+      "step": 7066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2202005714285715e-05,
+      "loss": 1.7124,
+      "step": 7067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2201737142857143e-05,
+      "loss": 1.1574,
+      "step": 7068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2201468571428573e-05,
+      "loss": 1.2221,
+      "step": 7069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.22012e-05,
+      "loss": 1.472,
+      "step": 7070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.220093142857143e-05,
+      "loss": 0.9827,
+      "step": 7071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2200662857142857e-05,
+      "loss": 1.3745,
+      "step": 7072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2200394285714286e-05,
+      "loss": 1.7389,
+      "step": 7073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2200125714285714e-05,
+      "loss": 1.4416,
+      "step": 7074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2199857142857144e-05,
+      "loss": 0.9829,
+      "step": 7075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2199588571428572e-05,
+      "loss": 1.3511,
+      "step": 7076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.219932e-05,
+      "loss": 0.8993,
+      "step": 7077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.219905142857143e-05,
+      "loss": 1.0083,
+      "step": 7078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2198782857142858e-05,
+      "loss": 1.4671,
+      "step": 7079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2198514285714286e-05,
+      "loss": 0.8702,
+      "step": 7080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2198245714285714e-05,
+      "loss": 1.2493,
+      "step": 7081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2197977142857144e-05,
+      "loss": 1.7872,
+      "step": 7082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.219770857142857e-05,
+      "loss": 1.5439,
+      "step": 7083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.219744e-05,
+      "loss": 1.2529,
+      "step": 7084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2197171428571429e-05,
+      "loss": 0.8412,
+      "step": 7085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2196902857142859e-05,
+      "loss": 1.2202,
+      "step": 7086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2196634285714285e-05,
+      "loss": 0.9741,
+      "step": 7087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2196365714285715e-05,
+      "loss": 1.22,
+      "step": 7088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2196097142857143e-05,
+      "loss": 1.5739,
+      "step": 7089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2195828571428571e-05,
+      "loss": 1.2127,
+      "step": 7090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2195560000000001e-05,
+      "loss": 1.4932,
+      "step": 7091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.219529142857143e-05,
+      "loss": 1.5021,
+      "step": 7092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2195022857142858e-05,
+      "loss": 1.5452,
+      "step": 7093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2194754285714286e-05,
+      "loss": 1.1929,
+      "step": 7094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2194485714285716e-05,
+      "loss": 1.3521,
+      "step": 7095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2194217142857142e-05,
+      "loss": 1.3401,
+      "step": 7096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2193948571428572e-05,
+      "loss": 1.3424,
+      "step": 7097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.219368e-05,
+      "loss": 1.192,
+      "step": 7098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.219341142857143e-05,
+      "loss": 1.4361,
+      "step": 7099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2193142857142856e-05,
+      "loss": 1.409,
+      "step": 7100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2192874285714286e-05,
+      "loss": 1.428,
+      "step": 7101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2192605714285715e-05,
+      "loss": 0.9854,
+      "step": 7102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2192337142857143e-05,
+      "loss": 1.3078,
+      "step": 7103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2192068571428573e-05,
+      "loss": 1.0148,
+      "step": 7104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.21918e-05,
+      "loss": 1.2916,
+      "step": 7105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2191531428571429e-05,
+      "loss": 1.587,
+      "step": 7106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2191262857142857e-05,
+      "loss": 1.1733,
+      "step": 7107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2190994285714287e-05,
+      "loss": 1.5901,
+      "step": 7108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2190725714285715e-05,
+      "loss": 1.3907,
+      "step": 7109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2190457142857143e-05,
+      "loss": 1.315,
+      "step": 7110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2190188571428571e-05,
+      "loss": 1.2928,
+      "step": 7111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2189920000000001e-05,
+      "loss": 1.5878,
+      "step": 7112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2189651428571428e-05,
+      "loss": 1.3668,
+      "step": 7113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2189382857142858e-05,
+      "loss": 1.5818,
+      "step": 7114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2189114285714286e-05,
+      "loss": 1.0328,
+      "step": 7115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2188845714285714e-05,
+      "loss": 1.2727,
+      "step": 7116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2188577142857144e-05,
+      "loss": 1.3618,
+      "step": 7117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2188308571428572e-05,
+      "loss": 1.1956,
+      "step": 7118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.218804e-05,
+      "loss": 1.1881,
+      "step": 7119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2187771428571428e-05,
+      "loss": 1.6325,
+      "step": 7120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2187502857142858e-05,
+      "loss": 1.8057,
+      "step": 7121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2187234285714287e-05,
+      "loss": 1.4126,
+      "step": 7122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2186965714285715e-05,
+      "loss": 1.6805,
+      "step": 7123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2186697142857143e-05,
+      "loss": 1.7632,
+      "step": 7124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2186428571428573e-05,
+      "loss": 1.3224,
+      "step": 7125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.218616e-05,
+      "loss": 1.057,
+      "step": 7126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2185891428571429e-05,
+      "loss": 0.8197,
+      "step": 7127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2185622857142857e-05,
+      "loss": 1.4913,
+      "step": 7128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2185354285714285e-05,
+      "loss": 1.2904,
+      "step": 7129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2185085714285715e-05,
+      "loss": 1.5074,
+      "step": 7130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2184817142857143e-05,
+      "loss": 1.1878,
+      "step": 7131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2184548571428573e-05,
+      "loss": 1.4748,
+      "step": 7132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.218428e-05,
+      "loss": 0.9178,
+      "step": 7133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.218401142857143e-05,
+      "loss": 1.4632,
+      "step": 7134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2183742857142858e-05,
+      "loss": 1.5409,
+      "step": 7135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2183474285714286e-05,
+      "loss": 1.4412,
+      "step": 7136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2183205714285714e-05,
+      "loss": 1.327,
+      "step": 7137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2182937142857144e-05,
+      "loss": 1.483,
+      "step": 7138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.218266857142857e-05,
+      "loss": 0.9988,
+      "step": 7139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.21824e-05,
+      "loss": 1.5507,
+      "step": 7140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2182131428571429e-05,
+      "loss": 1.2761,
+      "step": 7141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2181862857142857e-05,
+      "loss": 1.6966,
+      "step": 7142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2181594285714287e-05,
+      "loss": 1.4915,
+      "step": 7143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2181325714285715e-05,
+      "loss": 1.3075,
+      "step": 7144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2181057142857145e-05,
+      "loss": 1.3901,
+      "step": 7145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2180788571428571e-05,
+      "loss": 1.3968,
+      "step": 7146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2180520000000001e-05,
+      "loss": 1.1739,
+      "step": 7147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.218025142857143e-05,
+      "loss": 1.6725,
+      "step": 7148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2179982857142857e-05,
+      "loss": 1.1825,
+      "step": 7149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2179714285714286e-05,
+      "loss": 1.3401,
+      "step": 7150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2179445714285715e-05,
+      "loss": 1.1765,
+      "step": 7151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2179177142857142e-05,
+      "loss": 1.3826,
+      "step": 7152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2178908571428572e-05,
+      "loss": 1.1466,
+      "step": 7153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.217864e-05,
+      "loss": 1.3656,
+      "step": 7154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.217837142857143e-05,
+      "loss": 1.6001,
+      "step": 7155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2178102857142858e-05,
+      "loss": 1.1836,
+      "step": 7156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2177834285714286e-05,
+      "loss": 1.2681,
+      "step": 7157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2177565714285716e-05,
+      "loss": 1.2678,
+      "step": 7158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2177297142857143e-05,
+      "loss": 1.3574,
+      "step": 7159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2177028571428572e-05,
+      "loss": 1.2508,
+      "step": 7160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.217676e-05,
+      "loss": 1.606,
+      "step": 7161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2176491428571429e-05,
+      "loss": 1.557,
+      "step": 7162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2176222857142857e-05,
+      "loss": 1.4785,
+      "step": 7163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2175954285714287e-05,
+      "loss": 1.5215,
+      "step": 7164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2175685714285713e-05,
+      "loss": 1.6361,
+      "step": 7165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2175417142857143e-05,
+      "loss": 1.5926,
+      "step": 7166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2175148571428571e-05,
+      "loss": 1.1579,
+      "step": 7167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2174880000000001e-05,
+      "loss": 1.2613,
+      "step": 7168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.217461142857143e-05,
+      "loss": 1.3625,
+      "step": 7169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2174342857142858e-05,
+      "loss": 1.651,
+      "step": 7170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2174074285714287e-05,
+      "loss": 1.4357,
+      "step": 7171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2173805714285714e-05,
+      "loss": 1.428,
+      "step": 7172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2173537142857144e-05,
+      "loss": 1.1309,
+      "step": 7173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2173268571428572e-05,
+      "loss": 1.2176,
+      "step": 7174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2173e-05,
+      "loss": 1.7922,
+      "step": 7175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2172731428571428e-05,
+      "loss": 1.311,
+      "step": 7176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2172462857142858e-05,
+      "loss": 1.274,
+      "step": 7177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2172194285714286e-05,
+      "loss": 0.9546,
+      "step": 7178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2171925714285715e-05,
+      "loss": 1.1599,
+      "step": 7179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2171657142857143e-05,
+      "loss": 1.1782,
+      "step": 7180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2171388571428573e-05,
+      "loss": 1.1405,
+      "step": 7181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.217112e-05,
+      "loss": 1.3754,
+      "step": 7182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2170851428571429e-05,
+      "loss": 1.6671,
+      "step": 7183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2170582857142859e-05,
+      "loss": 1.5858,
+      "step": 7184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2170314285714285e-05,
+      "loss": 1.4789,
+      "step": 7185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2170045714285715e-05,
+      "loss": 1.0113,
+      "step": 7186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2169777142857143e-05,
+      "loss": 1.3689,
+      "step": 7187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2169508571428572e-05,
+      "loss": 1.5106,
+      "step": 7188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.216924e-05,
+      "loss": 1.7857,
+      "step": 7189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.216897142857143e-05,
+      "loss": 1.1946,
+      "step": 7190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2168702857142858e-05,
+      "loss": 1.4629,
+      "step": 7191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2168434285714286e-05,
+      "loss": 1.2559,
+      "step": 7192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2168165714285714e-05,
+      "loss": 1.4174,
+      "step": 7193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2167897142857144e-05,
+      "loss": 1.2209,
+      "step": 7194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2167628571428572e-05,
+      "loss": 1.4457,
+      "step": 7195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.216736e-05,
+      "loss": 1.092,
+      "step": 7196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.216709142857143e-05,
+      "loss": 1.8038,
+      "step": 7197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2166822857142857e-05,
+      "loss": 1.4972,
+      "step": 7198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2166554285714287e-05,
+      "loss": 1.3439,
+      "step": 7199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2166285714285715e-05,
+      "loss": 1.3282,
+      "step": 7200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2166017142857145e-05,
+      "loss": 1.4625,
+      "step": 7201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2165748571428571e-05,
+      "loss": 1.1262,
+      "step": 7202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2165480000000001e-05,
+      "loss": 1.7729,
+      "step": 7203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2165211428571429e-05,
+      "loss": 1.388,
+      "step": 7204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2164942857142857e-05,
+      "loss": 1.1633,
+      "step": 7205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2164674285714285e-05,
+      "loss": 1.297,
+      "step": 7206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2164405714285715e-05,
+      "loss": 1.5006,
+      "step": 7207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2164137142857144e-05,
+      "loss": 1.3542,
+      "step": 7208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2163868571428572e-05,
+      "loss": 1.2377,
+      "step": 7209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2163600000000002e-05,
+      "loss": 1.3448,
+      "step": 7210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2163331428571428e-05,
+      "loss": 1.3626,
+      "step": 7211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2163062857142858e-05,
+      "loss": 1.4723,
+      "step": 7212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2162794285714286e-05,
+      "loss": 1.69,
+      "step": 7213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2162525714285716e-05,
+      "loss": 1.4728,
+      "step": 7214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2162257142857142e-05,
+      "loss": 1.2467,
+      "step": 7215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2161988571428572e-05,
+      "loss": 0.9639,
+      "step": 7216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.216172e-05,
+      "loss": 1.4596,
+      "step": 7217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2161451428571429e-05,
+      "loss": 0.9316,
+      "step": 7218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2161182857142857e-05,
+      "loss": 1.3027,
+      "step": 7219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2160914285714287e-05,
+      "loss": 1.8315,
+      "step": 7220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2160645714285715e-05,
+      "loss": 1.2675,
+      "step": 7221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2160377142857143e-05,
+      "loss": 1.2431,
+      "step": 7222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2160108571428573e-05,
+      "loss": 1.3512,
+      "step": 7223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2159840000000001e-05,
+      "loss": 1.0611,
+      "step": 7224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.215957142857143e-05,
+      "loss": 1.4899,
+      "step": 7225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2159302857142857e-05,
+      "loss": 1.28,
+      "step": 7226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2159034285714287e-05,
+      "loss": 1.1916,
+      "step": 7227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2158765714285714e-05,
+      "loss": 1.1558,
+      "step": 7228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2158497142857144e-05,
+      "loss": 1.2627,
+      "step": 7229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2158228571428572e-05,
+      "loss": 1.204,
+      "step": 7230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.215796e-05,
+      "loss": 1.3235,
+      "step": 7231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2157691428571428e-05,
+      "loss": 1.3583,
+      "step": 7232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2157422857142858e-05,
+      "loss": 1.1152,
+      "step": 7233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2157154285714286e-05,
+      "loss": 1.4694,
+      "step": 7234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2156885714285714e-05,
+      "loss": 0.9069,
+      "step": 7235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2156617142857144e-05,
+      "loss": 1.5276,
+      "step": 7236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2156348571428573e-05,
+      "loss": 1.3823,
+      "step": 7237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.215608e-05,
+      "loss": 1.6657,
+      "step": 7238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2155811428571429e-05,
+      "loss": 1.4263,
+      "step": 7239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2155542857142859e-05,
+      "loss": 1.0893,
+      "step": 7240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2155274285714285e-05,
+      "loss": 1.18,
+      "step": 7241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2155005714285715e-05,
+      "loss": 1.4393,
+      "step": 7242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2154737142857143e-05,
+      "loss": 1.5934,
+      "step": 7243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2154468571428571e-05,
+      "loss": 1.1995,
+      "step": 7244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.21542e-05,
+      "loss": 1.3493,
+      "step": 7245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.215393142857143e-05,
+      "loss": 1.474,
+      "step": 7246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2153662857142858e-05,
+      "loss": 1.3405,
+      "step": 7247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2153394285714286e-05,
+      "loss": 1.243,
+      "step": 7248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2153125714285716e-05,
+      "loss": 1.5167,
+      "step": 7249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2152857142857144e-05,
+      "loss": 1.3344,
+      "step": 7250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2152588571428572e-05,
+      "loss": 1.0892,
+      "step": 7251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.215232e-05,
+      "loss": 1.2276,
+      "step": 7252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.215205142857143e-05,
+      "loss": 1.3819,
+      "step": 7253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2151782857142857e-05,
+      "loss": 1.3366,
+      "step": 7254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2151514285714286e-05,
+      "loss": 1.2967,
+      "step": 7255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2151245714285715e-05,
+      "loss": 1.2202,
+      "step": 7256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2150977142857143e-05,
+      "loss": 1.3176,
+      "step": 7257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2150708571428571e-05,
+      "loss": 1.1755,
+      "step": 7258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2150440000000001e-05,
+      "loss": 1.2265,
+      "step": 7259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2150171428571429e-05,
+      "loss": 1.504,
+      "step": 7260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2149902857142857e-05,
+      "loss": 1.1561,
+      "step": 7261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2149634285714287e-05,
+      "loss": 1.0876,
+      "step": 7262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2149365714285715e-05,
+      "loss": 1.2433,
+      "step": 7263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2149097142857143e-05,
+      "loss": 1.8227,
+      "step": 7264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2148828571428572e-05,
+      "loss": 0.9899,
+      "step": 7265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2148560000000001e-05,
+      "loss": 1.1582,
+      "step": 7266
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2148291428571428e-05,
+      "loss": 1.0807,
+      "step": 7267
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2148022857142858e-05,
+      "loss": 0.9555,
+      "step": 7268
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2147754285714286e-05,
+      "loss": 1.4238,
+      "step": 7269
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2147485714285714e-05,
+      "loss": 1.1555,
+      "step": 7270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2147217142857142e-05,
+      "loss": 1.2593,
+      "step": 7271
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2146948571428572e-05,
+      "loss": 1.2213,
+      "step": 7272
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.214668e-05,
+      "loss": 1.5262,
+      "step": 7273
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2146411428571429e-05,
+      "loss": 1.073,
+      "step": 7274
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2146142857142858e-05,
+      "loss": 1.448,
+      "step": 7275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2145874285714287e-05,
+      "loss": 1.7896,
+      "step": 7276
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2145605714285715e-05,
+      "loss": 1.7462,
+      "step": 7277
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2145337142857143e-05,
+      "loss": 1.0187,
+      "step": 7278
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2145068571428573e-05,
+      "loss": 1.2736,
+      "step": 7279
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.21448e-05,
+      "loss": 1.3244,
+      "step": 7280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.214453142857143e-05,
+      "loss": 1.3256,
+      "step": 7281
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2144262857142857e-05,
+      "loss": 1.1655,
+      "step": 7282
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2143994285714287e-05,
+      "loss": 1.1686,
+      "step": 7283
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2143725714285714e-05,
+      "loss": 1.5388,
+      "step": 7284
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2143457142857144e-05,
+      "loss": 1.3666,
+      "step": 7285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2143188571428572e-05,
+      "loss": 1.2911,
+      "step": 7286
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.214292e-05,
+      "loss": 1.2382,
+      "step": 7287
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.214265142857143e-05,
+      "loss": 1.2546,
+      "step": 7288
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2142382857142858e-05,
+      "loss": 0.8696,
+      "step": 7289
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2142114285714286e-05,
+      "loss": 1.6454,
+      "step": 7290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2141845714285714e-05,
+      "loss": 1.6259,
+      "step": 7291
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2141577142857144e-05,
+      "loss": 1.4753,
+      "step": 7292
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.214130857142857e-05,
+      "loss": 1.3224,
+      "step": 7293
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.214104e-05,
+      "loss": 1.68,
+      "step": 7294
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2140771428571429e-05,
+      "loss": 1.4205,
+      "step": 7295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2140502857142859e-05,
+      "loss": 1.5054,
+      "step": 7296
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2140234285714285e-05,
+      "loss": 1.0074,
+      "step": 7297
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2139965714285715e-05,
+      "loss": 1.2689,
+      "step": 7298
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2139697142857143e-05,
+      "loss": 1.6093,
+      "step": 7299
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2139428571428571e-05,
+      "loss": 1.2319,
+      "step": 7300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2139160000000001e-05,
+      "loss": 1.4923,
+      "step": 7301
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.213889142857143e-05,
+      "loss": 1.0944,
+      "step": 7302
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2138622857142858e-05,
+      "loss": 1.1255,
+      "step": 7303
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2138354285714286e-05,
+      "loss": 1.4365,
+      "step": 7304
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2138085714285716e-05,
+      "loss": 1.3868,
+      "step": 7305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2137817142857144e-05,
+      "loss": 1.0402,
+      "step": 7306
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2137548571428572e-05,
+      "loss": 1.0638,
+      "step": 7307
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.213728e-05,
+      "loss": 1.2778,
+      "step": 7308
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.213701142857143e-05,
+      "loss": 1.0366,
+      "step": 7309
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2136742857142856e-05,
+      "loss": 1.3795,
+      "step": 7310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2136474285714286e-05,
+      "loss": 1.1561,
+      "step": 7311
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2136205714285715e-05,
+      "loss": 1.1788,
+      "step": 7312
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2135937142857143e-05,
+      "loss": 1.3856,
+      "step": 7313
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2135668571428573e-05,
+      "loss": 1.1468,
+      "step": 7314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.21354e-05,
+      "loss": 1.2758,
+      "step": 7315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2135131428571429e-05,
+      "loss": 1.3224,
+      "step": 7316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2134862857142857e-05,
+      "loss": 1.4912,
+      "step": 7317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2134594285714287e-05,
+      "loss": 1.4482,
+      "step": 7318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2134325714285715e-05,
+      "loss": 1.3347,
+      "step": 7319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2134057142857143e-05,
+      "loss": 1.1153,
+      "step": 7320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2133788571428571e-05,
+      "loss": 1.1292,
+      "step": 7321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2133520000000001e-05,
+      "loss": 1.1929,
+      "step": 7322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2133251428571428e-05,
+      "loss": 1.3266,
+      "step": 7323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2132982857142858e-05,
+      "loss": 1.2007,
+      "step": 7324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2132714285714286e-05,
+      "loss": 0.9323,
+      "step": 7325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2132445714285714e-05,
+      "loss": 1.636,
+      "step": 7326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2132177142857144e-05,
+      "loss": 1.1585,
+      "step": 7327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2131908571428572e-05,
+      "loss": 1.3832,
+      "step": 7328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2131640000000002e-05,
+      "loss": 1.2543,
+      "step": 7329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2131371428571428e-05,
+      "loss": 1.5207,
+      "step": 7330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2131102857142858e-05,
+      "loss": 1.382,
+      "step": 7331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2130834285714287e-05,
+      "loss": 1.5854,
+      "step": 7332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2130565714285715e-05,
+      "loss": 1.2719,
+      "step": 7333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2130297142857143e-05,
+      "loss": 1.166,
+      "step": 7334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2130028571428573e-05,
+      "loss": 1.038,
+      "step": 7335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212976e-05,
+      "loss": 1.5187,
+      "step": 7336
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2129491428571429e-05,
+      "loss": 1.545,
+      "step": 7337
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2129222857142857e-05,
+      "loss": 1.3227,
+      "step": 7338
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2128954285714285e-05,
+      "loss": 1.2905,
+      "step": 7339
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2128685714285715e-05,
+      "loss": 1.7801,
+      "step": 7340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2128417142857143e-05,
+      "loss": 1.3947,
+      "step": 7341
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2128148571428573e-05,
+      "loss": 1.6668,
+      "step": 7342
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212788e-05,
+      "loss": 1.3649,
+      "step": 7343
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212761142857143e-05,
+      "loss": 1.3256,
+      "step": 7344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2127342857142858e-05,
+      "loss": 1.6088,
+      "step": 7345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2127074285714286e-05,
+      "loss": 1.7196,
+      "step": 7346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2126805714285714e-05,
+      "loss": 1.5711,
+      "step": 7347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2126537142857144e-05,
+      "loss": 1.1982,
+      "step": 7348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212626857142857e-05,
+      "loss": 1.5481,
+      "step": 7349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2126e-05,
+      "loss": 1.4513,
+      "step": 7350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2125731428571429e-05,
+      "loss": 1.6438,
+      "step": 7351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2125462857142858e-05,
+      "loss": 1.5347,
+      "step": 7352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2125194285714287e-05,
+      "loss": 1.0127,
+      "step": 7353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2124925714285715e-05,
+      "loss": 1.677,
+      "step": 7354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2124657142857145e-05,
+      "loss": 1.2305,
+      "step": 7355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2124388571428571e-05,
+      "loss": 1.5052,
+      "step": 7356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2124120000000001e-05,
+      "loss": 1.1722,
+      "step": 7357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212385142857143e-05,
+      "loss": 1.4112,
+      "step": 7358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2123582857142857e-05,
+      "loss": 0.9478,
+      "step": 7359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2123314285714286e-05,
+      "loss": 1.3841,
+      "step": 7360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2123045714285715e-05,
+      "loss": 1.298,
+      "step": 7361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2122777142857142e-05,
+      "loss": 1.186,
+      "step": 7362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2122508571428572e-05,
+      "loss": 1.3085,
+      "step": 7363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212224e-05,
+      "loss": 1.1773,
+      "step": 7364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212197142857143e-05,
+      "loss": 1.4854,
+      "step": 7365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2121702857142858e-05,
+      "loss": 1.0344,
+      "step": 7366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2121434285714286e-05,
+      "loss": 1.5688,
+      "step": 7367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2121165714285716e-05,
+      "loss": 1.3774,
+      "step": 7368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2120897142857143e-05,
+      "loss": 1.2957,
+      "step": 7369
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2120628571428572e-05,
+      "loss": 1.3213,
+      "step": 7370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.212036e-05,
+      "loss": 1.2799,
+      "step": 7371
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2120091428571429e-05,
+      "loss": 1.6212,
+      "step": 7372
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2119822857142857e-05,
+      "loss": 1.6097,
+      "step": 7373
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2119554285714287e-05,
+      "loss": 1.264,
+      "step": 7374
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2119285714285715e-05,
+      "loss": 1.3591,
+      "step": 7375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2119017142857143e-05,
+      "loss": 1.25,
+      "step": 7376
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2118748571428571e-05,
+      "loss": 0.9874,
+      "step": 7377
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2118480000000001e-05,
+      "loss": 0.879,
+      "step": 7378
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.211821142857143e-05,
+      "loss": 1.5326,
+      "step": 7379
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2117942857142858e-05,
+      "loss": 1.0116,
+      "step": 7380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2117674285714287e-05,
+      "loss": 1.632,
+      "step": 7381
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2117405714285714e-05,
+      "loss": 1.5435,
+      "step": 7382
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2117137142857144e-05,
+      "loss": 1.4673,
+      "step": 7383
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2116868571428572e-05,
+      "loss": 1.4223,
+      "step": 7384
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.21166e-05,
+      "loss": 1.3902,
+      "step": 7385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2116331428571428e-05,
+      "loss": 1.3718,
+      "step": 7386
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2116062857142858e-05,
+      "loss": 0.8939,
+      "step": 7387
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2115794285714286e-05,
+      "loss": 1.5871,
+      "step": 7388
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2115525714285715e-05,
+      "loss": 1.2329,
+      "step": 7389
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2115257142857143e-05,
+      "loss": 1.6351,
+      "step": 7390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2114988571428573e-05,
+      "loss": 1.3231,
+      "step": 7391
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2114719999999999e-05,
+      "loss": 0.7043,
+      "step": 7392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2114451428571429e-05,
+      "loss": 1.2631,
+      "step": 7393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2114182857142859e-05,
+      "loss": 1.2676,
+      "step": 7394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2113914285714285e-05,
+      "loss": 1.0144,
+      "step": 7395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2113645714285715e-05,
+      "loss": 1.4155,
+      "step": 7396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2113377142857143e-05,
+      "loss": 1.4848,
+      "step": 7397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2113108571428572e-05,
+      "loss": 1.4244,
+      "step": 7398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.211284e-05,
+      "loss": 1.2609,
+      "step": 7399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.211257142857143e-05,
+      "loss": 1.2375,
+      "step": 7400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2112302857142858e-05,
+      "loss": 1.4828,
+      "step": 7401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2112034285714286e-05,
+      "loss": 1.3123,
+      "step": 7402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2111765714285714e-05,
+      "loss": 1.0917,
+      "step": 7403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2111497142857144e-05,
+      "loss": 1.5232,
+      "step": 7404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.211122857142857e-05,
+      "loss": 1.2874,
+      "step": 7405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.211096e-05,
+      "loss": 1.65,
+      "step": 7406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.211069142857143e-05,
+      "loss": 1.3791,
+      "step": 7407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2110422857142857e-05,
+      "loss": 1.4672,
+      "step": 7408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2110154285714287e-05,
+      "loss": 1.5256,
+      "step": 7409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2109885714285715e-05,
+      "loss": 1.4437,
+      "step": 7410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2109617142857145e-05,
+      "loss": 1.3604,
+      "step": 7411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2109348571428571e-05,
+      "loss": 1.241,
+      "step": 7412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2109080000000001e-05,
+      "loss": 1.2571,
+      "step": 7413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2108811428571429e-05,
+      "loss": 1.3492,
+      "step": 7414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2108542857142857e-05,
+      "loss": 1.3532,
+      "step": 7415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2108274285714285e-05,
+      "loss": 1.5284,
+      "step": 7416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2108005714285715e-05,
+      "loss": 1.145,
+      "step": 7417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2107737142857142e-05,
+      "loss": 1.104,
+      "step": 7418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2107468571428572e-05,
+      "loss": 1.5955,
+      "step": 7419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2107200000000002e-05,
+      "loss": 1.3091,
+      "step": 7420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2106931428571428e-05,
+      "loss": 1.1529,
+      "step": 7421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2106662857142858e-05,
+      "loss": 1.6179,
+      "step": 7422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2106394285714286e-05,
+      "loss": 1.5456,
+      "step": 7423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2106125714285716e-05,
+      "loss": 0.8757,
+      "step": 7424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2105857142857142e-05,
+      "loss": 1.4376,
+      "step": 7425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2105588571428572e-05,
+      "loss": 1.6805,
+      "step": 7426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.210532e-05,
+      "loss": 1.2856,
+      "step": 7427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2105051428571429e-05,
+      "loss": 1.4948,
+      "step": 7428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2104782857142857e-05,
+      "loss": 1.1936,
+      "step": 7429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2104514285714287e-05,
+      "loss": 1.5865,
+      "step": 7430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2104245714285713e-05,
+      "loss": 1.3152,
+      "step": 7431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2103977142857143e-05,
+      "loss": 1.1824,
+      "step": 7432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2103708571428573e-05,
+      "loss": 1.4515,
+      "step": 7433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2103440000000001e-05,
+      "loss": 1.6806,
+      "step": 7434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.210317142857143e-05,
+      "loss": 1.621,
+      "step": 7435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2102902857142857e-05,
+      "loss": 1.2251,
+      "step": 7436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2102634285714287e-05,
+      "loss": 1.3809,
+      "step": 7437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2102365714285714e-05,
+      "loss": 1.6892,
+      "step": 7438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2102097142857144e-05,
+      "loss": 1.4388,
+      "step": 7439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2101828571428572e-05,
+      "loss": 0.8836,
+      "step": 7440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.210156e-05,
+      "loss": 1.4644,
+      "step": 7441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2101291428571428e-05,
+      "loss": 1.2053,
+      "step": 7442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2101022857142858e-05,
+      "loss": 1.579,
+      "step": 7443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2100754285714285e-05,
+      "loss": 1.1554,
+      "step": 7444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2100485714285714e-05,
+      "loss": 1.0284,
+      "step": 7445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2100217142857144e-05,
+      "loss": 1.3849,
+      "step": 7446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2099948571428573e-05,
+      "loss": 1.31,
+      "step": 7447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.209968e-05,
+      "loss": 1.274,
+      "step": 7448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2099411428571429e-05,
+      "loss": 1.0944,
+      "step": 7449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2099142857142859e-05,
+      "loss": 1.561,
+      "step": 7450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2098874285714285e-05,
+      "loss": 1.2595,
+      "step": 7451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2098605714285715e-05,
+      "loss": 1.515,
+      "step": 7452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2098337142857143e-05,
+      "loss": 1.3519,
+      "step": 7453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2098068571428571e-05,
+      "loss": 1.2768,
+      "step": 7454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.20978e-05,
+      "loss": 1.1461,
+      "step": 7455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.209753142857143e-05,
+      "loss": 1.319,
+      "step": 7456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2097262857142858e-05,
+      "loss": 1.4762,
+      "step": 7457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2096994285714286e-05,
+      "loss": 0.9096,
+      "step": 7458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2096725714285716e-05,
+      "loss": 1.0787,
+      "step": 7459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2096457142857144e-05,
+      "loss": 1.3775,
+      "step": 7460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2096188571428572e-05,
+      "loss": 1.5905,
+      "step": 7461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.209592e-05,
+      "loss": 1.3999,
+      "step": 7462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.209565142857143e-05,
+      "loss": 1.4286,
+      "step": 7463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2095382857142857e-05,
+      "loss": 1.3834,
+      "step": 7464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2095114285714286e-05,
+      "loss": 1.1618,
+      "step": 7465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2094845714285715e-05,
+      "loss": 0.9117,
+      "step": 7466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2094577142857143e-05,
+      "loss": 1.2573,
+      "step": 7467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2094308571428571e-05,
+      "loss": 1.4542,
+      "step": 7468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.209404e-05,
+      "loss": 1.4979,
+      "step": 7469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2093771428571429e-05,
+      "loss": 1.4924,
+      "step": 7470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2093502857142857e-05,
+      "loss": 1.0982,
+      "step": 7471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2093234285714287e-05,
+      "loss": 1.3665,
+      "step": 7472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2092965714285715e-05,
+      "loss": 1.187,
+      "step": 7473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2092697142857143e-05,
+      "loss": 1.3276,
+      "step": 7474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2092428571428572e-05,
+      "loss": 1.2877,
+      "step": 7475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2092160000000001e-05,
+      "loss": 1.453,
+      "step": 7476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2091891428571428e-05,
+      "loss": 1.4819,
+      "step": 7477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2091622857142858e-05,
+      "loss": 1.0874,
+      "step": 7478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2091354285714286e-05,
+      "loss": 1.5936,
+      "step": 7479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2091085714285716e-05,
+      "loss": 1.4722,
+      "step": 7480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2090817142857142e-05,
+      "loss": 1.4565,
+      "step": 7481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2090548571428572e-05,
+      "loss": 1.1309,
+      "step": 7482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.209028e-05,
+      "loss": 1.845,
+      "step": 7483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2090011428571429e-05,
+      "loss": 1.402,
+      "step": 7484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2089742857142858e-05,
+      "loss": 1.5604,
+      "step": 7485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2089474285714287e-05,
+      "loss": 1.4475,
+      "step": 7486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2089205714285715e-05,
+      "loss": 1.3952,
+      "step": 7487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2088937142857143e-05,
+      "loss": 1.2793,
+      "step": 7488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2088668571428573e-05,
+      "loss": 1.2431,
+      "step": 7489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.20884e-05,
+      "loss": 1.6014,
+      "step": 7490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208813142857143e-05,
+      "loss": 2.0764,
+      "step": 7491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2087862857142857e-05,
+      "loss": 1.3587,
+      "step": 7492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2087594285714287e-05,
+      "loss": 1.3567,
+      "step": 7493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2087325714285714e-05,
+      "loss": 1.6706,
+      "step": 7494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2087057142857144e-05,
+      "loss": 0.9634,
+      "step": 7495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2086788571428572e-05,
+      "loss": 1.3893,
+      "step": 7496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208652e-05,
+      "loss": 1.1397,
+      "step": 7497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2086251428571428e-05,
+      "loss": 1.3493,
+      "step": 7498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2085982857142858e-05,
+      "loss": 1.1365,
+      "step": 7499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2085714285714286e-05,
+      "loss": 1.198,
+      "step": 7500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2085445714285714e-05,
+      "loss": 1.3532,
+      "step": 7501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2085177142857144e-05,
+      "loss": 1.1336,
+      "step": 7502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2084908571428572e-05,
+      "loss": 1.4919,
+      "step": 7503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208464e-05,
+      "loss": 1.3454,
+      "step": 7504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2084371428571429e-05,
+      "loss": 1.1605,
+      "step": 7505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2084102857142859e-05,
+      "loss": 1.0314,
+      "step": 7506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2083834285714285e-05,
+      "loss": 1.5268,
+      "step": 7507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2083565714285715e-05,
+      "loss": 1.0861,
+      "step": 7508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2083297142857143e-05,
+      "loss": 1.3444,
+      "step": 7509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2083028571428571e-05,
+      "loss": 1.3619,
+      "step": 7510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208276e-05,
+      "loss": 1.3715,
+      "step": 7511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208249142857143e-05,
+      "loss": 1.7307,
+      "step": 7512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2082222857142858e-05,
+      "loss": 1.0816,
+      "step": 7513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2081954285714286e-05,
+      "loss": 1.5874,
+      "step": 7514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2081685714285716e-05,
+      "loss": 1.6035,
+      "step": 7515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2081417142857144e-05,
+      "loss": 1.3978,
+      "step": 7516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2081148571428572e-05,
+      "loss": 0.9415,
+      "step": 7517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208088e-05,
+      "loss": 1.681,
+      "step": 7518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208061142857143e-05,
+      "loss": 1.525,
+      "step": 7519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2080342857142856e-05,
+      "loss": 1.464,
+      "step": 7520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2080074285714286e-05,
+      "loss": 1.4625,
+      "step": 7521
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2079805714285715e-05,
+      "loss": 1.0421,
+      "step": 7522
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2079537142857143e-05,
+      "loss": 1.3186,
+      "step": 7523
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2079268571428571e-05,
+      "loss": 1.0892,
+      "step": 7524
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2079e-05,
+      "loss": 0.7964,
+      "step": 7525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.207873142857143e-05,
+      "loss": 1.3383,
+      "step": 7526
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2078462857142857e-05,
+      "loss": 1.2627,
+      "step": 7527
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2078194285714287e-05,
+      "loss": 1.1922,
+      "step": 7528
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2077925714285715e-05,
+      "loss": 1.5477,
+      "step": 7529
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2077657142857143e-05,
+      "loss": 1.2952,
+      "step": 7530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2077388571428571e-05,
+      "loss": 1.5987,
+      "step": 7531
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2077120000000001e-05,
+      "loss": 0.8131,
+      "step": 7532
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2076851428571428e-05,
+      "loss": 1.3559,
+      "step": 7533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2076582857142858e-05,
+      "loss": 0.9246,
+      "step": 7534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2076314285714286e-05,
+      "loss": 1.6186,
+      "step": 7535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2076045714285714e-05,
+      "loss": 1.1653,
+      "step": 7536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2075777142857142e-05,
+      "loss": 1.5866,
+      "step": 7537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2075508571428572e-05,
+      "loss": 1.3729,
+      "step": 7538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2075240000000002e-05,
+      "loss": 1.0753,
+      "step": 7539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2074971428571428e-05,
+      "loss": 1.1748,
+      "step": 7540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2074702857142858e-05,
+      "loss": 1.2439,
+      "step": 7541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2074434285714287e-05,
+      "loss": 1.1359,
+      "step": 7542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2074165714285715e-05,
+      "loss": 0.8075,
+      "step": 7543
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2073897142857143e-05,
+      "loss": 1.4454,
+      "step": 7544
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2073628571428573e-05,
+      "loss": 1.2836,
+      "step": 7545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.207336e-05,
+      "loss": 1.8783,
+      "step": 7546
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2073091428571429e-05,
+      "loss": 0.9378,
+      "step": 7547
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2072822857142857e-05,
+      "loss": 1.1251,
+      "step": 7548
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2072554285714287e-05,
+      "loss": 1.3863,
+      "step": 7549
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2072285714285714e-05,
+      "loss": 1.5787,
+      "step": 7550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2072017142857143e-05,
+      "loss": 1.3321,
+      "step": 7551
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2071748571428573e-05,
+      "loss": 1.1047,
+      "step": 7552
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.207148e-05,
+      "loss": 1.6603,
+      "step": 7553
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.207121142857143e-05,
+      "loss": 1.3608,
+      "step": 7554
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2070942857142858e-05,
+      "loss": 1.2847,
+      "step": 7555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2070674285714286e-05,
+      "loss": 1.3589,
+      "step": 7556
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2070405714285714e-05,
+      "loss": 1.327,
+      "step": 7557
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2070137142857144e-05,
+      "loss": 1.6215,
+      "step": 7558
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.206986857142857e-05,
+      "loss": 1.2148,
+      "step": 7559
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.20696e-05,
+      "loss": 1.1305,
+      "step": 7560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2069331428571429e-05,
+      "loss": 1.2137,
+      "step": 7561
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2069062857142858e-05,
+      "loss": 1.198,
+      "step": 7562
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2068794285714285e-05,
+      "loss": 1.2841,
+      "step": 7563
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2068525714285715e-05,
+      "loss": 1.2831,
+      "step": 7564
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2068257142857145e-05,
+      "loss": 1.9406,
+      "step": 7565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2067988571428571e-05,
+      "loss": 1.3379,
+      "step": 7566
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2067720000000001e-05,
+      "loss": 1.5858,
+      "step": 7567
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.206745142857143e-05,
+      "loss": 1.3547,
+      "step": 7568
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2067182857142857e-05,
+      "loss": 1.2306,
+      "step": 7569
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2066914285714286e-05,
+      "loss": 1.3628,
+      "step": 7570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2066645714285715e-05,
+      "loss": 1.6468,
+      "step": 7571
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2066377142857142e-05,
+      "loss": 1.3598,
+      "step": 7572
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2066108571428572e-05,
+      "loss": 1.4018,
+      "step": 7573
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.206584e-05,
+      "loss": 1.3941,
+      "step": 7574
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.206557142857143e-05,
+      "loss": 1.5563,
+      "step": 7575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2065302857142856e-05,
+      "loss": 1.4176,
+      "step": 7576
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2065034285714286e-05,
+      "loss": 1.1583,
+      "step": 7577
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2064765714285716e-05,
+      "loss": 1.0605,
+      "step": 7578
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2064497142857143e-05,
+      "loss": 1.0959,
+      "step": 7579
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2064228571428572e-05,
+      "loss": 1.6058,
+      "step": 7580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.206396e-05,
+      "loss": 1.4035,
+      "step": 7581
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2063691428571429e-05,
+      "loss": 1.4714,
+      "step": 7582
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2063422857142857e-05,
+      "loss": 1.6503,
+      "step": 7583
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2063154285714287e-05,
+      "loss": 1.2091,
+      "step": 7584
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2062885714285715e-05,
+      "loss": 1.246,
+      "step": 7585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2062617142857143e-05,
+      "loss": 1.1461,
+      "step": 7586
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2062348571428571e-05,
+      "loss": 1.4448,
+      "step": 7587
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2062080000000001e-05,
+      "loss": 1.3546,
+      "step": 7588
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2061811428571428e-05,
+      "loss": 1.354,
+      "step": 7589
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2061542857142858e-05,
+      "loss": 1.4061,
+      "step": 7590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2061274285714287e-05,
+      "loss": 1.26,
+      "step": 7591
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2061005714285714e-05,
+      "loss": 1.733,
+      "step": 7592
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2060737142857144e-05,
+      "loss": 1.5391,
+      "step": 7593
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2060468571428572e-05,
+      "loss": 0.9199,
+      "step": 7594
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.20602e-05,
+      "loss": 1.3562,
+      "step": 7595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2059931428571428e-05,
+      "loss": 1.1697,
+      "step": 7596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2059662857142858e-05,
+      "loss": 1.3889,
+      "step": 7597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2059394285714286e-05,
+      "loss": 1.3823,
+      "step": 7598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2059125714285715e-05,
+      "loss": 1.2612,
+      "step": 7599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2058857142857143e-05,
+      "loss": 1.2332,
+      "step": 7600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2058588571428573e-05,
+      "loss": 1.1723,
+      "step": 7601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2058319999999999e-05,
+      "loss": 1.143,
+      "step": 7602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2058051428571429e-05,
+      "loss": 1.4619,
+      "step": 7603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2057782857142857e-05,
+      "loss": 1.3497,
+      "step": 7604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2057514285714285e-05,
+      "loss": 1.4652,
+      "step": 7605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2057245714285715e-05,
+      "loss": 1.0712,
+      "step": 7606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2056977142857143e-05,
+      "loss": 1.2495,
+      "step": 7607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2056708571428573e-05,
+      "loss": 1.5878,
+      "step": 7608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.205644e-05,
+      "loss": 1.1758,
+      "step": 7609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.205617142857143e-05,
+      "loss": 1.3716,
+      "step": 7610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2055902857142858e-05,
+      "loss": 1.2988,
+      "step": 7611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2055634285714286e-05,
+      "loss": 1.2135,
+      "step": 7612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2055365714285714e-05,
+      "loss": 1.3333,
+      "step": 7613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2055097142857144e-05,
+      "loss": 1.4943,
+      "step": 7614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.205482857142857e-05,
+      "loss": 1.345,
+      "step": 7615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.205456e-05,
+      "loss": 1.3671,
+      "step": 7616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2054291428571429e-05,
+      "loss": 0.8658,
+      "step": 7617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2054022857142857e-05,
+      "loss": 1.2791,
+      "step": 7618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2053754285714287e-05,
+      "loss": 0.9047,
+      "step": 7619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2053485714285715e-05,
+      "loss": 1.2349,
+      "step": 7620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2053217142857145e-05,
+      "loss": 1.7678,
+      "step": 7621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2052948571428571e-05,
+      "loss": 1.2858,
+      "step": 7622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2052680000000001e-05,
+      "loss": 1.203,
+      "step": 7623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2052411428571429e-05,
+      "loss": 1.2907,
+      "step": 7624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2052142857142857e-05,
+      "loss": 1.4609,
+      "step": 7625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2051874285714285e-05,
+      "loss": 1.3132,
+      "step": 7626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2051605714285715e-05,
+      "loss": 1.2172,
+      "step": 7627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2051337142857142e-05,
+      "loss": 1.4876,
+      "step": 7628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2051068571428572e-05,
+      "loss": 1.0331,
+      "step": 7629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.20508e-05,
+      "loss": 1.3157,
+      "step": 7630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.205053142857143e-05,
+      "loss": 1.3776,
+      "step": 7631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2050262857142858e-05,
+      "loss": 1.134,
+      "step": 7632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2049994285714286e-05,
+      "loss": 1.0728,
+      "step": 7633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2049725714285716e-05,
+      "loss": 1.1718,
+      "step": 7634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2049457142857142e-05,
+      "loss": 1.6436,
+      "step": 7635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2049188571428572e-05,
+      "loss": 1.7583,
+      "step": 7636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.204892e-05,
+      "loss": 1.441,
+      "step": 7637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2048651428571429e-05,
+      "loss": 1.3758,
+      "step": 7638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2048382857142857e-05,
+      "loss": 1.6053,
+      "step": 7639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2048114285714287e-05,
+      "loss": 1.333,
+      "step": 7640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2047845714285713e-05,
+      "loss": 1.2549,
+      "step": 7641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2047577142857143e-05,
+      "loss": 1.3039,
+      "step": 7642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2047308571428571e-05,
+      "loss": 1.1747,
+      "step": 7643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2047040000000001e-05,
+      "loss": 1.43,
+      "step": 7644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.204677142857143e-05,
+      "loss": 1.2481,
+      "step": 7645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2046502857142857e-05,
+      "loss": 1.1743,
+      "step": 7646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2046234285714287e-05,
+      "loss": 1.27,
+      "step": 7647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2045965714285714e-05,
+      "loss": 1.5826,
+      "step": 7648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2045697142857144e-05,
+      "loss": 1.726,
+      "step": 7649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2045428571428572e-05,
+      "loss": 1.2358,
+      "step": 7650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.204516e-05,
+      "loss": 1.1964,
+      "step": 7651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2044891428571428e-05,
+      "loss": 1.7463,
+      "step": 7652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2044622857142858e-05,
+      "loss": 1.0798,
+      "step": 7653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2044354285714286e-05,
+      "loss": 1.4021,
+      "step": 7654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2044085714285714e-05,
+      "loss": 1.1302,
+      "step": 7655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2043817142857143e-05,
+      "loss": 1.6339,
+      "step": 7656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2043548571428572e-05,
+      "loss": 1.435,
+      "step": 7657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.204328e-05,
+      "loss": 1.0234,
+      "step": 7658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2043011428571429e-05,
+      "loss": 1.0318,
+      "step": 7659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2042742857142859e-05,
+      "loss": 1.0777,
+      "step": 7660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2042474285714285e-05,
+      "loss": 1.3146,
+      "step": 7661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2042205714285715e-05,
+      "loss": 1.6714,
+      "step": 7662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2041937142857143e-05,
+      "loss": 1.2172,
+      "step": 7663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2041668571428571e-05,
+      "loss": 1.3875,
+      "step": 7664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.20414e-05,
+      "loss": 1.5378,
+      "step": 7665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.204113142857143e-05,
+      "loss": 1.7085,
+      "step": 7666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2040862857142858e-05,
+      "loss": 1.502,
+      "step": 7667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2040594285714286e-05,
+      "loss": 0.7877,
+      "step": 7668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2040325714285714e-05,
+      "loss": 1.2386,
+      "step": 7669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2040057142857144e-05,
+      "loss": 1.556,
+      "step": 7670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2039788571428572e-05,
+      "loss": 1.0646,
+      "step": 7671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.203952e-05,
+      "loss": 1.565,
+      "step": 7672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.203925142857143e-05,
+      "loss": 1.2531,
+      "step": 7673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2038982857142857e-05,
+      "loss": 1.5041,
+      "step": 7674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2038714285714286e-05,
+      "loss": 1.1134,
+      "step": 7675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2038445714285715e-05,
+      "loss": 1.3987,
+      "step": 7676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2038177142857144e-05,
+      "loss": 1.8199,
+      "step": 7677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2037908571428571e-05,
+      "loss": 1.591,
+      "step": 7678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.203764e-05,
+      "loss": 1.4537,
+      "step": 7679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2037371428571429e-05,
+      "loss": 1.5236,
+      "step": 7680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2037102857142857e-05,
+      "loss": 1.2879,
+      "step": 7681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2036834285714285e-05,
+      "loss": 1.6275,
+      "step": 7682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2036565714285715e-05,
+      "loss": 1.1298,
+      "step": 7683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2036297142857143e-05,
+      "loss": 1.2774,
+      "step": 7684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2036028571428572e-05,
+      "loss": 1.5982,
+      "step": 7685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2035760000000001e-05,
+      "loss": 1.4245,
+      "step": 7686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2035491428571428e-05,
+      "loss": 1.6082,
+      "step": 7687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2035222857142858e-05,
+      "loss": 1.3679,
+      "step": 7688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2034954285714286e-05,
+      "loss": 1.3591,
+      "step": 7689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2034685714285716e-05,
+      "loss": 1.2494,
+      "step": 7690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2034417142857142e-05,
+      "loss": 0.9428,
+      "step": 7691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2034148571428572e-05,
+      "loss": 1.7274,
+      "step": 7692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.203388e-05,
+      "loss": 1.3922,
+      "step": 7693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2033611428571429e-05,
+      "loss": 1.608,
+      "step": 7694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2033342857142857e-05,
+      "loss": 1.5027,
+      "step": 7695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2033074285714287e-05,
+      "loss": 1.5794,
+      "step": 7696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2032805714285715e-05,
+      "loss": 0.891,
+      "step": 7697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2032537142857143e-05,
+      "loss": 1.5046,
+      "step": 7698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2032268571428573e-05,
+      "loss": 1.0895,
+      "step": 7699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2032000000000001e-05,
+      "loss": 1.2227,
+      "step": 7700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.203173142857143e-05,
+      "loss": 1.2392,
+      "step": 7701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2031462857142857e-05,
+      "loss": 0.9142,
+      "step": 7702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2031194285714287e-05,
+      "loss": 1.2139,
+      "step": 7703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2030925714285714e-05,
+      "loss": 0.8572,
+      "step": 7704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2030657142857144e-05,
+      "loss": 1.3706,
+      "step": 7705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2030388571428572e-05,
+      "loss": 1.412,
+      "step": 7706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.203012e-05,
+      "loss": 1.3406,
+      "step": 7707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2029851428571428e-05,
+      "loss": 1.4242,
+      "step": 7708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2029582857142858e-05,
+      "loss": 1.2151,
+      "step": 7709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2029314285714286e-05,
+      "loss": 1.1868,
+      "step": 7710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2029045714285714e-05,
+      "loss": 1.3545,
+      "step": 7711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2028777142857144e-05,
+      "loss": 1.213,
+      "step": 7712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2028508571428572e-05,
+      "loss": 0.9613,
+      "step": 7713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.202824e-05,
+      "loss": 1.3114,
+      "step": 7714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2027971428571429e-05,
+      "loss": 1.31,
+      "step": 7715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2027702857142859e-05,
+      "loss": 1.5695,
+      "step": 7716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2027434285714285e-05,
+      "loss": 0.9091,
+      "step": 7717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2027165714285715e-05,
+      "loss": 1.8,
+      "step": 7718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2026897142857143e-05,
+      "loss": 1.938,
+      "step": 7719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2026628571428571e-05,
+      "loss": 1.5083,
+      "step": 7720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.202636e-05,
+      "loss": 1.2399,
+      "step": 7721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.202609142857143e-05,
+      "loss": 1.2782,
+      "step": 7722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2025822857142858e-05,
+      "loss": 1.3911,
+      "step": 7723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2025554285714286e-05,
+      "loss": 1.224,
+      "step": 7724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2025285714285716e-05,
+      "loss": 1.7096,
+      "step": 7725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2025017142857144e-05,
+      "loss": 1.232,
+      "step": 7726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2024748571428572e-05,
+      "loss": 1.3985,
+      "step": 7727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.202448e-05,
+      "loss": 1.6225,
+      "step": 7728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.202421142857143e-05,
+      "loss": 1.1903,
+      "step": 7729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2023942857142856e-05,
+      "loss": 1.1608,
+      "step": 7730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2023674285714286e-05,
+      "loss": 1.2798,
+      "step": 7731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2023405714285715e-05,
+      "loss": 1.5275,
+      "step": 7732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2023137142857143e-05,
+      "loss": 1.3398,
+      "step": 7733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2022868571428571e-05,
+      "loss": 1.3296,
+      "step": 7734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.20226e-05,
+      "loss": 1.3637,
+      "step": 7735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2022331428571429e-05,
+      "loss": 1.2707,
+      "step": 7736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2022062857142857e-05,
+      "loss": 1.3838,
+      "step": 7737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2021794285714287e-05,
+      "loss": 1.6037,
+      "step": 7738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2021525714285715e-05,
+      "loss": 0.8712,
+      "step": 7739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2021257142857143e-05,
+      "loss": 1.2342,
+      "step": 7740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2020988571428571e-05,
+      "loss": 1.0689,
+      "step": 7741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2020720000000001e-05,
+      "loss": 1.3598,
+      "step": 7742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2020451428571428e-05,
+      "loss": 1.2883,
+      "step": 7743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2020182857142858e-05,
+      "loss": 1.1367,
+      "step": 7744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2019914285714286e-05,
+      "loss": 1.1405,
+      "step": 7745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2019645714285714e-05,
+      "loss": 1.6171,
+      "step": 7746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2019377142857142e-05,
+      "loss": 1.6317,
+      "step": 7747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2019108571428572e-05,
+      "loss": 1.195,
+      "step": 7748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.201884e-05,
+      "loss": 1.0267,
+      "step": 7749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2018571428571428e-05,
+      "loss": 1.3376,
+      "step": 7750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2018302857142858e-05,
+      "loss": 1.4138,
+      "step": 7751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2018034285714287e-05,
+      "loss": 1.3916,
+      "step": 7752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2017765714285715e-05,
+      "loss": 1.3403,
+      "step": 7753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2017497142857143e-05,
+      "loss": 1.5703,
+      "step": 7754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2017228571428573e-05,
+      "loss": 1.5009,
+      "step": 7755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.201696e-05,
+      "loss": 1.5007,
+      "step": 7756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2016691428571429e-05,
+      "loss": 1.5152,
+      "step": 7757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2016422857142857e-05,
+      "loss": 1.6811,
+      "step": 7758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2016154285714287e-05,
+      "loss": 1.3153,
+      "step": 7759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2015885714285714e-05,
+      "loss": 1.3553,
+      "step": 7760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2015617142857143e-05,
+      "loss": 1.2728,
+      "step": 7761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2015348571428572e-05,
+      "loss": 1.2102,
+      "step": 7762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.201508e-05,
+      "loss": 1.1287,
+      "step": 7763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.201481142857143e-05,
+      "loss": 0.9456,
+      "step": 7764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2014542857142858e-05,
+      "loss": 1.5162,
+      "step": 7765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2014274285714286e-05,
+      "loss": 1.7613,
+      "step": 7766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2014005714285714e-05,
+      "loss": 1.3578,
+      "step": 7767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2013737142857144e-05,
+      "loss": 1.0295,
+      "step": 7768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.201346857142857e-05,
+      "loss": 1.0069,
+      "step": 7769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.20132e-05,
+      "loss": 1.0398,
+      "step": 7770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2012931428571429e-05,
+      "loss": 1.702,
+      "step": 7771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2012662857142858e-05,
+      "loss": 0.9963,
+      "step": 7772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2012394285714285e-05,
+      "loss": 1.7925,
+      "step": 7773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2012125714285715e-05,
+      "loss": 1.0922,
+      "step": 7774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2011857142857143e-05,
+      "loss": 1.2956,
+      "step": 7775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2011588571428571e-05,
+      "loss": 1.3696,
+      "step": 7776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2011320000000001e-05,
+      "loss": 1.2042,
+      "step": 7777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.201105142857143e-05,
+      "loss": 0.9947,
+      "step": 7778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2010782857142857e-05,
+      "loss": 1.6875,
+      "step": 7779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2010514285714286e-05,
+      "loss": 1.0557,
+      "step": 7780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2010245714285715e-05,
+      "loss": 1.199,
+      "step": 7781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2009977142857144e-05,
+      "loss": 1.5166,
+      "step": 7782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2009708571428572e-05,
+      "loss": 1.3976,
+      "step": 7783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.200944e-05,
+      "loss": 1.2645,
+      "step": 7784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.200917142857143e-05,
+      "loss": 1.0744,
+      "step": 7785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2008902857142856e-05,
+      "loss": 1.6228,
+      "step": 7786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2008634285714286e-05,
+      "loss": 1.3625,
+      "step": 7787
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008365714285714e-05,
+      "loss": 1.2227,
+      "step": 7788
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008097142857143e-05,
+      "loss": 1.2217,
+      "step": 7789
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2007828571428572e-05,
+      "loss": 0.9689,
+      "step": 7790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200756e-05,
+      "loss": 1.6828,
+      "step": 7791
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2007291428571429e-05,
+      "loss": 1.2995,
+      "step": 7792
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2007022857142857e-05,
+      "loss": 1.4686,
+      "step": 7793
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2006754285714287e-05,
+      "loss": 1.3112,
+      "step": 7794
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2006485714285715e-05,
+      "loss": 1.8431,
+      "step": 7795
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2006217142857143e-05,
+      "loss": 1.3047,
+      "step": 7796
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005948571428571e-05,
+      "loss": 1.2296,
+      "step": 7797
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005680000000001e-05,
+      "loss": 1.2915,
+      "step": 7798
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005411428571428e-05,
+      "loss": 1.4136,
+      "step": 7799
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005142857142858e-05,
+      "loss": 1.2926,
+      "step": 7800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004874285714286e-05,
+      "loss": 1.6282,
+      "step": 7801
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004605714285714e-05,
+      "loss": 1.1363,
+      "step": 7802
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004337142857144e-05,
+      "loss": 1.364,
+      "step": 7803
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004068571428572e-05,
+      "loss": 1.0019,
+      "step": 7804
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003800000000002e-05,
+      "loss": 1.4118,
+      "step": 7805
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003531428571428e-05,
+      "loss": 1.4092,
+      "step": 7806
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003262857142858e-05,
+      "loss": 1.5163,
+      "step": 7807
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002994285714286e-05,
+      "loss": 1.3971,
+      "step": 7808
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002725714285715e-05,
+      "loss": 1.0777,
+      "step": 7809
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002457142857143e-05,
+      "loss": 1.2957,
+      "step": 7810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002188571428573e-05,
+      "loss": 1.5146,
+      "step": 7811
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001919999999999e-05,
+      "loss": 1.6272,
+      "step": 7812
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001651428571429e-05,
+      "loss": 1.3542,
+      "step": 7813
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001382857142857e-05,
+      "loss": 1.0093,
+      "step": 7814
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001114285714285e-05,
+      "loss": 0.9111,
+      "step": 7815
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2000845714285715e-05,
+      "loss": 0.9258,
+      "step": 7816
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2000577142857143e-05,
+      "loss": 0.8929,
+      "step": 7817
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2000308571428573e-05,
+      "loss": 1.4021,
+      "step": 7818
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200004e-05,
+      "loss": 1.4407,
+      "step": 7819
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199977142857143e-05,
+      "loss": 1.1308,
+      "step": 7820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999502857142858e-05,
+      "loss": 0.8754,
+      "step": 7821
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999234285714286e-05,
+      "loss": 1.3811,
+      "step": 7822
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998965714285714e-05,
+      "loss": 1.1994,
+      "step": 7823
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998697142857144e-05,
+      "loss": 1.4757,
+      "step": 7824
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199842857142857e-05,
+      "loss": 1.152,
+      "step": 7825
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199816e-05,
+      "loss": 1.1911,
+      "step": 7826
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997891428571429e-05,
+      "loss": 1.5103,
+      "step": 7827
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997622857142858e-05,
+      "loss": 1.1874,
+      "step": 7828
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997354285714287e-05,
+      "loss": 1.4427,
+      "step": 7829
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997085714285715e-05,
+      "loss": 1.3007,
+      "step": 7830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996817142857145e-05,
+      "loss": 1.7197,
+      "step": 7831
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996548571428571e-05,
+      "loss": 1.2628,
+      "step": 7832
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996280000000001e-05,
+      "loss": 1.6043,
+      "step": 7833
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996011428571429e-05,
+      "loss": 1.6393,
+      "step": 7834
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995742857142857e-05,
+      "loss": 1.4215,
+      "step": 7835
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995474285714285e-05,
+      "loss": 1.1676,
+      "step": 7836
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995205714285715e-05,
+      "loss": 1.5022,
+      "step": 7837
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1994937142857142e-05,
+      "loss": 1.3366,
+      "step": 7838
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1994668571428572e-05,
+      "loss": 0.9837,
+      "step": 7839
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.19944e-05,
+      "loss": 1.0522,
+      "step": 7840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199413142857143e-05,
+      "loss": 1.3979,
+      "step": 7841
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1993862857142858e-05,
+      "loss": 1.361,
+      "step": 7842
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1993594285714286e-05,
+      "loss": 1.4679,
+      "step": 7843
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1993325714285716e-05,
+      "loss": 1.544,
+      "step": 7844
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1993057142857142e-05,
+      "loss": 1.3457,
+      "step": 7845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1992788571428572e-05,
+      "loss": 1.2543,
+      "step": 7846
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199252e-05,
+      "loss": 1.5395,
+      "step": 7847
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1992251428571429e-05,
+      "loss": 1.2354,
+      "step": 7848
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991982857142857e-05,
+      "loss": 1.0562,
+      "step": 7849
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991714285714287e-05,
+      "loss": 1.0361,
+      "step": 7850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991445714285715e-05,
+      "loss": 1.2542,
+      "step": 7851
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991177142857143e-05,
+      "loss": 1.0449,
+      "step": 7852
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990908571428571e-05,
+      "loss": 1.2995,
+      "step": 7853
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990640000000001e-05,
+      "loss": 1.8619,
+      "step": 7854
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199037142857143e-05,
+      "loss": 1.6265,
+      "step": 7855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990102857142857e-05,
+      "loss": 1.3833,
+      "step": 7856
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989834285714287e-05,
+      "loss": 1.6483,
+      "step": 7857
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989565714285714e-05,
+      "loss": 1.7493,
+      "step": 7858
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989297142857144e-05,
+      "loss": 0.7626,
+      "step": 7859
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989028571428572e-05,
+      "loss": 1.3767,
+      "step": 7860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198876e-05,
+      "loss": 1.3184,
+      "step": 7861
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1988491428571428e-05,
+      "loss": 1.5235,
+      "step": 7862
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1988222857142858e-05,
+      "loss": 1.2378,
+      "step": 7863
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987954285714286e-05,
+      "loss": 1.4901,
+      "step": 7864
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987685714285714e-05,
+      "loss": 1.6728,
+      "step": 7865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987417142857143e-05,
+      "loss": 1.1457,
+      "step": 7866
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987148571428572e-05,
+      "loss": 1.5597,
+      "step": 7867
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198688e-05,
+      "loss": 1.5231,
+      "step": 7868
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1986611428571429e-05,
+      "loss": 1.538,
+      "step": 7869
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1986342857142859e-05,
+      "loss": 1.1666,
+      "step": 7870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1986074285714285e-05,
+      "loss": 1.5353,
+      "step": 7871
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985805714285715e-05,
+      "loss": 1.3957,
+      "step": 7872
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985537142857143e-05,
+      "loss": 1.571,
+      "step": 7873
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985268571428573e-05,
+      "loss": 1.0186,
+      "step": 7874
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985e-05,
+      "loss": 1.185,
+      "step": 7875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198473142857143e-05,
+      "loss": 1.4762,
+      "step": 7876
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1984462857142858e-05,
+      "loss": 1.0916,
+      "step": 7877
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1984194285714286e-05,
+      "loss": 1.0486,
+      "step": 7878
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983925714285714e-05,
+      "loss": 1.083,
+      "step": 7879
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983657142857144e-05,
+      "loss": 1.549,
+      "step": 7880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983388571428572e-05,
+      "loss": 1.4108,
+      "step": 7881
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198312e-05,
+      "loss": 1.2151,
+      "step": 7882
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198285142857143e-05,
+      "loss": 1.4434,
+      "step": 7883
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1982582857142857e-05,
+      "loss": 1.2906,
+      "step": 7884
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1982314285714286e-05,
+      "loss": 1.4958,
+      "step": 7885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1982045714285715e-05,
+      "loss": 1.0105,
+      "step": 7886
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1981777142857144e-05,
+      "loss": 0.997,
+      "step": 7887
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1981508571428571e-05,
+      "loss": 1.4186,
+      "step": 7888
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198124e-05,
+      "loss": 1.0374,
+      "step": 7889
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980971428571429e-05,
+      "loss": 1.1157,
+      "step": 7890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980702857142857e-05,
+      "loss": 1.4765,
+      "step": 7891
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980434285714285e-05,
+      "loss": 1.2499,
+      "step": 7892
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980165714285715e-05,
+      "loss": 1.6608,
+      "step": 7893
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979897142857143e-05,
+      "loss": 1.4722,
+      "step": 7894
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979628571428572e-05,
+      "loss": 1.5456,
+      "step": 7895
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979360000000001e-05,
+      "loss": 1.4082,
+      "step": 7896
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979091428571428e-05,
+      "loss": 1.3487,
+      "step": 7897
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978822857142858e-05,
+      "loss": 1.3294,
+      "step": 7898
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978554285714286e-05,
+      "loss": 1.3646,
+      "step": 7899
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978285714285716e-05,
+      "loss": 1.5251,
+      "step": 7900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978017142857142e-05,
+      "loss": 1.6248,
+      "step": 7901
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1977748571428572e-05,
+      "loss": 1.2419,
+      "step": 7902
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197748e-05,
+      "loss": 1.2887,
+      "step": 7903
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1977211428571429e-05,
+      "loss": 1.6522,
+      "step": 7904
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976942857142857e-05,
+      "loss": 1.419,
+      "step": 7905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976674285714287e-05,
+      "loss": 1.0534,
+      "step": 7906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976405714285715e-05,
+      "loss": 1.5338,
+      "step": 7907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976137142857143e-05,
+      "loss": 1.2888,
+      "step": 7908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975868571428573e-05,
+      "loss": 0.9147,
+      "step": 7909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975600000000001e-05,
+      "loss": 1.4617,
+      "step": 7910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197533142857143e-05,
+      "loss": 1.1497,
+      "step": 7911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975062857142857e-05,
+      "loss": 1.139,
+      "step": 7912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1974794285714287e-05,
+      "loss": 1.3957,
+      "step": 7913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1974525714285714e-05,
+      "loss": 1.3032,
+      "step": 7914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1974257142857144e-05,
+      "loss": 1.3905,
+      "step": 7915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973988571428572e-05,
+      "loss": 1.2112,
+      "step": 7916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197372e-05,
+      "loss": 1.3697,
+      "step": 7917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973451428571428e-05,
+      "loss": 1.7576,
+      "step": 7918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973182857142858e-05,
+      "loss": 1.0867,
+      "step": 7919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972914285714286e-05,
+      "loss": 1.4813,
+      "step": 7920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972645714285714e-05,
+      "loss": 1.3401,
+      "step": 7921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972377142857144e-05,
+      "loss": 1.2508,
+      "step": 7922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972108571428572e-05,
+      "loss": 1.4694,
+      "step": 7923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197184e-05,
+      "loss": 1.59,
+      "step": 7924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971571428571429e-05,
+      "loss": 1.0334,
+      "step": 7925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971302857142859e-05,
+      "loss": 0.9028,
+      "step": 7926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971034285714285e-05,
+      "loss": 1.1132,
+      "step": 7927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970765714285715e-05,
+      "loss": 1.074,
+      "step": 7928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970497142857143e-05,
+      "loss": 1.0013,
+      "step": 7929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970228571428571e-05,
+      "loss": 1.1474,
+      "step": 7930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196996e-05,
+      "loss": 1.1254,
+      "step": 7931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196969142857143e-05,
+      "loss": 1.73,
+      "step": 7932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1969422857142858e-05,
+      "loss": 1.1303,
+      "step": 7933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1969154285714286e-05,
+      "loss": 1.025,
+      "step": 7934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968885714285716e-05,
+      "loss": 1.7156,
+      "step": 7935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968617142857144e-05,
+      "loss": 1.7894,
+      "step": 7936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968348571428572e-05,
+      "loss": 1.1105,
+      "step": 7937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196808e-05,
+      "loss": 1.6505,
+      "step": 7938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196781142857143e-05,
+      "loss": 1.2865,
+      "step": 7939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1967542857142856e-05,
+      "loss": 0.9872,
+      "step": 7940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1967274285714286e-05,
+      "loss": 1.3802,
+      "step": 7941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1967005714285715e-05,
+      "loss": 1.4495,
+      "step": 7942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966737142857143e-05,
+      "loss": 1.4801,
+      "step": 7943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966468571428571e-05,
+      "loss": 1.4573,
+      "step": 7944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.19662e-05,
+      "loss": 1.1255,
+      "step": 7945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1965931428571429e-05,
+      "loss": 0.984,
+      "step": 7946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1965662857142857e-05,
+      "loss": 1.4522,
+      "step": 7947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1965394285714287e-05,
+      "loss": 1.5043,
+      "step": 7948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1965125714285715e-05,
+      "loss": 1.1594,
+      "step": 7949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964857142857143e-05,
+      "loss": 1.7417,
+      "step": 7950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964588571428571e-05,
+      "loss": 1.3859,
+      "step": 7951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964320000000001e-05,
+      "loss": 1.0623,
+      "step": 7952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964051428571428e-05,
+      "loss": 1.0657,
+      "step": 7953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963782857142858e-05,
+      "loss": 1.0376,
+      "step": 7954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963514285714286e-05,
+      "loss": 1.2401,
+      "step": 7955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963245714285716e-05,
+      "loss": 1.3182,
+      "step": 7956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962977142857142e-05,
+      "loss": 1.6898,
+      "step": 7957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962708571428572e-05,
+      "loss": 1.1938,
+      "step": 7958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196244e-05,
+      "loss": 1.357,
+      "step": 7959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962171428571428e-05,
+      "loss": 1.1995,
+      "step": 7960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1961902857142858e-05,
+      "loss": 0.9578,
+      "step": 7961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1961634285714286e-05,
+      "loss": 1.6762,
+      "step": 7962
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1961365714285715e-05,
+      "loss": 1.765,
+      "step": 7963
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1961097142857143e-05,
+      "loss": 1.2392,
+      "step": 7964
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1960828571428573e-05,
+      "loss": 1.5405,
+      "step": 7965
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.196056e-05,
+      "loss": 1.3879,
+      "step": 7966
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1960291428571429e-05,
+      "loss": 1.268,
+      "step": 7967
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1960022857142857e-05,
+      "loss": 1.3681,
+      "step": 7968
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1959754285714287e-05,
+      "loss": 1.6215,
+      "step": 7969
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1959485714285714e-05,
+      "loss": 1.1244,
+      "step": 7970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1959217142857143e-05,
+      "loss": 1.442,
+      "step": 7971
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1958948571428572e-05,
+      "loss": 1.5921,
+      "step": 7972
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.195868e-05,
+      "loss": 1.3074,
+      "step": 7973
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.195841142857143e-05,
+      "loss": 1.6059,
+      "step": 7974
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1958142857142858e-05,
+      "loss": 1.2281,
+      "step": 7975
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1957874285714286e-05,
+      "loss": 1.162,
+      "step": 7976
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1957605714285714e-05,
+      "loss": 1.7299,
+      "step": 7977
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1957337142857144e-05,
+      "loss": 1.2953,
+      "step": 7978
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1957068571428572e-05,
+      "loss": 1.7784,
+      "step": 7979
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.19568e-05,
+      "loss": 0.8985,
+      "step": 7980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1956531428571429e-05,
+      "loss": 0.9937,
+      "step": 7981
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1956262857142858e-05,
+      "loss": 1.192,
+      "step": 7982
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1955994285714285e-05,
+      "loss": 0.9627,
+      "step": 7983
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1955725714285715e-05,
+      "loss": 1.5904,
+      "step": 7984
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1955457142857143e-05,
+      "loss": 1.129,
+      "step": 7985
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1955188571428571e-05,
+      "loss": 1.4771,
+      "step": 7986
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1954920000000001e-05,
+      "loss": 1.4203,
+      "step": 7987
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.195465142857143e-05,
+      "loss": 1.1424,
+      "step": 7988
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1954382857142857e-05,
+      "loss": 1.4236,
+      "step": 7989
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1954114285714286e-05,
+      "loss": 1.1843,
+      "step": 7990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1953845714285715e-05,
+      "loss": 1.4815,
+      "step": 7991
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1953577142857144e-05,
+      "loss": 1.3449,
+      "step": 7992
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1953308571428572e-05,
+      "loss": 1.2335,
+      "step": 7993
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.195304e-05,
+      "loss": 1.3674,
+      "step": 7994
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.195277142857143e-05,
+      "loss": 1.436,
+      "step": 7995
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1952502857142856e-05,
+      "loss": 1.3028,
+      "step": 7996
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1952234285714286e-05,
+      "loss": 1.1575,
+      "step": 7997
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1951965714285714e-05,
+      "loss": 1.2712,
+      "step": 7998
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1951697142857143e-05,
+      "loss": 1.3386,
+      "step": 7999
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1951428571428572e-05,
+      "loss": 1.843,
+      "step": 8000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.195116e-05,
+      "loss": 1.2059,
+      "step": 8001
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.195089142857143e-05,
+      "loss": 1.566,
+      "step": 8002
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1950622857142857e-05,
+      "loss": 0.8632,
+      "step": 8003
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1950354285714287e-05,
+      "loss": 1.4016,
+      "step": 8004
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1950085714285715e-05,
+      "loss": 1.5144,
+      "step": 8005
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1949817142857143e-05,
+      "loss": 1.0432,
+      "step": 8006
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1949548571428571e-05,
+      "loss": 1.7064,
+      "step": 8007
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1949280000000001e-05,
+      "loss": 1.1843,
+      "step": 8008
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1949011428571428e-05,
+      "loss": 1.2561,
+      "step": 8009
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1948742857142858e-05,
+      "loss": 1.3264,
+      "step": 8010
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1948474285714286e-05,
+      "loss": 1.2613,
+      "step": 8011
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1948205714285714e-05,
+      "loss": 1.3511,
+      "step": 8012
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1947937142857144e-05,
+      "loss": 1.4278,
+      "step": 8013
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1947668571428572e-05,
+      "loss": 0.9619,
+      "step": 8014
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1947400000000002e-05,
+      "loss": 1.3747,
+      "step": 8015
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1947131428571428e-05,
+      "loss": 1.0315,
+      "step": 8016
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1946862857142858e-05,
+      "loss": 0.9944,
+      "step": 8017
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1946594285714286e-05,
+      "loss": 1.3239,
+      "step": 8018
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1946325714285715e-05,
+      "loss": 1.0217,
+      "step": 8019
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1946057142857143e-05,
+      "loss": 1.4751,
+      "step": 8020
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1945788571428573e-05,
+      "loss": 1.1499,
+      "step": 8021
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1945519999999999e-05,
+      "loss": 1.6048,
+      "step": 8022
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1945251428571429e-05,
+      "loss": 1.4533,
+      "step": 8023
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1944982857142857e-05,
+      "loss": 1.1829,
+      "step": 8024
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1944714285714287e-05,
+      "loss": 1.1986,
+      "step": 8025
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1944445714285715e-05,
+      "loss": 2.1733,
+      "step": 8026
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1944177142857143e-05,
+      "loss": 1.4927,
+      "step": 8027
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1943908571428573e-05,
+      "loss": 1.2812,
+      "step": 8028
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.194364e-05,
+      "loss": 1.5499,
+      "step": 8029
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.194337142857143e-05,
+      "loss": 1.0889,
+      "step": 8030
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1943102857142858e-05,
+      "loss": 1.0552,
+      "step": 8031
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1942834285714286e-05,
+      "loss": 0.99,
+      "step": 8032
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1942565714285714e-05,
+      "loss": 1.1269,
+      "step": 8033
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1942297142857144e-05,
+      "loss": 0.9658,
+      "step": 8034
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.194202857142857e-05,
+      "loss": 0.9936,
+      "step": 8035
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.194176e-05,
+      "loss": 1.191,
+      "step": 8036
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1941491428571429e-05,
+      "loss": 1.6827,
+      "step": 8037
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1941222857142858e-05,
+      "loss": 1.1169,
+      "step": 8038
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1940954285714287e-05,
+      "loss": 1.2947,
+      "step": 8039
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1940685714285715e-05,
+      "loss": 1.655,
+      "step": 8040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1940417142857145e-05,
+      "loss": 1.179,
+      "step": 8041
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1940148571428571e-05,
+      "loss": 1.0769,
+      "step": 8042
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1939880000000001e-05,
+      "loss": 1.1808,
+      "step": 8043
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1939611428571429e-05,
+      "loss": 1.242,
+      "step": 8044
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1939342857142857e-05,
+      "loss": 1.0969,
+      "step": 8045
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1939074285714285e-05,
+      "loss": 1.6414,
+      "step": 8046
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1938805714285715e-05,
+      "loss": 1.1292,
+      "step": 8047
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1938537142857144e-05,
+      "loss": 1.5353,
+      "step": 8048
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1938268571428572e-05,
+      "loss": 1.098,
+      "step": 8049
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1938e-05,
+      "loss": 1.0333,
+      "step": 8050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.193773142857143e-05,
+      "loss": 1.1505,
+      "step": 8051
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1937462857142858e-05,
+      "loss": 0.9194,
+      "step": 8052
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1937194285714286e-05,
+      "loss": 1.543,
+      "step": 8053
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1936925714285716e-05,
+      "loss": 1.4321,
+      "step": 8054
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1936657142857142e-05,
+      "loss": 1.7831,
+      "step": 8055
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1936388571428572e-05,
+      "loss": 1.38,
+      "step": 8056
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.193612e-05,
+      "loss": 1.5206,
+      "step": 8057
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1935851428571429e-05,
+      "loss": 1.654,
+      "step": 8058
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1935582857142857e-05,
+      "loss": 1.2979,
+      "step": 8059
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1935314285714287e-05,
+      "loss": 1.2378,
+      "step": 8060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1935045714285715e-05,
+      "loss": 1.663,
+      "step": 8061
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1934777142857143e-05,
+      "loss": 1.44,
+      "step": 8062
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1934508571428571e-05,
+      "loss": 0.9835,
+      "step": 8063
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1934240000000001e-05,
+      "loss": 1.3127,
+      "step": 8064
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.193397142857143e-05,
+      "loss": 1.1678,
+      "step": 8065
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1933702857142857e-05,
+      "loss": 0.9933,
+      "step": 8066
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1933434285714287e-05,
+      "loss": 1.5581,
+      "step": 8067
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1933165714285714e-05,
+      "loss": 1.4583,
+      "step": 8068
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1932897142857144e-05,
+      "loss": 1.32,
+      "step": 8069
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1932628571428572e-05,
+      "loss": 1.4195,
+      "step": 8070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.193236e-05,
+      "loss": 1.1888,
+      "step": 8071
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1932091428571428e-05,
+      "loss": 1.1829,
+      "step": 8072
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1931822857142858e-05,
+      "loss": 1.5203,
+      "step": 8073
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1931554285714286e-05,
+      "loss": 1.2552,
+      "step": 8074
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1931285714285714e-05,
+      "loss": 1.0601,
+      "step": 8075
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1931017142857143e-05,
+      "loss": 0.9784,
+      "step": 8076
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1930748571428572e-05,
+      "loss": 1.0446,
+      "step": 8077
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.193048e-05,
+      "loss": 1.0792,
+      "step": 8078
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1930211428571429e-05,
+      "loss": 1.1474,
+      "step": 8079
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929942857142859e-05,
+      "loss": 0.9638,
+      "step": 8080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929674285714285e-05,
+      "loss": 1.5287,
+      "step": 8081
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929405714285715e-05,
+      "loss": 1.45,
+      "step": 8082
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929137142857143e-05,
+      "loss": 1.8768,
+      "step": 8083
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1928868571428573e-05,
+      "loss": 1.1335,
+      "step": 8084
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.19286e-05,
+      "loss": 1.3626,
+      "step": 8085
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192833142857143e-05,
+      "loss": 1.1369,
+      "step": 8086
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1928062857142858e-05,
+      "loss": 1.3843,
+      "step": 8087
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927794285714286e-05,
+      "loss": 1.5926,
+      "step": 8088
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927525714285714e-05,
+      "loss": 1.2434,
+      "step": 8089
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927257142857144e-05,
+      "loss": 1.0851,
+      "step": 8090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1926988571428572e-05,
+      "loss": 1.2751,
+      "step": 8091
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192672e-05,
+      "loss": 1.4665,
+      "step": 8092
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192645142857143e-05,
+      "loss": 1.4686,
+      "step": 8093
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1926182857142857e-05,
+      "loss": 1.3285,
+      "step": 8094
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925914285714286e-05,
+      "loss": 1.0459,
+      "step": 8095
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925645714285715e-05,
+      "loss": 0.9706,
+      "step": 8096
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925377142857144e-05,
+      "loss": 1.5734,
+      "step": 8097
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925108571428571e-05,
+      "loss": 1.3693,
+      "step": 8098
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192484e-05,
+      "loss": 1.5874,
+      "step": 8099
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924571428571429e-05,
+      "loss": 1.5378,
+      "step": 8100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924302857142857e-05,
+      "loss": 1.3764,
+      "step": 8101
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924034285714285e-05,
+      "loss": 1.27,
+      "step": 8102
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923765714285715e-05,
+      "loss": 1.1308,
+      "step": 8103
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923497142857143e-05,
+      "loss": 1.6796,
+      "step": 8104
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923228571428572e-05,
+      "loss": 1.0007,
+      "step": 8105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922960000000001e-05,
+      "loss": 1.1138,
+      "step": 8106
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192269142857143e-05,
+      "loss": 1.4358,
+      "step": 8107
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922422857142858e-05,
+      "loss": 0.9707,
+      "step": 8108
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922154285714286e-05,
+      "loss": 1.2795,
+      "step": 8109
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1921885714285716e-05,
+      "loss": 1.2614,
+      "step": 8110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1921617142857142e-05,
+      "loss": 0.9325,
+      "step": 8111
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1921348571428572e-05,
+      "loss": 1.355,
+      "step": 8112
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192108e-05,
+      "loss": 1.2681,
+      "step": 8113
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1920811428571429e-05,
+      "loss": 1.108,
+      "step": 8114
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1920542857142857e-05,
+      "loss": 1.3947,
+      "step": 8115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1920274285714287e-05,
+      "loss": 1.4582,
+      "step": 8116
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1920005714285715e-05,
+      "loss": 1.2413,
+      "step": 8117
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919737142857143e-05,
+      "loss": 1.4271,
+      "step": 8118
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919468571428573e-05,
+      "loss": 1.4576,
+      "step": 8119
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919200000000001e-05,
+      "loss": 1.3307,
+      "step": 8120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191893142857143e-05,
+      "loss": 1.2077,
+      "step": 8121
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918662857142857e-05,
+      "loss": 1.3807,
+      "step": 8122
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918394285714287e-05,
+      "loss": 1.6927,
+      "step": 8123
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918125714285714e-05,
+      "loss": 1.0137,
+      "step": 8124
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917857142857144e-05,
+      "loss": 1.4156,
+      "step": 8125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917588571428572e-05,
+      "loss": 1.3179,
+      "step": 8126
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191732e-05,
+      "loss": 1.3884,
+      "step": 8127
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917051428571428e-05,
+      "loss": 1.4292,
+      "step": 8128
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1916782857142858e-05,
+      "loss": 1.1165,
+      "step": 8129
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1916514285714286e-05,
+      "loss": 1.2335,
+      "step": 8130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1916245714285714e-05,
+      "loss": 1.3437,
+      "step": 8131
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1915977142857144e-05,
+      "loss": 1.1679,
+      "step": 8132
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1915708571428572e-05,
+      "loss": 1.4961,
+      "step": 8133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191544e-05,
+      "loss": 1.3023,
+      "step": 8134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1915171428571429e-05,
+      "loss": 1.059,
+      "step": 8135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1914902857142859e-05,
+      "loss": 0.8496,
+      "step": 8136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1914634285714285e-05,
+      "loss": 1.795,
+      "step": 8137
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1914365714285715e-05,
+      "loss": 1.2149,
+      "step": 8138
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1914097142857143e-05,
+      "loss": 1.289,
+      "step": 8139
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1913828571428571e-05,
+      "loss": 1.4959,
+      "step": 8140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.191356e-05,
+      "loss": 0.9435,
+      "step": 8141
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.191329142857143e-05,
+      "loss": 1.168,
+      "step": 8142
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1913022857142858e-05,
+      "loss": 1.3266,
+      "step": 8143
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1912754285714286e-05,
+      "loss": 0.9092,
+      "step": 8144
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1912485714285716e-05,
+      "loss": 1.6251,
+      "step": 8145
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1912217142857144e-05,
+      "loss": 1.0827,
+      "step": 8146
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1911948571428572e-05,
+      "loss": 1.2102,
+      "step": 8147
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.191168e-05,
+      "loss": 1.1368,
+      "step": 8148
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.191141142857143e-05,
+      "loss": 1.3073,
+      "step": 8149
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1911142857142856e-05,
+      "loss": 0.8596,
+      "step": 8150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1910874285714286e-05,
+      "loss": 1.4002,
+      "step": 8151
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1910605714285715e-05,
+      "loss": 1.4523,
+      "step": 8152
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1910337142857144e-05,
+      "loss": 1.1575,
+      "step": 8153
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1910068571428571e-05,
+      "loss": 1.3901,
+      "step": 8154
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.19098e-05,
+      "loss": 1.7431,
+      "step": 8155
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1909531428571429e-05,
+      "loss": 1.8055,
+      "step": 8156
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1909262857142857e-05,
+      "loss": 1.1121,
+      "step": 8157
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1908994285714287e-05,
+      "loss": 1.1274,
+      "step": 8158
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1908725714285715e-05,
+      "loss": 1.0161,
+      "step": 8159
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1908457142857143e-05,
+      "loss": 1.5143,
+      "step": 8160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1908188571428571e-05,
+      "loss": 1.3439,
+      "step": 8161
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1907920000000001e-05,
+      "loss": 0.9682,
+      "step": 8162
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1907651428571428e-05,
+      "loss": 1.1299,
+      "step": 8163
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1907382857142858e-05,
+      "loss": 1.0781,
+      "step": 8164
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1907114285714286e-05,
+      "loss": 1.2595,
+      "step": 8165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1906845714285716e-05,
+      "loss": 1.2969,
+      "step": 8166
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1906577142857142e-05,
+      "loss": 1.3112,
+      "step": 8167
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1906308571428572e-05,
+      "loss": 1.2797,
+      "step": 8168
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.190604e-05,
+      "loss": 1.4972,
+      "step": 8169
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1905771428571428e-05,
+      "loss": 1.8045,
+      "step": 8170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1905502857142858e-05,
+      "loss": 1.0804,
+      "step": 8171
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1905234285714286e-05,
+      "loss": 1.4213,
+      "step": 8172
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1904965714285715e-05,
+      "loss": 1.3586,
+      "step": 8173
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1904697142857143e-05,
+      "loss": 1.3282,
+      "step": 8174
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1904428571428573e-05,
+      "loss": 1.3382,
+      "step": 8175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1904160000000001e-05,
+      "loss": 1.2558,
+      "step": 8176
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1903891428571429e-05,
+      "loss": 1.2011,
+      "step": 8177
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1903622857142857e-05,
+      "loss": 1.169,
+      "step": 8178
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1903354285714287e-05,
+      "loss": 1.54,
+      "step": 8179
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1903085714285714e-05,
+      "loss": 1.3617,
+      "step": 8180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1902817142857143e-05,
+      "loss": 1.402,
+      "step": 8181
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1902548571428572e-05,
+      "loss": 1.232,
+      "step": 8182
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.190228e-05,
+      "loss": 1.3621,
+      "step": 8183
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.190201142857143e-05,
+      "loss": 1.3749,
+      "step": 8184
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1901742857142858e-05,
+      "loss": 1.0741,
+      "step": 8185
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1901474285714286e-05,
+      "loss": 1.341,
+      "step": 8186
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1901205714285714e-05,
+      "loss": 1.4055,
+      "step": 8187
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1900937142857144e-05,
+      "loss": 1.1185,
+      "step": 8188
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1900668571428572e-05,
+      "loss": 1.1757,
+      "step": 8189
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.19004e-05,
+      "loss": 1.2922,
+      "step": 8190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1900131428571429e-05,
+      "loss": 1.5978,
+      "step": 8191
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1899862857142858e-05,
+      "loss": 1.0458,
+      "step": 8192
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1899594285714285e-05,
+      "loss": 1.3054,
+      "step": 8193
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1899325714285715e-05,
+      "loss": 1.2387,
+      "step": 8194
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1899057142857143e-05,
+      "loss": 1.0558,
+      "step": 8195
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1898788571428571e-05,
+      "loss": 1.414,
+      "step": 8196
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1898520000000001e-05,
+      "loss": 1.7523,
+      "step": 8197
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.189825142857143e-05,
+      "loss": 0.8714,
+      "step": 8198
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1897982857142859e-05,
+      "loss": 1.505,
+      "step": 8199
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1897714285714286e-05,
+      "loss": 1.514,
+      "step": 8200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1897445714285715e-05,
+      "loss": 1.4488,
+      "step": 8201
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1897177142857144e-05,
+      "loss": 1.6196,
+      "step": 8202
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1896908571428572e-05,
+      "loss": 1.515,
+      "step": 8203
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.189664e-05,
+      "loss": 1.4669,
+      "step": 8204
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.189637142857143e-05,
+      "loss": 1.3507,
+      "step": 8205
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1896102857142856e-05,
+      "loss": 1.1239,
+      "step": 8206
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1895834285714286e-05,
+      "loss": 1.1024,
+      "step": 8207
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1895565714285714e-05,
+      "loss": 0.9504,
+      "step": 8208
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1895297142857143e-05,
+      "loss": 1.321,
+      "step": 8209
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1895028571428572e-05,
+      "loss": 1.4021,
+      "step": 8210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.189476e-05,
+      "loss": 1.6982,
+      "step": 8211
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.189449142857143e-05,
+      "loss": 1.5159,
+      "step": 8212
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1894222857142857e-05,
+      "loss": 1.5133,
+      "step": 8213
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1893954285714287e-05,
+      "loss": 0.952,
+      "step": 8214
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1893685714285715e-05,
+      "loss": 1.6163,
+      "step": 8215
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1893417142857143e-05,
+      "loss": 1.2104,
+      "step": 8216
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1893148571428571e-05,
+      "loss": 1.2107,
+      "step": 8217
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1892880000000001e-05,
+      "loss": 1.1374,
+      "step": 8218
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1892611428571428e-05,
+      "loss": 1.3234,
+      "step": 8219
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1892342857142858e-05,
+      "loss": 1.7976,
+      "step": 8220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1892074285714286e-05,
+      "loss": 1.0942,
+      "step": 8221
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1891805714285716e-05,
+      "loss": 1.0756,
+      "step": 8222
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1891537142857144e-05,
+      "loss": 1.3619,
+      "step": 8223
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1891268571428572e-05,
+      "loss": 1.1959,
+      "step": 8224
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1891000000000002e-05,
+      "loss": 1.3535,
+      "step": 8225
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1890731428571428e-05,
+      "loss": 1.3041,
+      "step": 8226
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1890462857142858e-05,
+      "loss": 1.473,
+      "step": 8227
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1890194285714286e-05,
+      "loss": 0.918,
+      "step": 8228
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1889925714285715e-05,
+      "loss": 1.5533,
+      "step": 8229
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1889657142857143e-05,
+      "loss": 1.4122,
+      "step": 8230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1889388571428573e-05,
+      "loss": 1.4976,
+      "step": 8231
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1889119999999999e-05,
+      "loss": 1.3317,
+      "step": 8232
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1888851428571429e-05,
+      "loss": 1.5574,
+      "step": 8233
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1888582857142857e-05,
+      "loss": 1.1521,
+      "step": 8234
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1888314285714287e-05,
+      "loss": 1.4332,
+      "step": 8235
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1888045714285715e-05,
+      "loss": 1.3757,
+      "step": 8236
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1887777142857143e-05,
+      "loss": 1.3895,
+      "step": 8237
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1887508571428573e-05,
+      "loss": 1.2,
+      "step": 8238
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.188724e-05,
+      "loss": 1.0517,
+      "step": 8239
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.188697142857143e-05,
+      "loss": 1.4271,
+      "step": 8240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1886702857142858e-05,
+      "loss": 1.2996,
+      "step": 8241
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1886434285714286e-05,
+      "loss": 1.163,
+      "step": 8242
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1886165714285714e-05,
+      "loss": 1.2701,
+      "step": 8243
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1885897142857144e-05,
+      "loss": 1.4877,
+      "step": 8244
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.188562857142857e-05,
+      "loss": 0.8342,
+      "step": 8245
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.188536e-05,
+      "loss": 1.2747,
+      "step": 8246
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1885091428571429e-05,
+      "loss": 0.7942,
+      "step": 8247
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1884822857142858e-05,
+      "loss": 1.0478,
+      "step": 8248
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1884554285714287e-05,
+      "loss": 1.217,
+      "step": 8249
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1884285714285715e-05,
+      "loss": 1.2228,
+      "step": 8250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1884017142857145e-05,
+      "loss": 1.1255,
+      "step": 8251
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1883748571428571e-05,
+      "loss": 1.1501,
+      "step": 8252
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1883480000000001e-05,
+      "loss": 1.4349,
+      "step": 8253
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1883211428571429e-05,
+      "loss": 1.4083,
+      "step": 8254
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1882942857142857e-05,
+      "loss": 1.1789,
+      "step": 8255
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1882674285714285e-05,
+      "loss": 1.2248,
+      "step": 8256
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1882405714285715e-05,
+      "loss": 1.7123,
+      "step": 8257
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1882137142857144e-05,
+      "loss": 1.1778,
+      "step": 8258
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1881868571428572e-05,
+      "loss": 1.1855,
+      "step": 8259
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.18816e-05,
+      "loss": 1.6068,
+      "step": 8260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.188133142857143e-05,
+      "loss": 1.4546,
+      "step": 8261
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1881062857142858e-05,
+      "loss": 1.225,
+      "step": 8262
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1880794285714286e-05,
+      "loss": 1.4201,
+      "step": 8263
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1880525714285716e-05,
+      "loss": 1.0787,
+      "step": 8264
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1880257142857142e-05,
+      "loss": 1.2202,
+      "step": 8265
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1879988571428572e-05,
+      "loss": 1.2684,
+      "step": 8266
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.187972e-05,
+      "loss": 1.3067,
+      "step": 8267
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1879451428571429e-05,
+      "loss": 1.5019,
+      "step": 8268
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1879182857142857e-05,
+      "loss": 1.2641,
+      "step": 8269
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1878914285714287e-05,
+      "loss": 1.1449,
+      "step": 8270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1878645714285715e-05,
+      "loss": 0.9977,
+      "step": 8271
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1878377142857143e-05,
+      "loss": 0.9509,
+      "step": 8272
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1878108571428571e-05,
+      "loss": 0.9822,
+      "step": 8273
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1877840000000001e-05,
+      "loss": 1.0376,
+      "step": 8274
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.187757142857143e-05,
+      "loss": 1.166,
+      "step": 8275
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1877302857142857e-05,
+      "loss": 1.2767,
+      "step": 8276
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1877034285714287e-05,
+      "loss": 1.3179,
+      "step": 8277
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1876765714285714e-05,
+      "loss": 1.2401,
+      "step": 8278
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1876497142857144e-05,
+      "loss": 1.3792,
+      "step": 8279
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1876228571428572e-05,
+      "loss": 1.5547,
+      "step": 8280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1875960000000002e-05,
+      "loss": 1.2987,
+      "step": 8281
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1875691428571428e-05,
+      "loss": 1.1767,
+      "step": 8282
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1875422857142858e-05,
+      "loss": 1.5482,
+      "step": 8283
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1875154285714286e-05,
+      "loss": 1.3788,
+      "step": 8284
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1874885714285714e-05,
+      "loss": 1.4566,
+      "step": 8285
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1874617142857143e-05,
+      "loss": 1.6716,
+      "step": 8286
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1874348571428572e-05,
+      "loss": 1.4311,
+      "step": 8287
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.187408e-05,
+      "loss": 1.267,
+      "step": 8288
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1873811428571429e-05,
+      "loss": 1.4516,
+      "step": 8289
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1873542857142859e-05,
+      "loss": 0.9717,
+      "step": 8290
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1873274285714285e-05,
+      "loss": 1.2728,
+      "step": 8291
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1873005714285715e-05,
+      "loss": 1.2116,
+      "step": 8292
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1872737142857143e-05,
+      "loss": 1.4285,
+      "step": 8293
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1872468571428573e-05,
+      "loss": 1.3384,
+      "step": 8294
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.18722e-05,
+      "loss": 0.9562,
+      "step": 8295
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.187193142857143e-05,
+      "loss": 1.1376,
+      "step": 8296
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1871662857142858e-05,
+      "loss": 1.0441,
+      "step": 8297
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1871394285714286e-05,
+      "loss": 1.2403,
+      "step": 8298
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1871125714285714e-05,
+      "loss": 1.1386,
+      "step": 8299
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1870857142857144e-05,
+      "loss": 0.9066,
+      "step": 8300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.187058857142857e-05,
+      "loss": 1.3373,
+      "step": 8301
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.187032e-05,
+      "loss": 1.2533,
+      "step": 8302
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.187005142857143e-05,
+      "loss": 1.2297,
+      "step": 8303
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1869782857142858e-05,
+      "loss": 1.3461,
+      "step": 8304
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1869514285714286e-05,
+      "loss": 1.5131,
+      "step": 8305
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1869245714285715e-05,
+      "loss": 1.1536,
+      "step": 8306
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1868977142857144e-05,
+      "loss": 1.2405,
+      "step": 8307
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1868708571428571e-05,
+      "loss": 1.2288,
+      "step": 8308
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.186844e-05,
+      "loss": 1.4003,
+      "step": 8309
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1868171428571429e-05,
+      "loss": 1.278,
+      "step": 8310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1867902857142857e-05,
+      "loss": 1.1363,
+      "step": 8311
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1867634285714285e-05,
+      "loss": 1.2296,
+      "step": 8312
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1867365714285715e-05,
+      "loss": 1.2902,
+      "step": 8313
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1867097142857142e-05,
+      "loss": 1.6778,
+      "step": 8314
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1866828571428572e-05,
+      "loss": 1.4533,
+      "step": 8315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1866560000000001e-05,
+      "loss": 1.2445,
+      "step": 8316
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.186629142857143e-05,
+      "loss": 1.3167,
+      "step": 8317
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1866022857142858e-05,
+      "loss": 1.0198,
+      "step": 8318
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1865754285714286e-05,
+      "loss": 1.0693,
+      "step": 8319
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1865485714285716e-05,
+      "loss": 0.9396,
+      "step": 8320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1865217142857142e-05,
+      "loss": 1.1666,
+      "step": 8321
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1864948571428572e-05,
+      "loss": 1.4672,
+      "step": 8322
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.186468e-05,
+      "loss": 1.1475,
+      "step": 8323
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1864411428571429e-05,
+      "loss": 1.5037,
+      "step": 8324
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1864142857142857e-05,
+      "loss": 1.6829,
+      "step": 8325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1863874285714287e-05,
+      "loss": 1.0756,
+      "step": 8326
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1863605714285715e-05,
+      "loss": 1.2732,
+      "step": 8327
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1863337142857143e-05,
+      "loss": 1.2438,
+      "step": 8328
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1863068571428573e-05,
+      "loss": 1.2026,
+      "step": 8329
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1862800000000001e-05,
+      "loss": 1.5281,
+      "step": 8330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.186253142857143e-05,
+      "loss": 1.5086,
+      "step": 8331
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1862262857142857e-05,
+      "loss": 1.2883,
+      "step": 8332
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1861994285714287e-05,
+      "loss": 1.5441,
+      "step": 8333
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1861725714285714e-05,
+      "loss": 1.1747,
+      "step": 8334
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1861457142857144e-05,
+      "loss": 1.397,
+      "step": 8335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1861188571428572e-05,
+      "loss": 1.0829,
+      "step": 8336
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.186092e-05,
+      "loss": 1.461,
+      "step": 8337
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1860651428571428e-05,
+      "loss": 1.2014,
+      "step": 8338
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1860382857142858e-05,
+      "loss": 1.3122,
+      "step": 8339
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1860114285714286e-05,
+      "loss": 1.3287,
+      "step": 8340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1859845714285714e-05,
+      "loss": 1.0966,
+      "step": 8341
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1859577142857144e-05,
+      "loss": 1.3424,
+      "step": 8342
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1859308571428572e-05,
+      "loss": 1.3679,
+      "step": 8343
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.185904e-05,
+      "loss": 1.0265,
+      "step": 8344
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1858771428571429e-05,
+      "loss": 1.2536,
+      "step": 8345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1858502857142859e-05,
+      "loss": 1.3034,
+      "step": 8346
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1858234285714285e-05,
+      "loss": 1.5815,
+      "step": 8347
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1857965714285715e-05,
+      "loss": 1.4666,
+      "step": 8348
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1857697142857143e-05,
+      "loss": 0.9434,
+      "step": 8349
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1857428571428573e-05,
+      "loss": 1.0927,
+      "step": 8350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.185716e-05,
+      "loss": 1.8487,
+      "step": 8351
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.185689142857143e-05,
+      "loss": 1.4407,
+      "step": 8352
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1856622857142858e-05,
+      "loss": 1.099,
+      "step": 8353
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1856354285714286e-05,
+      "loss": 1.4298,
+      "step": 8354
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1856085714285716e-05,
+      "loss": 1.682,
+      "step": 8355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1855817142857144e-05,
+      "loss": 1.31,
+      "step": 8356
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1855548571428572e-05,
+      "loss": 1.2791,
+      "step": 8357
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.185528e-05,
+      "loss": 1.512,
+      "step": 8358
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.185501142857143e-05,
+      "loss": 1.4435,
+      "step": 8359
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1854742857142856e-05,
+      "loss": 1.252,
+      "step": 8360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1854474285714286e-05,
+      "loss": 1.2316,
+      "step": 8361
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1854205714285715e-05,
+      "loss": 0.9855,
+      "step": 8362
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1853937142857144e-05,
+      "loss": 1.1217,
+      "step": 8363
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1853668571428571e-05,
+      "loss": 1.9429,
+      "step": 8364
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.18534e-05,
+      "loss": 1.4653,
+      "step": 8365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1853131428571429e-05,
+      "loss": 1.4412,
+      "step": 8366
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1852862857142857e-05,
+      "loss": 1.1659,
+      "step": 8367
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1852594285714287e-05,
+      "loss": 1.1207,
+      "step": 8368
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1852325714285715e-05,
+      "loss": 1.4456,
+      "step": 8369
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1852057142857143e-05,
+      "loss": 1.1842,
+      "step": 8370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1851788571428571e-05,
+      "loss": 1.4313,
+      "step": 8371
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1851520000000001e-05,
+      "loss": 1.5269,
+      "step": 8372
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.185125142857143e-05,
+      "loss": 1.6899,
+      "step": 8373
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1850982857142858e-05,
+      "loss": 1.1575,
+      "step": 8374
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1850714285714286e-05,
+      "loss": 1.2981,
+      "step": 8375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1850445714285716e-05,
+      "loss": 1.4074,
+      "step": 8376
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1850177142857142e-05,
+      "loss": 1.2916,
+      "step": 8377
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1849908571428572e-05,
+      "loss": 1.0652,
+      "step": 8378
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.184964e-05,
+      "loss": 1.0327,
+      "step": 8379
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1849371428571428e-05,
+      "loss": 1.0418,
+      "step": 8380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1849102857142858e-05,
+      "loss": 1.2153,
+      "step": 8381
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1848834285714286e-05,
+      "loss": 1.585,
+      "step": 8382
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1848565714285715e-05,
+      "loss": 1.0588,
+      "step": 8383
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1848297142857143e-05,
+      "loss": 1.0463,
+      "step": 8384
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1848028571428573e-05,
+      "loss": 1.6089,
+      "step": 8385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1847760000000001e-05,
+      "loss": 1.37,
+      "step": 8386
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1847491428571429e-05,
+      "loss": 1.4423,
+      "step": 8387
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1847222857142857e-05,
+      "loss": 1.3291,
+      "step": 8388
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1846954285714287e-05,
+      "loss": 1.1373,
+      "step": 8389
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1846685714285714e-05,
+      "loss": 1.2984,
+      "step": 8390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1846417142857143e-05,
+      "loss": 1.2682,
+      "step": 8391
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1846148571428572e-05,
+      "loss": 1.0002,
+      "step": 8392
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.184588e-05,
+      "loss": 1.3081,
+      "step": 8393
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.184561142857143e-05,
+      "loss": 1.4196,
+      "step": 8394
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1845342857142858e-05,
+      "loss": 1.4362,
+      "step": 8395
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1845074285714286e-05,
+      "loss": 1.4725,
+      "step": 8396
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1844805714285714e-05,
+      "loss": 1.0083,
+      "step": 8397
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1844537142857144e-05,
+      "loss": 1.0723,
+      "step": 8398
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1844268571428572e-05,
+      "loss": 1.8204,
+      "step": 8399
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1844e-05,
+      "loss": 0.9772,
+      "step": 8400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1843731428571429e-05,
+      "loss": 1.3374,
+      "step": 8401
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1843462857142858e-05,
+      "loss": 1.296,
+      "step": 8402
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1843194285714285e-05,
+      "loss": 1.6553,
+      "step": 8403
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1842925714285715e-05,
+      "loss": 1.2175,
+      "step": 8404
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1842657142857143e-05,
+      "loss": 1.4547,
+      "step": 8405
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1842388571428571e-05,
+      "loss": 1.0864,
+      "step": 8406
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.184212e-05,
+      "loss": 1.298,
+      "step": 8407
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.184185142857143e-05,
+      "loss": 1.1906,
+      "step": 8408
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1841582857142859e-05,
+      "loss": 1.322,
+      "step": 8409
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1841314285714286e-05,
+      "loss": 1.0958,
+      "step": 8410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1841045714285715e-05,
+      "loss": 1.2744,
+      "step": 8411
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1840777142857144e-05,
+      "loss": 1.4234,
+      "step": 8412
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1840508571428572e-05,
+      "loss": 1.4074,
+      "step": 8413
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.184024e-05,
+      "loss": 1.7197,
+      "step": 8414
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.183997142857143e-05,
+      "loss": 1.2933,
+      "step": 8415
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1839702857142856e-05,
+      "loss": 0.9227,
+      "step": 8416
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1839434285714286e-05,
+      "loss": 1.2492,
+      "step": 8417
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1839165714285714e-05,
+      "loss": 1.4404,
+      "step": 8418
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1838897142857143e-05,
+      "loss": 1.3528,
+      "step": 8419
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.183862857142857e-05,
+      "loss": 1.5687,
+      "step": 8420
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.183836e-05,
+      "loss": 1.3914,
+      "step": 8421
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.183809142857143e-05,
+      "loss": 1.3759,
+      "step": 8422
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1837822857142857e-05,
+      "loss": 1.3226,
+      "step": 8423
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1837554285714287e-05,
+      "loss": 1.3645,
+      "step": 8424
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1837285714285715e-05,
+      "loss": 1.339,
+      "step": 8425
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1837017142857143e-05,
+      "loss": 1.2936,
+      "step": 8426
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1836748571428571e-05,
+      "loss": 1.562,
+      "step": 8427
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1836480000000001e-05,
+      "loss": 1.2597,
+      "step": 8428
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1836211428571428e-05,
+      "loss": 1.1144,
+      "step": 8429
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1835942857142858e-05,
+      "loss": 1.3319,
+      "step": 8430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1835674285714286e-05,
+      "loss": 1.6425,
+      "step": 8431
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1835405714285716e-05,
+      "loss": 1.6933,
+      "step": 8432
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1835137142857142e-05,
+      "loss": 1.0608,
+      "step": 8433
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1834868571428572e-05,
+      "loss": 0.8754,
+      "step": 8434
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1834600000000002e-05,
+      "loss": 1.0722,
+      "step": 8435
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1834331428571428e-05,
+      "loss": 1.1108,
+      "step": 8436
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1834062857142858e-05,
+      "loss": 1.5277,
+      "step": 8437
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1833794285714286e-05,
+      "loss": 1.6937,
+      "step": 8438
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1833525714285715e-05,
+      "loss": 1.3993,
+      "step": 8439
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1833257142857143e-05,
+      "loss": 1.5471,
+      "step": 8440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1832988571428573e-05,
+      "loss": 1.5489,
+      "step": 8441
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1832719999999999e-05,
+      "loss": 1.6278,
+      "step": 8442
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1832451428571429e-05,
+      "loss": 1.0425,
+      "step": 8443
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1832182857142857e-05,
+      "loss": 1.9964,
+      "step": 8444
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1831914285714287e-05,
+      "loss": 1.3559,
+      "step": 8445
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1831645714285713e-05,
+      "loss": 0.7429,
+      "step": 8446
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1831377142857143e-05,
+      "loss": 1.3727,
+      "step": 8447
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1831108571428573e-05,
+      "loss": 1.0406,
+      "step": 8448
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.183084e-05,
+      "loss": 1.0262,
+      "step": 8449
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.183057142857143e-05,
+      "loss": 1.4848,
+      "step": 8450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1830302857142858e-05,
+      "loss": 1.4552,
+      "step": 8451
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1830034285714286e-05,
+      "loss": 1.9151,
+      "step": 8452
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1829765714285714e-05,
+      "loss": 1.478,
+      "step": 8453
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1829497142857144e-05,
+      "loss": 0.8945,
+      "step": 8454
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1829228571428572e-05,
+      "loss": 1.132,
+      "step": 8455
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.182896e-05,
+      "loss": 1.7951,
+      "step": 8456
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1828691428571429e-05,
+      "loss": 1.1291,
+      "step": 8457
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1828422857142858e-05,
+      "loss": 1.222,
+      "step": 8458
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1828154285714285e-05,
+      "loss": 0.9409,
+      "step": 8459
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1827885714285715e-05,
+      "loss": 1.1823,
+      "step": 8460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1827617142857145e-05,
+      "loss": 1.343,
+      "step": 8461
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1827348571428571e-05,
+      "loss": 1.5273,
+      "step": 8462
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1827080000000001e-05,
+      "loss": 1.5281,
+      "step": 8463
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1826811428571429e-05,
+      "loss": 1.4493,
+      "step": 8464
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1826542857142857e-05,
+      "loss": 1.4194,
+      "step": 8465
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1826274285714285e-05,
+      "loss": 0.9979,
+      "step": 8466
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1826005714285715e-05,
+      "loss": 1.5363,
+      "step": 8467
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1825737142857144e-05,
+      "loss": 1.1286,
+      "step": 8468
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1825468571428572e-05,
+      "loss": 1.1862,
+      "step": 8469
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.18252e-05,
+      "loss": 1.5039,
+      "step": 8470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.182493142857143e-05,
+      "loss": 0.9246,
+      "step": 8471
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1824662857142856e-05,
+      "loss": 1.026,
+      "step": 8472
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1824394285714286e-05,
+      "loss": 0.6919,
+      "step": 8473
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1824125714285716e-05,
+      "loss": 1.7091,
+      "step": 8474
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1823857142857142e-05,
+      "loss": 1.2412,
+      "step": 8475
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1823588571428572e-05,
+      "loss": 1.3986,
+      "step": 8476
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.182332e-05,
+      "loss": 1.0992,
+      "step": 8477
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.182305142857143e-05,
+      "loss": 0.8939,
+      "step": 8478
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1822782857142857e-05,
+      "loss": 1.7116,
+      "step": 8479
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1822514285714287e-05,
+      "loss": 1.0746,
+      "step": 8480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1822245714285715e-05,
+      "loss": 1.4474,
+      "step": 8481
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1821977142857143e-05,
+      "loss": 1.434,
+      "step": 8482
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1821708571428571e-05,
+      "loss": 1.2536,
+      "step": 8483
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1821440000000001e-05,
+      "loss": 1.379,
+      "step": 8484
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1821171428571428e-05,
+      "loss": 1.0893,
+      "step": 8485
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1820902857142857e-05,
+      "loss": 1.8247,
+      "step": 8486
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1820634285714287e-05,
+      "loss": 1.491,
+      "step": 8487
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1820365714285714e-05,
+      "loss": 1.555,
+      "step": 8488
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1820097142857144e-05,
+      "loss": 1.4942,
+      "step": 8489
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1819828571428572e-05,
+      "loss": 0.8728,
+      "step": 8490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1819560000000002e-05,
+      "loss": 1.214,
+      "step": 8491
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1819291428571428e-05,
+      "loss": 1.0802,
+      "step": 8492
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1819022857142858e-05,
+      "loss": 1.299,
+      "step": 8493
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1818754285714286e-05,
+      "loss": 1.3294,
+      "step": 8494
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1818485714285714e-05,
+      "loss": 1.46,
+      "step": 8495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1818217142857143e-05,
+      "loss": 1.3275,
+      "step": 8496
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1817948571428572e-05,
+      "loss": 1.2589,
+      "step": 8497
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1817679999999999e-05,
+      "loss": 1.6347,
+      "step": 8498
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1817411428571429e-05,
+      "loss": 1.0871,
+      "step": 8499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1817142857142859e-05,
+      "loss": 1.1189,
+      "step": 8500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1816874285714287e-05,
+      "loss": 1.1477,
+      "step": 8501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1816605714285715e-05,
+      "loss": 1.4348,
+      "step": 8502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1816337142857143e-05,
+      "loss": 1.208,
+      "step": 8503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1816068571428573e-05,
+      "loss": 0.9396,
+      "step": 8504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.18158e-05,
+      "loss": 1.0895,
+      "step": 8505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.181553142857143e-05,
+      "loss": 1.516,
+      "step": 8506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1815262857142858e-05,
+      "loss": 1.4029,
+      "step": 8507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1814994285714286e-05,
+      "loss": 1.2282,
+      "step": 8508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1814725714285714e-05,
+      "loss": 1.3391,
+      "step": 8509
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1814457142857144e-05,
+      "loss": 1.2452,
+      "step": 8510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.181418857142857e-05,
+      "loss": 1.5356,
+      "step": 8511
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.181392e-05,
+      "loss": 1.4233,
+      "step": 8512
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1813651428571428e-05,
+      "loss": 1.2212,
+      "step": 8513
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1813382857142858e-05,
+      "loss": 1.6091,
+      "step": 8514
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1813114285714286e-05,
+      "loss": 1.2071,
+      "step": 8515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1812845714285715e-05,
+      "loss": 1.1316,
+      "step": 8516
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1812577142857144e-05,
+      "loss": 1.8982,
+      "step": 8517
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1812308571428571e-05,
+      "loss": 1.2291,
+      "step": 8518
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.181204e-05,
+      "loss": 1.3412,
+      "step": 8519
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1811771428571429e-05,
+      "loss": 0.8032,
+      "step": 8520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1811502857142857e-05,
+      "loss": 1.3773,
+      "step": 8521
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1811234285714285e-05,
+      "loss": 1.5215,
+      "step": 8522
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1810965714285715e-05,
+      "loss": 1.4232,
+      "step": 8523
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1810697142857143e-05,
+      "loss": 1.5403,
+      "step": 8524
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1810428571428572e-05,
+      "loss": 1.1873,
+      "step": 8525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.181016e-05,
+      "loss": 1.4216,
+      "step": 8526
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.180989142857143e-05,
+      "loss": 1.5677,
+      "step": 8527
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1809622857142858e-05,
+      "loss": 1.1641,
+      "step": 8528
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1809354285714286e-05,
+      "loss": 1.0659,
+      "step": 8529
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1809085714285716e-05,
+      "loss": 1.6234,
+      "step": 8530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1808817142857142e-05,
+      "loss": 1.5261,
+      "step": 8531
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1808548571428572e-05,
+      "loss": 1.2785,
+      "step": 8532
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.180828e-05,
+      "loss": 1.1346,
+      "step": 8533
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1808011428571429e-05,
+      "loss": 1.0602,
+      "step": 8534
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1807742857142857e-05,
+      "loss": 1.3086,
+      "step": 8535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1807474285714287e-05,
+      "loss": 1.7552,
+      "step": 8536
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1807205714285715e-05,
+      "loss": 1.6556,
+      "step": 8537
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1806937142857143e-05,
+      "loss": 1.1902,
+      "step": 8538
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1806668571428571e-05,
+      "loss": 0.9321,
+      "step": 8539
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1806400000000001e-05,
+      "loss": 1.603,
+      "step": 8540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.180613142857143e-05,
+      "loss": 1.2691,
+      "step": 8541
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1805862857142857e-05,
+      "loss": 1.4224,
+      "step": 8542
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1805594285714287e-05,
+      "loss": 1.496,
+      "step": 8543
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1805325714285714e-05,
+      "loss": 1.5999,
+      "step": 8544
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1805057142857144e-05,
+      "loss": 1.1912,
+      "step": 8545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1804788571428572e-05,
+      "loss": 0.9895,
+      "step": 8546
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1804520000000002e-05,
+      "loss": 1.5882,
+      "step": 8547
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1804251428571428e-05,
+      "loss": 1.0061,
+      "step": 8548
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1803982857142858e-05,
+      "loss": 1.2323,
+      "step": 8549
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1803714285714286e-05,
+      "loss": 1.8276,
+      "step": 8550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1803445714285714e-05,
+      "loss": 1.2122,
+      "step": 8551
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1803177142857143e-05,
+      "loss": 1.3624,
+      "step": 8552
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1802908571428572e-05,
+      "loss": 1.7109,
+      "step": 8553
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.180264e-05,
+      "loss": 1.1189,
+      "step": 8554
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1802371428571429e-05,
+      "loss": 1.1645,
+      "step": 8555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1802102857142859e-05,
+      "loss": 1.0411,
+      "step": 8556
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1801834285714285e-05,
+      "loss": 1.1222,
+      "step": 8557
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1801565714285715e-05,
+      "loss": 1.0838,
+      "step": 8558
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1801297142857143e-05,
+      "loss": 1.1128,
+      "step": 8559
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1801028571428573e-05,
+      "loss": 1.2203,
+      "step": 8560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.180076e-05,
+      "loss": 1.3172,
+      "step": 8561
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.180049142857143e-05,
+      "loss": 1.2513,
+      "step": 8562
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1800222857142858e-05,
+      "loss": 1.3631,
+      "step": 8563
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1799954285714286e-05,
+      "loss": 1.2488,
+      "step": 8564
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1799685714285714e-05,
+      "loss": 1.4661,
+      "step": 8565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1799417142857144e-05,
+      "loss": 1.398,
+      "step": 8566
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1799148571428572e-05,
+      "loss": 0.9876,
+      "step": 8567
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.179888e-05,
+      "loss": 1.4444,
+      "step": 8568
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.179861142857143e-05,
+      "loss": 1.1833,
+      "step": 8569
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1798342857142856e-05,
+      "loss": 0.9048,
+      "step": 8570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1798074285714286e-05,
+      "loss": 1.3945,
+      "step": 8571
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1797805714285715e-05,
+      "loss": 1.1871,
+      "step": 8572
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1797537142857144e-05,
+      "loss": 1.3497,
+      "step": 8573
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1797268571428571e-05,
+      "loss": 1.6248,
+      "step": 8574
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1797e-05,
+      "loss": 1.3713,
+      "step": 8575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1796731428571429e-05,
+      "loss": 1.4827,
+      "step": 8576
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1796462857142857e-05,
+      "loss": 1.1719,
+      "step": 8577
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1796194285714285e-05,
+      "loss": 1.2704,
+      "step": 8578
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1795925714285715e-05,
+      "loss": 1.6802,
+      "step": 8579
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1795657142857143e-05,
+      "loss": 1.1135,
+      "step": 8580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1795388571428571e-05,
+      "loss": 1.6381,
+      "step": 8581
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1795120000000001e-05,
+      "loss": 1.0295,
+      "step": 8582
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.179485142857143e-05,
+      "loss": 1.4004,
+      "step": 8583
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1794582857142858e-05,
+      "loss": 1.251,
+      "step": 8584
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1794314285714286e-05,
+      "loss": 1.298,
+      "step": 8585
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1794045714285716e-05,
+      "loss": 1.3977,
+      "step": 8586
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1793777142857142e-05,
+      "loss": 1.2631,
+      "step": 8587
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1793508571428572e-05,
+      "loss": 1.2809,
+      "step": 8588
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.179324e-05,
+      "loss": 1.5598,
+      "step": 8589
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1792971428571428e-05,
+      "loss": 1.6053,
+      "step": 8590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1792702857142857e-05,
+      "loss": 1.3107,
+      "step": 8591
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1792434285714286e-05,
+      "loss": 0.9154,
+      "step": 8592
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1792165714285715e-05,
+      "loss": 1.2391,
+      "step": 8593
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1791897142857143e-05,
+      "loss": 0.9398,
+      "step": 8594
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1791628571428573e-05,
+      "loss": 1.1406,
+      "step": 8595
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1791360000000001e-05,
+      "loss": 1.2596,
+      "step": 8596
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1791091428571429e-05,
+      "loss": 1.6039,
+      "step": 8597
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1790822857142857e-05,
+      "loss": 1.3013,
+      "step": 8598
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1790554285714287e-05,
+      "loss": 0.8682,
+      "step": 8599
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1790285714285714e-05,
+      "loss": 1.5636,
+      "step": 8600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1790017142857143e-05,
+      "loss": 1.2055,
+      "step": 8601
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1789748571428572e-05,
+      "loss": 1.2802,
+      "step": 8602
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.178948e-05,
+      "loss": 1.5922,
+      "step": 8603
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1789211428571428e-05,
+      "loss": 1.1153,
+      "step": 8604
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1788942857142858e-05,
+      "loss": 1.5377,
+      "step": 8605
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1788674285714288e-05,
+      "loss": 0.6847,
+      "step": 8606
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1788405714285714e-05,
+      "loss": 1.1327,
+      "step": 8607
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1788137142857144e-05,
+      "loss": 1.5255,
+      "step": 8608
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1787868571428572e-05,
+      "loss": 1.198,
+      "step": 8609
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.17876e-05,
+      "loss": 1.2608,
+      "step": 8610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1787331428571429e-05,
+      "loss": 1.476,
+      "step": 8611
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1787062857142858e-05,
+      "loss": 1.2633,
+      "step": 8612
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1786794285714285e-05,
+      "loss": 1.1478,
+      "step": 8613
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1786525714285715e-05,
+      "loss": 1.1466,
+      "step": 8614
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1786257142857143e-05,
+      "loss": 1.6249,
+      "step": 8615
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1785988571428571e-05,
+      "loss": 1.5959,
+      "step": 8616
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.178572e-05,
+      "loss": 0.9049,
+      "step": 8617
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.178545142857143e-05,
+      "loss": 1.0433,
+      "step": 8618
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1785182857142857e-05,
+      "loss": 1.3806,
+      "step": 8619
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1784914285714286e-05,
+      "loss": 1.239,
+      "step": 8620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1784645714285715e-05,
+      "loss": 1.4853,
+      "step": 8621
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1784377142857144e-05,
+      "loss": 1.0602,
+      "step": 8622
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1784108571428572e-05,
+      "loss": 0.9955,
+      "step": 8623
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.178384e-05,
+      "loss": 1.1553,
+      "step": 8624
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.178357142857143e-05,
+      "loss": 1.1975,
+      "step": 8625
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1783302857142856e-05,
+      "loss": 1.2169,
+      "step": 8626
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1783034285714286e-05,
+      "loss": 1.4421,
+      "step": 8627
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1782765714285714e-05,
+      "loss": 1.4845,
+      "step": 8628
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1782497142857144e-05,
+      "loss": 1.3742,
+      "step": 8629
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.178222857142857e-05,
+      "loss": 1.1805,
+      "step": 8630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.178196e-05,
+      "loss": 1.4817,
+      "step": 8631
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1781691428571429e-05,
+      "loss": 1.2937,
+      "step": 8632
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1781422857142857e-05,
+      "loss": 1.3995,
+      "step": 8633
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1781154285714287e-05,
+      "loss": 1.1751,
+      "step": 8634
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1780885714285715e-05,
+      "loss": 1.3783,
+      "step": 8635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1780617142857143e-05,
+      "loss": 1.0813,
+      "step": 8636
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1780348571428571e-05,
+      "loss": 1.2364,
+      "step": 8637
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1780080000000001e-05,
+      "loss": 1.5874,
+      "step": 8638
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1779811428571428e-05,
+      "loss": 1.1484,
+      "step": 8639
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1779542857142858e-05,
+      "loss": 1.3837,
+      "step": 8640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1779274285714286e-05,
+      "loss": 1.2495,
+      "step": 8641
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1779005714285716e-05,
+      "loss": 1.2257,
+      "step": 8642
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1778737142857142e-05,
+      "loss": 1.0784,
+      "step": 8643
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1778468571428572e-05,
+      "loss": 1.4653,
+      "step": 8644
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.17782e-05,
+      "loss": 1.1829,
+      "step": 8645
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1777931428571428e-05,
+      "loss": 1.3933,
+      "step": 8646
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1777662857142858e-05,
+      "loss": 1.4652,
+      "step": 8647
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1777394285714286e-05,
+      "loss": 1.4206,
+      "step": 8648
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1777125714285715e-05,
+      "loss": 1.6902,
+      "step": 8649
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1776857142857143e-05,
+      "loss": 1.1902,
+      "step": 8650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1776588571428573e-05,
+      "loss": 1.2819,
+      "step": 8651
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.177632e-05,
+      "loss": 1.3656,
+      "step": 8652
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1776051428571429e-05,
+      "loss": 1.2257,
+      "step": 8653
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1775782857142857e-05,
+      "loss": 1.5264,
+      "step": 8654
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1775514285714287e-05,
+      "loss": 1.0902,
+      "step": 8655
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1775245714285713e-05,
+      "loss": 1.3081,
+      "step": 8656
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1774977142857143e-05,
+      "loss": 1.7096,
+      "step": 8657
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1774708571428572e-05,
+      "loss": 0.9804,
+      "step": 8658
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.177444e-05,
+      "loss": 0.9145,
+      "step": 8659
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.177417142857143e-05,
+      "loss": 1.4005,
+      "step": 8660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1773902857142858e-05,
+      "loss": 0.7828,
+      "step": 8661
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1773634285714286e-05,
+      "loss": 1.2795,
+      "step": 8662
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1773365714285714e-05,
+      "loss": 1.2179,
+      "step": 8663
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1773097142857144e-05,
+      "loss": 1.4464,
+      "step": 8664
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1772828571428572e-05,
+      "loss": 1.1806,
+      "step": 8665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.177256e-05,
+      "loss": 1.5805,
+      "step": 8666
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1772291428571429e-05,
+      "loss": 1.1064,
+      "step": 8667
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1772022857142858e-05,
+      "loss": 1.4859,
+      "step": 8668
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1771754285714285e-05,
+      "loss": 1.1601,
+      "step": 8669
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1771485714285715e-05,
+      "loss": 1.6661,
+      "step": 8670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1771217142857143e-05,
+      "loss": 1.2768,
+      "step": 8671
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1770948571428571e-05,
+      "loss": 1.6154,
+      "step": 8672
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1770680000000001e-05,
+      "loss": 1.3731,
+      "step": 8673
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1770411428571429e-05,
+      "loss": 1.3534,
+      "step": 8674
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1770142857142859e-05,
+      "loss": 1.0661,
+      "step": 8675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1769874285714285e-05,
+      "loss": 1.0577,
+      "step": 8676
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1769605714285715e-05,
+      "loss": 1.3814,
+      "step": 8677
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1769337142857144e-05,
+      "loss": 1.2961,
+      "step": 8678
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1769068571428572e-05,
+      "loss": 1.091,
+      "step": 8679
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.17688e-05,
+      "loss": 1.3947,
+      "step": 8680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.176853142857143e-05,
+      "loss": 1.6381,
+      "step": 8681
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1768262857142856e-05,
+      "loss": 1.4283,
+      "step": 8682
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1767994285714286e-05,
+      "loss": 1.3503,
+      "step": 8683
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1767725714285714e-05,
+      "loss": 1.0452,
+      "step": 8684
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1767457142857142e-05,
+      "loss": 1.3657,
+      "step": 8685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1767188571428572e-05,
+      "loss": 1.7132,
+      "step": 8686
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.176692e-05,
+      "loss": 1.3516,
+      "step": 8687
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.176665142857143e-05,
+      "loss": 1.5451,
+      "step": 8688
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1766382857142857e-05,
+      "loss": 1.2222,
+      "step": 8689
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1766114285714287e-05,
+      "loss": 1.5244,
+      "step": 8690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1765845714285715e-05,
+      "loss": 1.0518,
+      "step": 8691
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1765577142857143e-05,
+      "loss": 1.235,
+      "step": 8692
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1765308571428571e-05,
+      "loss": 1.2531,
+      "step": 8693
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1765040000000001e-05,
+      "loss": 1.5634,
+      "step": 8694
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1764771428571428e-05,
+      "loss": 1.5331,
+      "step": 8695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1764502857142857e-05,
+      "loss": 1.2123,
+      "step": 8696
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1764234285714286e-05,
+      "loss": 1.5204,
+      "step": 8697
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1763965714285716e-05,
+      "loss": 1.1916,
+      "step": 8698
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1763697142857144e-05,
+      "loss": 1.3632,
+      "step": 8699
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1763428571428572e-05,
+      "loss": 1.4393,
+      "step": 8700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1763160000000002e-05,
+      "loss": 1.0607,
+      "step": 8701
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1762891428571428e-05,
+      "loss": 1.2841,
+      "step": 8702
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1762622857142858e-05,
+      "loss": 1.1658,
+      "step": 8703
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1762354285714286e-05,
+      "loss": 1.4011,
+      "step": 8704
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1762085714285714e-05,
+      "loss": 1.1547,
+      "step": 8705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1761817142857143e-05,
+      "loss": 1.319,
+      "step": 8706
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1761548571428572e-05,
+      "loss": 1.0889,
+      "step": 8707
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1761279999999999e-05,
+      "loss": 1.0344,
+      "step": 8708
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1761011428571429e-05,
+      "loss": 1.2759,
+      "step": 8709
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1760742857142857e-05,
+      "loss": 1.2778,
+      "step": 8710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1760474285714287e-05,
+      "loss": 1.4693,
+      "step": 8711
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1760205714285715e-05,
+      "loss": 1.508,
+      "step": 8712
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1759937142857143e-05,
+      "loss": 1.4659,
+      "step": 8713
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1759668571428573e-05,
+      "loss": 1.302,
+      "step": 8714
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.17594e-05,
+      "loss": 1.4567,
+      "step": 8715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.175913142857143e-05,
+      "loss": 0.859,
+      "step": 8716
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1758862857142858e-05,
+      "loss": 1.2429,
+      "step": 8717
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1758594285714286e-05,
+      "loss": 1.3289,
+      "step": 8718
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1758325714285714e-05,
+      "loss": 1.4201,
+      "step": 8719
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1758057142857144e-05,
+      "loss": 1.423,
+      "step": 8720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1757788571428572e-05,
+      "loss": 1.3795,
+      "step": 8721
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.175752e-05,
+      "loss": 1.4259,
+      "step": 8722
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1757251428571428e-05,
+      "loss": 1.3177,
+      "step": 8723
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1756982857142858e-05,
+      "loss": 1.112,
+      "step": 8724
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1756714285714286e-05,
+      "loss": 1.0192,
+      "step": 8725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1756445714285715e-05,
+      "loss": 1.4317,
+      "step": 8726
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1756177142857144e-05,
+      "loss": 1.069,
+      "step": 8727
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1755908571428571e-05,
+      "loss": 1.2252,
+      "step": 8728
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.175564e-05,
+      "loss": 1.0514,
+      "step": 8729
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1755371428571429e-05,
+      "loss": 1.464,
+      "step": 8730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1755102857142857e-05,
+      "loss": 1.3449,
+      "step": 8731
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1754834285714285e-05,
+      "loss": 1.2469,
+      "step": 8732
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1754565714285715e-05,
+      "loss": 1.7249,
+      "step": 8733
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1754297142857143e-05,
+      "loss": 1.383,
+      "step": 8734
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1754028571428572e-05,
+      "loss": 1.0471,
+      "step": 8735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.175376e-05,
+      "loss": 1.1048,
+      "step": 8736
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.175349142857143e-05,
+      "loss": 1.5163,
+      "step": 8737
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1753222857142858e-05,
+      "loss": 1.0972,
+      "step": 8738
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1752954285714286e-05,
+      "loss": 1.3847,
+      "step": 8739
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1752685714285716e-05,
+      "loss": 1.6173,
+      "step": 8740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1752417142857142e-05,
+      "loss": 1.4725,
+      "step": 8741
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1752148571428572e-05,
+      "loss": 1.3206,
+      "step": 8742
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.175188e-05,
+      "loss": 1.4088,
+      "step": 8743
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1751611428571429e-05,
+      "loss": 1.406,
+      "step": 8744
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1751342857142857e-05,
+      "loss": 1.1295,
+      "step": 8745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1751074285714287e-05,
+      "loss": 1.4209,
+      "step": 8746
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1750805714285715e-05,
+      "loss": 1.3659,
+      "step": 8747
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1750537142857143e-05,
+      "loss": 1.2852,
+      "step": 8748
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1750268571428571e-05,
+      "loss": 1.2003,
+      "step": 8749
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1750000000000001e-05,
+      "loss": 1.683,
+      "step": 8750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.174973142857143e-05,
+      "loss": 1.3967,
+      "step": 8751
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1749462857142857e-05,
+      "loss": 1.0027,
+      "step": 8752
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1749194285714287e-05,
+      "loss": 1.8121,
+      "step": 8753
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1748925714285714e-05,
+      "loss": 1.0932,
+      "step": 8754
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1748657142857144e-05,
+      "loss": 1.3435,
+      "step": 8755
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1748388571428572e-05,
+      "loss": 1.767,
+      "step": 8756
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1748120000000002e-05,
+      "loss": 1.5007,
+      "step": 8757
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1747851428571428e-05,
+      "loss": 1.1471,
+      "step": 8758
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1747582857142858e-05,
+      "loss": 1.0606,
+      "step": 8759
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1747314285714286e-05,
+      "loss": 1.3136,
+      "step": 8760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1747045714285714e-05,
+      "loss": 1.3669,
+      "step": 8761
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1746777142857143e-05,
+      "loss": 1.6709,
+      "step": 8762
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1746508571428572e-05,
+      "loss": 1.4314,
+      "step": 8763
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.174624e-05,
+      "loss": 1.5466,
+      "step": 8764
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1745971428571429e-05,
+      "loss": 1.0671,
+      "step": 8765
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1745702857142859e-05,
+      "loss": 1.204,
+      "step": 8766
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1745434285714285e-05,
+      "loss": 1.0275,
+      "step": 8767
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1745165714285715e-05,
+      "loss": 1.3464,
+      "step": 8768
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1744897142857143e-05,
+      "loss": 1.7878,
+      "step": 8769
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1744628571428573e-05,
+      "loss": 1.2096,
+      "step": 8770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.174436e-05,
+      "loss": 1.1241,
+      "step": 8771
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.174409142857143e-05,
+      "loss": 1.0233,
+      "step": 8772
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1743822857142858e-05,
+      "loss": 1.5536,
+      "step": 8773
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1743554285714286e-05,
+      "loss": 1.0802,
+      "step": 8774
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1743285714285714e-05,
+      "loss": 1.1781,
+      "step": 8775
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1743017142857144e-05,
+      "loss": 1.1446,
+      "step": 8776
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1742748571428572e-05,
+      "loss": 1.4267,
+      "step": 8777
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.174248e-05,
+      "loss": 1.7983,
+      "step": 8778
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.174221142857143e-05,
+      "loss": 1.0333,
+      "step": 8779
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1741942857142858e-05,
+      "loss": 1.2521,
+      "step": 8780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1741674285714286e-05,
+      "loss": 1.1092,
+      "step": 8781
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1741405714285714e-05,
+      "loss": 1.1081,
+      "step": 8782
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1741137142857144e-05,
+      "loss": 0.5824,
+      "step": 8783
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1740868571428571e-05,
+      "loss": 1.4834,
+      "step": 8784
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.17406e-05,
+      "loss": 1.6717,
+      "step": 8785
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1740331428571429e-05,
+      "loss": 1.5568,
+      "step": 8786
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1740062857142857e-05,
+      "loss": 1.2773,
+      "step": 8787
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1739794285714285e-05,
+      "loss": 1.4248,
+      "step": 8788
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1739525714285715e-05,
+      "loss": 1.035,
+      "step": 8789
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1739257142857143e-05,
+      "loss": 1.6306,
+      "step": 8790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1738988571428571e-05,
+      "loss": 1.658,
+      "step": 8791
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1738720000000001e-05,
+      "loss": 1.079,
+      "step": 8792
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.173845142857143e-05,
+      "loss": 1.6962,
+      "step": 8793
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1738182857142858e-05,
+      "loss": 1.0862,
+      "step": 8794
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1737914285714286e-05,
+      "loss": 1.0141,
+      "step": 8795
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1737645714285716e-05,
+      "loss": 1.578,
+      "step": 8796
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1737377142857142e-05,
+      "loss": 1.2992,
+      "step": 8797
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1737108571428572e-05,
+      "loss": 1.5201,
+      "step": 8798
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.173684e-05,
+      "loss": 1.3292,
+      "step": 8799
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1736571428571428e-05,
+      "loss": 1.0349,
+      "step": 8800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1736302857142857e-05,
+      "loss": 1.2468,
+      "step": 8801
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1736034285714286e-05,
+      "loss": 1.2753,
+      "step": 8802
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1735765714285715e-05,
+      "loss": 0.9048,
+      "step": 8803
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1735497142857143e-05,
+      "loss": 1.5306,
+      "step": 8804
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1735228571428573e-05,
+      "loss": 1.0866,
+      "step": 8805
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1734960000000001e-05,
+      "loss": 1.1649,
+      "step": 8806
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1734691428571429e-05,
+      "loss": 1.1749,
+      "step": 8807
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1734422857142857e-05,
+      "loss": 1.2201,
+      "step": 8808
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1734154285714287e-05,
+      "loss": 1.2018,
+      "step": 8809
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1733885714285714e-05,
+      "loss": 1.015,
+      "step": 8810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1733617142857143e-05,
+      "loss": 1.0521,
+      "step": 8811
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1733348571428572e-05,
+      "loss": 1.0495,
+      "step": 8812
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.173308e-05,
+      "loss": 1.0197,
+      "step": 8813
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1732811428571428e-05,
+      "loss": 1.2798,
+      "step": 8814
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1732542857142858e-05,
+      "loss": 1.5735,
+      "step": 8815
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1732274285714286e-05,
+      "loss": 1.2062,
+      "step": 8816
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1732005714285714e-05,
+      "loss": 1.3985,
+      "step": 8817
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1731737142857144e-05,
+      "loss": 1.4648,
+      "step": 8818
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1731468571428572e-05,
+      "loss": 1.3133,
+      "step": 8819
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.17312e-05,
+      "loss": 1.5522,
+      "step": 8820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1730931428571429e-05,
+      "loss": 1.3017,
+      "step": 8821
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1730662857142858e-05,
+      "loss": 1.7904,
+      "step": 8822
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1730394285714285e-05,
+      "loss": 1.3463,
+      "step": 8823
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1730125714285715e-05,
+      "loss": 0.7158,
+      "step": 8824
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1729857142857143e-05,
+      "loss": 1.3679,
+      "step": 8825
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1729588571428573e-05,
+      "loss": 1.1951,
+      "step": 8826
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.172932e-05,
+      "loss": 1.436,
+      "step": 8827
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.172905142857143e-05,
+      "loss": 1.2883,
+      "step": 8828
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1728782857142857e-05,
+      "loss": 1.2568,
+      "step": 8829
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1728514285714286e-05,
+      "loss": 1.486,
+      "step": 8830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1728245714285715e-05,
+      "loss": 1.0912,
+      "step": 8831
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1727977142857144e-05,
+      "loss": 0.8288,
+      "step": 8832
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1727708571428572e-05,
+      "loss": 1.361,
+      "step": 8833
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.172744e-05,
+      "loss": 1.3271,
+      "step": 8834
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.172717142857143e-05,
+      "loss": 1.3479,
+      "step": 8835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1726902857142856e-05,
+      "loss": 1.2778,
+      "step": 8836
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1726634285714286e-05,
+      "loss": 1.5994,
+      "step": 8837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1726365714285714e-05,
+      "loss": 1.6757,
+      "step": 8838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1726097142857144e-05,
+      "loss": 1.0869,
+      "step": 8839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.172582857142857e-05,
+      "loss": 1.1537,
+      "step": 8840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.172556e-05,
+      "loss": 1.2978,
+      "step": 8841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1725291428571429e-05,
+      "loss": 0.911,
+      "step": 8842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1725022857142857e-05,
+      "loss": 1.0671,
+      "step": 8843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1724754285714287e-05,
+      "loss": 1.5493,
+      "step": 8844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1724485714285715e-05,
+      "loss": 1.3246,
+      "step": 8845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1724217142857143e-05,
+      "loss": 1.0716,
+      "step": 8846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1723948571428571e-05,
+      "loss": 1.2129,
+      "step": 8847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1723680000000001e-05,
+      "loss": 1.0744,
+      "step": 8848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.172341142857143e-05,
+      "loss": 1.1944,
+      "step": 8849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1723142857142858e-05,
+      "loss": 1.4666,
+      "step": 8850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1722874285714286e-05,
+      "loss": 1.3741,
+      "step": 8851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1722605714285716e-05,
+      "loss": 1.8786,
+      "step": 8852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1722337142857142e-05,
+      "loss": 1.491,
+      "step": 8853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1722068571428572e-05,
+      "loss": 0.8528,
+      "step": 8854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.17218e-05,
+      "loss": 1.2783,
+      "step": 8855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1721531428571428e-05,
+      "loss": 1.1384,
+      "step": 8856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1721262857142858e-05,
+      "loss": 1.0452,
+      "step": 8857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1720994285714286e-05,
+      "loss": 0.9309,
+      "step": 8858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1720725714285715e-05,
+      "loss": 0.7908,
+      "step": 8859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1720457142857143e-05,
+      "loss": 1.8811,
+      "step": 8860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1720188571428573e-05,
+      "loss": 1.2446,
+      "step": 8861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.171992e-05,
+      "loss": 1.3889,
+      "step": 8862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1719651428571429e-05,
+      "loss": 1.4564,
+      "step": 8863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1719382857142857e-05,
+      "loss": 1.0048,
+      "step": 8864
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1719114285714287e-05,
+      "loss": 1.2193,
+      "step": 8865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1718845714285713e-05,
+      "loss": 0.8983,
+      "step": 8866
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1718577142857143e-05,
+      "loss": 1.2072,
+      "step": 8867
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1718308571428572e-05,
+      "loss": 1.1462,
+      "step": 8868
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.171804e-05,
+      "loss": 1.3213,
+      "step": 8869
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.171777142857143e-05,
+      "loss": 1.2742,
+      "step": 8870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1717502857142858e-05,
+      "loss": 1.2166,
+      "step": 8871
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1717234285714288e-05,
+      "loss": 1.3513,
+      "step": 8872
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1716965714285714e-05,
+      "loss": 1.3698,
+      "step": 8873
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1716697142857144e-05,
+      "loss": 1.2663,
+      "step": 8874
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1716428571428572e-05,
+      "loss": 1.3887,
+      "step": 8875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.171616e-05,
+      "loss": 1.587,
+      "step": 8876
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1715891428571429e-05,
+      "loss": 1.2204,
+      "step": 8877
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1715622857142858e-05,
+      "loss": 1.0594,
+      "step": 8878
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1715354285714285e-05,
+      "loss": 1.4024,
+      "step": 8879
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1715085714285715e-05,
+      "loss": 1.2459,
+      "step": 8880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1714817142857143e-05,
+      "loss": 1.4637,
+      "step": 8881
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1714548571428571e-05,
+      "loss": 1.1273,
+      "step": 8882
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1714280000000001e-05,
+      "loss": 1.4297,
+      "step": 8883
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1714011428571429e-05,
+      "loss": 1.5248,
+      "step": 8884
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1713742857142859e-05,
+      "loss": 1.3822,
+      "step": 8885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1713474285714285e-05,
+      "loss": 1.1638,
+      "step": 8886
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1713205714285715e-05,
+      "loss": 0.6479,
+      "step": 8887
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1712937142857144e-05,
+      "loss": 1.0311,
+      "step": 8888
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1712668571428572e-05,
+      "loss": 1.4167,
+      "step": 8889
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.17124e-05,
+      "loss": 1.2716,
+      "step": 8890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.171213142857143e-05,
+      "loss": 1.0663,
+      "step": 8891
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1711862857142856e-05,
+      "loss": 1.4324,
+      "step": 8892
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1711594285714286e-05,
+      "loss": 1.2434,
+      "step": 8893
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1711325714285714e-05,
+      "loss": 1.2976,
+      "step": 8894
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1711057142857142e-05,
+      "loss": 1.3711,
+      "step": 8895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1710788571428572e-05,
+      "loss": 1.4661,
+      "step": 8896
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.171052e-05,
+      "loss": 1.5399,
+      "step": 8897
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.171025142857143e-05,
+      "loss": 1.5303,
+      "step": 8898
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1709982857142857e-05,
+      "loss": 1.1442,
+      "step": 8899
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1709714285714287e-05,
+      "loss": 1.0189,
+      "step": 8900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1709445714285715e-05,
+      "loss": 1.3801,
+      "step": 8901
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1709177142857143e-05,
+      "loss": 1.2298,
+      "step": 8902
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1708908571428571e-05,
+      "loss": 1.5729,
+      "step": 8903
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1708640000000001e-05,
+      "loss": 1.3384,
+      "step": 8904
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1708371428571428e-05,
+      "loss": 1.5473,
+      "step": 8905
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1708102857142857e-05,
+      "loss": 1.0849,
+      "step": 8906
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1707834285714286e-05,
+      "loss": 1.7398,
+      "step": 8907
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1707565714285716e-05,
+      "loss": 1.5219,
+      "step": 8908
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1707297142857144e-05,
+      "loss": 1.1776,
+      "step": 8909
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1707028571428572e-05,
+      "loss": 1.2048,
+      "step": 8910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1706760000000002e-05,
+      "loss": 1.4086,
+      "step": 8911
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1706491428571428e-05,
+      "loss": 1.2288,
+      "step": 8912
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1706222857142858e-05,
+      "loss": 1.1901,
+      "step": 8913
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1705954285714286e-05,
+      "loss": 1.0692,
+      "step": 8914
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1705685714285714e-05,
+      "loss": 1.1387,
+      "step": 8915
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1705417142857143e-05,
+      "loss": 1.5807,
+      "step": 8916
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1705148571428572e-05,
+      "loss": 1.0598,
+      "step": 8917
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1704879999999999e-05,
+      "loss": 1.2228,
+      "step": 8918
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1704611428571429e-05,
+      "loss": 1.3884,
+      "step": 8919
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1704342857142857e-05,
+      "loss": 1.0537,
+      "step": 8920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1704074285714287e-05,
+      "loss": 1.4823,
+      "step": 8921
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1703805714285715e-05,
+      "loss": 1.3396,
+      "step": 8922
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1703537142857143e-05,
+      "loss": 1.2223,
+      "step": 8923
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1703268571428573e-05,
+      "loss": 1.1878,
+      "step": 8924
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1703e-05,
+      "loss": 1.548,
+      "step": 8925
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.170273142857143e-05,
+      "loss": 1.6437,
+      "step": 8926
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1702462857142858e-05,
+      "loss": 1.2415,
+      "step": 8927
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1702194285714286e-05,
+      "loss": 1.0269,
+      "step": 8928
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1701925714285714e-05,
+      "loss": 1.0762,
+      "step": 8929
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1701657142857144e-05,
+      "loss": 1.0664,
+      "step": 8930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1701388571428572e-05,
+      "loss": 1.1584,
+      "step": 8931
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.170112e-05,
+      "loss": 1.7354,
+      "step": 8932
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1700851428571428e-05,
+      "loss": 1.2775,
+      "step": 8933
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1700582857142858e-05,
+      "loss": 1.2722,
+      "step": 8934
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1700314285714286e-05,
+      "loss": 0.8799,
+      "step": 8935
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1700045714285715e-05,
+      "loss": 1.2638,
+      "step": 8936
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1699777142857144e-05,
+      "loss": 1.2116,
+      "step": 8937
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1699508571428571e-05,
+      "loss": 0.7829,
+      "step": 8938
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.169924e-05,
+      "loss": 1.2581,
+      "step": 8939
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1698971428571429e-05,
+      "loss": 1.2161,
+      "step": 8940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1698702857142857e-05,
+      "loss": 1.5309,
+      "step": 8941
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1698434285714285e-05,
+      "loss": 1.1487,
+      "step": 8942
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1698165714285715e-05,
+      "loss": 1.8238,
+      "step": 8943
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1697897142857143e-05,
+      "loss": 1.164,
+      "step": 8944
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1697628571428572e-05,
+      "loss": 1.1201,
+      "step": 8945
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.169736e-05,
+      "loss": 1.2229,
+      "step": 8946
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.169709142857143e-05,
+      "loss": 1.6033,
+      "step": 8947
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1696822857142858e-05,
+      "loss": 1.3264,
+      "step": 8948
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1696554285714286e-05,
+      "loss": 1.0849,
+      "step": 8949
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1696285714285716e-05,
+      "loss": 1.0132,
+      "step": 8950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1696017142857142e-05,
+      "loss": 1.2534,
+      "step": 8951
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1695748571428572e-05,
+      "loss": 1.407,
+      "step": 8952
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.169548e-05,
+      "loss": 1.2837,
+      "step": 8953
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.169521142857143e-05,
+      "loss": 1.5471,
+      "step": 8954
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1694942857142857e-05,
+      "loss": 1.1589,
+      "step": 8955
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1694674285714287e-05,
+      "loss": 1.3383,
+      "step": 8956
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1694405714285715e-05,
+      "loss": 1.1935,
+      "step": 8957
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1694137142857143e-05,
+      "loss": 1.2425,
+      "step": 8958
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1693868571428571e-05,
+      "loss": 1.6184,
+      "step": 8959
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1693600000000001e-05,
+      "loss": 1.4737,
+      "step": 8960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.169333142857143e-05,
+      "loss": 1.1616,
+      "step": 8961
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1693062857142857e-05,
+      "loss": 1.0839,
+      "step": 8962
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1692794285714287e-05,
+      "loss": 1.3053,
+      "step": 8963
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1692525714285714e-05,
+      "loss": 1.33,
+      "step": 8964
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1692257142857144e-05,
+      "loss": 0.9322,
+      "step": 8965
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1691988571428572e-05,
+      "loss": 1.5746,
+      "step": 8966
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1691720000000002e-05,
+      "loss": 1.4042,
+      "step": 8967
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1691451428571428e-05,
+      "loss": 1.2348,
+      "step": 8968
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1691182857142858e-05,
+      "loss": 1.5657,
+      "step": 8969
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1690914285714286e-05,
+      "loss": 1.3361,
+      "step": 8970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1690645714285714e-05,
+      "loss": 1.4572,
+      "step": 8971
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1690377142857143e-05,
+      "loss": 1.3368,
+      "step": 8972
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1690108571428572e-05,
+      "loss": 1.32,
+      "step": 8973
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.168984e-05,
+      "loss": 1.4992,
+      "step": 8974
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1689571428571429e-05,
+      "loss": 1.2238,
+      "step": 8975
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1689302857142859e-05,
+      "loss": 1.0948,
+      "step": 8976
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1689034285714287e-05,
+      "loss": 1.5426,
+      "step": 8977
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1688765714285715e-05,
+      "loss": 1.5156,
+      "step": 8978
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1688497142857143e-05,
+      "loss": 1.2312,
+      "step": 8979
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1688228571428573e-05,
+      "loss": 1.3327,
+      "step": 8980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.168796e-05,
+      "loss": 1.1801,
+      "step": 8981
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.168769142857143e-05,
+      "loss": 1.8366,
+      "step": 8982
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1687422857142858e-05,
+      "loss": 1.1593,
+      "step": 8983
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1687154285714286e-05,
+      "loss": 1.1627,
+      "step": 8984
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1686885714285714e-05,
+      "loss": 1.6146,
+      "step": 8985
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1686617142857144e-05,
+      "loss": 1.1769,
+      "step": 8986
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1686348571428572e-05,
+      "loss": 1.4904,
+      "step": 8987
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.168608e-05,
+      "loss": 1.0213,
+      "step": 8988
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.168581142857143e-05,
+      "loss": 1.5297,
+      "step": 8989
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1685542857142858e-05,
+      "loss": 1.2693,
+      "step": 8990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1685274285714286e-05,
+      "loss": 1.294,
+      "step": 8991
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1685005714285714e-05,
+      "loss": 1.4834,
+      "step": 8992
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1684737142857144e-05,
+      "loss": 1.2701,
+      "step": 8993
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1684468571428571e-05,
+      "loss": 1.3229,
+      "step": 8994
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.16842e-05,
+      "loss": 1.4429,
+      "step": 8995
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1683931428571429e-05,
+      "loss": 0.9954,
+      "step": 8996
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1683662857142857e-05,
+      "loss": 1.1942,
+      "step": 8997
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1683394285714285e-05,
+      "loss": 1.1705,
+      "step": 8998
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1683125714285715e-05,
+      "loss": 1.2518,
+      "step": 8999
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1682857142857143e-05,
+      "loss": 1.8289,
+      "step": 9000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1682588571428571e-05,
+      "loss": 1.2207,
+      "step": 9001
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1682320000000001e-05,
+      "loss": 1.4608,
+      "step": 9002
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.168205142857143e-05,
+      "loss": 1.0322,
+      "step": 9003
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1681782857142858e-05,
+      "loss": 1.4363,
+      "step": 9004
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1681514285714286e-05,
+      "loss": 1.1419,
+      "step": 9005
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1681245714285716e-05,
+      "loss": 0.761,
+      "step": 9006
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1680977142857142e-05,
+      "loss": 1.2563,
+      "step": 9007
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1680708571428572e-05,
+      "loss": 1.1846,
+      "step": 9008
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.168044e-05,
+      "loss": 1.5591,
+      "step": 9009
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1680171428571428e-05,
+      "loss": 1.5296,
+      "step": 9010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1679902857142857e-05,
+      "loss": 1.2253,
+      "step": 9011
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.1679634285714286e-05,
+      "loss": 1.2933,
+      "step": 9012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1679365714285715e-05,
+      "loss": 1.3692,
+      "step": 9013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1679097142857143e-05,
+      "loss": 1.4134,
+      "step": 9014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1678828571428573e-05,
+      "loss": 1.215,
+      "step": 9015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1678560000000001e-05,
+      "loss": 1.5656,
+      "step": 9016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1678291428571429e-05,
+      "loss": 1.5289,
+      "step": 9017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1678022857142857e-05,
+      "loss": 1.5254,
+      "step": 9018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1677754285714287e-05,
+      "loss": 1.2455,
+      "step": 9019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1677485714285714e-05,
+      "loss": 1.3129,
+      "step": 9020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1677217142857143e-05,
+      "loss": 1.4155,
+      "step": 9021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1676948571428572e-05,
+      "loss": 1.2482,
+      "step": 9022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1676680000000002e-05,
+      "loss": 1.1237,
+      "step": 9023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1676411428571428e-05,
+      "loss": 1.1372,
+      "step": 9024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1676142857142858e-05,
+      "loss": 1.0137,
+      "step": 9025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1675874285714286e-05,
+      "loss": 1.0843,
+      "step": 9026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1675605714285714e-05,
+      "loss": 1.412,
+      "step": 9027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1675337142857144e-05,
+      "loss": 1.5886,
+      "step": 9028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1675068571428572e-05,
+      "loss": 1.4994,
+      "step": 9029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.16748e-05,
+      "loss": 1.1983,
+      "step": 9030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1674531428571429e-05,
+      "loss": 1.7812,
+      "step": 9031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1674262857142858e-05,
+      "loss": 1.3993,
+      "step": 9032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1673994285714285e-05,
+      "loss": 1.2082,
+      "step": 9033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1673725714285715e-05,
+      "loss": 1.3999,
+      "step": 9034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1673457142857143e-05,
+      "loss": 1.4154,
+      "step": 9035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1673188571428573e-05,
+      "loss": 1.1895,
+      "step": 9036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.167292e-05,
+      "loss": 1.5279,
+      "step": 9037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.167265142857143e-05,
+      "loss": 0.8691,
+      "step": 9038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1672382857142857e-05,
+      "loss": 1.0012,
+      "step": 9039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1672114285714286e-05,
+      "loss": 1.2517,
+      "step": 9040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1671845714285715e-05,
+      "loss": 1.8736,
+      "step": 9041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1671577142857144e-05,
+      "loss": 1.405,
+      "step": 9042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1671308571428572e-05,
+      "loss": 1.5981,
+      "step": 9043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.167104e-05,
+      "loss": 1.3315,
+      "step": 9044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.167077142857143e-05,
+      "loss": 1.2532,
+      "step": 9045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1670502857142858e-05,
+      "loss": 1.4282,
+      "step": 9046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1670234285714286e-05,
+      "loss": 1.1101,
+      "step": 9047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1669965714285714e-05,
+      "loss": 1.2158,
+      "step": 9048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1669697142857144e-05,
+      "loss": 1.6146,
+      "step": 9049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.166942857142857e-05,
+      "loss": 1.3578,
+      "step": 9050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.166916e-05,
+      "loss": 1.3073,
+      "step": 9051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1668891428571429e-05,
+      "loss": 1.2108,
+      "step": 9052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1668622857142857e-05,
+      "loss": 1.0453,
+      "step": 9053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1668354285714287e-05,
+      "loss": 1.1998,
+      "step": 9054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1668085714285715e-05,
+      "loss": 1.0753,
+      "step": 9055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1667817142857143e-05,
+      "loss": 1.3784,
+      "step": 9056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1667548571428571e-05,
+      "loss": 0.8946,
+      "step": 9057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1667280000000001e-05,
+      "loss": 1.4481,
+      "step": 9058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.166701142857143e-05,
+      "loss": 1.3385,
+      "step": 9059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1666742857142858e-05,
+      "loss": 1.3182,
+      "step": 9060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1666474285714286e-05,
+      "loss": 1.1126,
+      "step": 9061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1666205714285716e-05,
+      "loss": 1.2781,
+      "step": 9062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1665937142857142e-05,
+      "loss": 1.1042,
+      "step": 9063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1665668571428572e-05,
+      "loss": 1.2223,
+      "step": 9064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.16654e-05,
+      "loss": 1.0086,
+      "step": 9065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1665131428571428e-05,
+      "loss": 0.9685,
+      "step": 9066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1664862857142858e-05,
+      "loss": 1.3786,
+      "step": 9067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1664594285714286e-05,
+      "loss": 1.4481,
+      "step": 9068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1664325714285715e-05,
+      "loss": 1.4394,
+      "step": 9069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1664057142857143e-05,
+      "loss": 0.9166,
+      "step": 9070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1663788571428573e-05,
+      "loss": 1.4112,
+      "step": 9071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.166352e-05,
+      "loss": 1.6334,
+      "step": 9072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1663251428571429e-05,
+      "loss": 1.4543,
+      "step": 9073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1662982857142857e-05,
+      "loss": 1.5281,
+      "step": 9074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1662714285714287e-05,
+      "loss": 1.4553,
+      "step": 9075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1662445714285713e-05,
+      "loss": 1.2549,
+      "step": 9076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1662177142857143e-05,
+      "loss": 1.152,
+      "step": 9077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1661908571428572e-05,
+      "loss": 1.3926,
+      "step": 9078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.166164e-05,
+      "loss": 1.7195,
+      "step": 9079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.166137142857143e-05,
+      "loss": 1.3379,
+      "step": 9080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1661102857142858e-05,
+      "loss": 1.271,
+      "step": 9081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1660834285714288e-05,
+      "loss": 1.2389,
+      "step": 9082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1660565714285714e-05,
+      "loss": 0.9801,
+      "step": 9083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1660297142857144e-05,
+      "loss": 1.2135,
+      "step": 9084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1660028571428572e-05,
+      "loss": 1.6011,
+      "step": 9085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.165976e-05,
+      "loss": 1.2281,
+      "step": 9086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1659491428571428e-05,
+      "loss": 1.521,
+      "step": 9087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1659222857142858e-05,
+      "loss": 1.047,
+      "step": 9088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1658954285714285e-05,
+      "loss": 1.1379,
+      "step": 9089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1658685714285715e-05,
+      "loss": 0.9118,
+      "step": 9090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1658417142857143e-05,
+      "loss": 1.0275,
+      "step": 9091
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1658148571428571e-05,
+      "loss": 1.3752,
+      "step": 9092
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1657880000000001e-05,
+      "loss": 1.3602,
+      "step": 9093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1657611428571429e-05,
+      "loss": 1.2072,
+      "step": 9094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1657342857142859e-05,
+      "loss": 1.2991,
+      "step": 9095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1657074285714285e-05,
+      "loss": 1.4626,
+      "step": 9096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1656805714285715e-05,
+      "loss": 1.4284,
+      "step": 9097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1656537142857144e-05,
+      "loss": 1.0303,
+      "step": 9098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1656268571428572e-05,
+      "loss": 0.7567,
+      "step": 9099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1656e-05,
+      "loss": 1.2774,
+      "step": 9100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.165573142857143e-05,
+      "loss": 0.9836,
+      "step": 9101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1655462857142856e-05,
+      "loss": 1.3008,
+      "step": 9102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1655194285714286e-05,
+      "loss": 1.2841,
+      "step": 9103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1654925714285714e-05,
+      "loss": 0.852,
+      "step": 9104
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1654657142857144e-05,
+      "loss": 1.2265,
+      "step": 9105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1654388571428572e-05,
+      "loss": 1.1435,
+      "step": 9106
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.165412e-05,
+      "loss": 1.0537,
+      "step": 9107
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.165385142857143e-05,
+      "loss": 1.699,
+      "step": 9108
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1653582857142857e-05,
+      "loss": 1.1476,
+      "step": 9109
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1653314285714287e-05,
+      "loss": 1.0907,
+      "step": 9110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1653045714285715e-05,
+      "loss": 1.4941,
+      "step": 9111
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1652777142857143e-05,
+      "loss": 0.9714,
+      "step": 9112
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1652508571428571e-05,
+      "loss": 1.454,
+      "step": 9113
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1652240000000001e-05,
+      "loss": 1.4317,
+      "step": 9114
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1651971428571428e-05,
+      "loss": 1.0945,
+      "step": 9115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1651702857142857e-05,
+      "loss": 1.1817,
+      "step": 9116
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1651434285714286e-05,
+      "loss": 1.2534,
+      "step": 9117
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1651165714285716e-05,
+      "loss": 1.0993,
+      "step": 9118
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1650897142857144e-05,
+      "loss": 0.8707,
+      "step": 9119
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1650628571428572e-05,
+      "loss": 1.4532,
+      "step": 9120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1650360000000002e-05,
+      "loss": 1.2408,
+      "step": 9121
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1650091428571428e-05,
+      "loss": 1.4552,
+      "step": 9122
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1649822857142858e-05,
+      "loss": 1.2022,
+      "step": 9123
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1649554285714286e-05,
+      "loss": 1.5763,
+      "step": 9124
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1649285714285714e-05,
+      "loss": 1.1073,
+      "step": 9125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1649017142857143e-05,
+      "loss": 1.4434,
+      "step": 9126
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1648748571428572e-05,
+      "loss": 1.0742,
+      "step": 9127
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.164848e-05,
+      "loss": 1.1575,
+      "step": 9128
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1648211428571429e-05,
+      "loss": 1.4205,
+      "step": 9129
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1647942857142857e-05,
+      "loss": 1.3713,
+      "step": 9130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1647674285714287e-05,
+      "loss": 1.0288,
+      "step": 9131
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1647405714285715e-05,
+      "loss": 1.4165,
+      "step": 9132
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1647137142857143e-05,
+      "loss": 1.8361,
+      "step": 9133
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1646868571428573e-05,
+      "loss": 0.9636,
+      "step": 9134
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.16466e-05,
+      "loss": 1.2392,
+      "step": 9135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.164633142857143e-05,
+      "loss": 1.615,
+      "step": 9136
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1646062857142858e-05,
+      "loss": 1.122,
+      "step": 9137
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1645794285714286e-05,
+      "loss": 1.3618,
+      "step": 9138
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1645525714285714e-05,
+      "loss": 1.1528,
+      "step": 9139
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1645257142857144e-05,
+      "loss": 1.3457,
+      "step": 9140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1644988571428572e-05,
+      "loss": 1.0658,
+      "step": 9141
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.164472e-05,
+      "loss": 1.3691,
+      "step": 9142
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1644451428571428e-05,
+      "loss": 1.0176,
+      "step": 9143
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1644182857142858e-05,
+      "loss": 1.1713,
+      "step": 9144
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1643914285714286e-05,
+      "loss": 1.3075,
+      "step": 9145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1643645714285715e-05,
+      "loss": 1.5556,
+      "step": 9146
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1643377142857144e-05,
+      "loss": 1.3734,
+      "step": 9147
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1643108571428571e-05,
+      "loss": 1.1679,
+      "step": 9148
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.164284e-05,
+      "loss": 1.0825,
+      "step": 9149
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1642571428571429e-05,
+      "loss": 1.4125,
+      "step": 9150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1642302857142859e-05,
+      "loss": 0.9919,
+      "step": 9151
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1642034285714285e-05,
+      "loss": 1.2672,
+      "step": 9152
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1641765714285715e-05,
+      "loss": 1.4477,
+      "step": 9153
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1641497142857143e-05,
+      "loss": 0.7865,
+      "step": 9154
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1641228571428572e-05,
+      "loss": 1.1689,
+      "step": 9155
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.164096e-05,
+      "loss": 1.5548,
+      "step": 9156
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.164069142857143e-05,
+      "loss": 1.0563,
+      "step": 9157
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1640422857142858e-05,
+      "loss": 1.4379,
+      "step": 9158
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1640154285714286e-05,
+      "loss": 0.979,
+      "step": 9159
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1639885714285716e-05,
+      "loss": 1.3126,
+      "step": 9160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1639617142857142e-05,
+      "loss": 1.1678,
+      "step": 9161
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1639348571428572e-05,
+      "loss": 1.1398,
+      "step": 9162
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.163908e-05,
+      "loss": 1.0608,
+      "step": 9163
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.163881142857143e-05,
+      "loss": 0.975,
+      "step": 9164
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1638542857142857e-05,
+      "loss": 1.1296,
+      "step": 9165
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1638274285714287e-05,
+      "loss": 1.1668,
+      "step": 9166
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1638005714285715e-05,
+      "loss": 0.8623,
+      "step": 9167
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1637737142857143e-05,
+      "loss": 1.1395,
+      "step": 9168
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1637468571428571e-05,
+      "loss": 1.0115,
+      "step": 9169
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1637200000000001e-05,
+      "loss": 1.6274,
+      "step": 9170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.163693142857143e-05,
+      "loss": 1.4012,
+      "step": 9171
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1636662857142857e-05,
+      "loss": 1.2018,
+      "step": 9172
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1636394285714287e-05,
+      "loss": 1.1944,
+      "step": 9173
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1636125714285715e-05,
+      "loss": 1.3496,
+      "step": 9174
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1635857142857144e-05,
+      "loss": 1.2544,
+      "step": 9175
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1635588571428572e-05,
+      "loss": 1.2383,
+      "step": 9176
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1635320000000002e-05,
+      "loss": 1.3823,
+      "step": 9177
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1635051428571428e-05,
+      "loss": 1.6891,
+      "step": 9178
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1634782857142858e-05,
+      "loss": 1.2624,
+      "step": 9179
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1634514285714286e-05,
+      "loss": 1.31,
+      "step": 9180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1634245714285714e-05,
+      "loss": 1.0705,
+      "step": 9181
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1633977142857143e-05,
+      "loss": 1.1352,
+      "step": 9182
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1633708571428572e-05,
+      "loss": 1.3198,
+      "step": 9183
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.163344e-05,
+      "loss": 1.4653,
+      "step": 9184
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1633171428571429e-05,
+      "loss": 1.48,
+      "step": 9185
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1632902857142859e-05,
+      "loss": 1.3284,
+      "step": 9186
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.1632634285714287e-05,
+      "loss": 1.4538,
+      "step": 9187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1632365714285715e-05,
+      "loss": 1.1762,
+      "step": 9188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1632097142857143e-05,
+      "loss": 1.2616,
+      "step": 9189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1631828571428573e-05,
+      "loss": 1.1972,
+      "step": 9190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.163156e-05,
+      "loss": 0.8788,
+      "step": 9191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.163129142857143e-05,
+      "loss": 1.5148,
+      "step": 9192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1631022857142858e-05,
+      "loss": 0.7967,
+      "step": 9193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1630754285714286e-05,
+      "loss": 1.3198,
+      "step": 9194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1630485714285714e-05,
+      "loss": 1.0485,
+      "step": 9195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1630217142857144e-05,
+      "loss": 0.9084,
+      "step": 9196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1629948571428572e-05,
+      "loss": 1.0856,
+      "step": 9197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.162968e-05,
+      "loss": 1.0541,
+      "step": 9198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.162941142857143e-05,
+      "loss": 1.2419,
+      "step": 9199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1629142857142858e-05,
+      "loss": 1.6124,
+      "step": 9200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1628874285714286e-05,
+      "loss": 1.4053,
+      "step": 9201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1628605714285714e-05,
+      "loss": 1.3513,
+      "step": 9202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1628337142857144e-05,
+      "loss": 1.1941,
+      "step": 9203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.162806857142857e-05,
+      "loss": 1.4203,
+      "step": 9204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.16278e-05,
+      "loss": 1.4072,
+      "step": 9205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1627531428571429e-05,
+      "loss": 0.9754,
+      "step": 9206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1627262857142857e-05,
+      "loss": 1.1917,
+      "step": 9207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1626994285714285e-05,
+      "loss": 1.6594,
+      "step": 9208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1626725714285715e-05,
+      "loss": 1.2284,
+      "step": 9209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1626457142857143e-05,
+      "loss": 1.5954,
+      "step": 9210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1626188571428571e-05,
+      "loss": 1.4247,
+      "step": 9211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1625920000000001e-05,
+      "loss": 0.7648,
+      "step": 9212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.162565142857143e-05,
+      "loss": 1.3753,
+      "step": 9213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1625382857142858e-05,
+      "loss": 1.5796,
+      "step": 9214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1625114285714286e-05,
+      "loss": 1.4347,
+      "step": 9215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1624845714285716e-05,
+      "loss": 1.6131,
+      "step": 9216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1624577142857142e-05,
+      "loss": 1.1937,
+      "step": 9217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1624308571428572e-05,
+      "loss": 1.5914,
+      "step": 9218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.162404e-05,
+      "loss": 1.6354,
+      "step": 9219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.162377142857143e-05,
+      "loss": 1.3009,
+      "step": 9220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1623502857142857e-05,
+      "loss": 1.6589,
+      "step": 9221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1623234285714286e-05,
+      "loss": 1.3079,
+      "step": 9222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1622965714285715e-05,
+      "loss": 1.5889,
+      "step": 9223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1622697142857143e-05,
+      "loss": 1.2206,
+      "step": 9224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1622428571428573e-05,
+      "loss": 0.9806,
+      "step": 9225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1622160000000001e-05,
+      "loss": 0.9627,
+      "step": 9226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1621891428571429e-05,
+      "loss": 1.0385,
+      "step": 9227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1621622857142857e-05,
+      "loss": 1.1114,
+      "step": 9228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1621354285714287e-05,
+      "loss": 1.4605,
+      "step": 9229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1621085714285714e-05,
+      "loss": 1.0232,
+      "step": 9230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1620817142857143e-05,
+      "loss": 0.9244,
+      "step": 9231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1620548571428572e-05,
+      "loss": 1.4119,
+      "step": 9232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1620280000000002e-05,
+      "loss": 1.4286,
+      "step": 9233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1620011428571428e-05,
+      "loss": 0.9728,
+      "step": 9234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1619742857142858e-05,
+      "loss": 0.8905,
+      "step": 9235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1619474285714286e-05,
+      "loss": 1.2844,
+      "step": 9236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1619205714285714e-05,
+      "loss": 1.279,
+      "step": 9237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1618937142857144e-05,
+      "loss": 1.5254,
+      "step": 9238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1618668571428572e-05,
+      "loss": 1.3689,
+      "step": 9239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.16184e-05,
+      "loss": 1.461,
+      "step": 9240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1618131428571429e-05,
+      "loss": 1.2047,
+      "step": 9241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1617862857142858e-05,
+      "loss": 0.8064,
+      "step": 9242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1617594285714285e-05,
+      "loss": 1.5671,
+      "step": 9243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1617325714285715e-05,
+      "loss": 0.9669,
+      "step": 9244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1617057142857143e-05,
+      "loss": 1.059,
+      "step": 9245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1616788571428573e-05,
+      "loss": 1.4766,
+      "step": 9246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.161652e-05,
+      "loss": 1.1642,
+      "step": 9247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.161625142857143e-05,
+      "loss": 0.7786,
+      "step": 9248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1615982857142857e-05,
+      "loss": 0.9064,
+      "step": 9249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1615714285714286e-05,
+      "loss": 1.2161,
+      "step": 9250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1615445714285715e-05,
+      "loss": 1.6484,
+      "step": 9251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1615177142857144e-05,
+      "loss": 1.399,
+      "step": 9252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1614908571428572e-05,
+      "loss": 1.2501,
+      "step": 9253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.161464e-05,
+      "loss": 1.4388,
+      "step": 9254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.161437142857143e-05,
+      "loss": 1.6279,
+      "step": 9255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1614102857142858e-05,
+      "loss": 1.1019,
+      "step": 9256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1613834285714286e-05,
+      "loss": 1.7798,
+      "step": 9257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1613565714285714e-05,
+      "loss": 1.6018,
+      "step": 9258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1613297142857144e-05,
+      "loss": 0.9854,
+      "step": 9259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.161302857142857e-05,
+      "loss": 1.1337,
+      "step": 9260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.161276e-05,
+      "loss": 1.0937,
+      "step": 9261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1612491428571429e-05,
+      "loss": 0.8904,
+      "step": 9262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1612222857142857e-05,
+      "loss": 1.2943,
+      "step": 9263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1611954285714287e-05,
+      "loss": 1.3454,
+      "step": 9264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1611685714285715e-05,
+      "loss": 1.4804,
+      "step": 9265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1611417142857143e-05,
+      "loss": 1.6224,
+      "step": 9266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1611148571428571e-05,
+      "loss": 1.4389,
+      "step": 9267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1610880000000001e-05,
+      "loss": 1.1448,
+      "step": 9268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.161061142857143e-05,
+      "loss": 1.286,
+      "step": 9269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1610342857142858e-05,
+      "loss": 1.3487,
+      "step": 9270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1610074285714286e-05,
+      "loss": 1.319,
+      "step": 9271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1609805714285716e-05,
+      "loss": 1.1352,
+      "step": 9272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1609537142857142e-05,
+      "loss": 1.3652,
+      "step": 9273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1609268571428572e-05,
+      "loss": 1.551,
+      "step": 9274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1609e-05,
+      "loss": 1.3808,
+      "step": 9275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1608731428571428e-05,
+      "loss": 1.0561,
+      "step": 9276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1608462857142858e-05,
+      "loss": 1.3944,
+      "step": 9277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1608194285714286e-05,
+      "loss": 0.9501,
+      "step": 9278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1607925714285716e-05,
+      "loss": 1.2941,
+      "step": 9279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1607657142857143e-05,
+      "loss": 1.5541,
+      "step": 9280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1607388571428573e-05,
+      "loss": 1.3954,
+      "step": 9281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.160712e-05,
+      "loss": 1.4164,
+      "step": 9282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1606851428571429e-05,
+      "loss": 1.4331,
+      "step": 9283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1606582857142857e-05,
+      "loss": 1.4672,
+      "step": 9284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1606314285714287e-05,
+      "loss": 1.3328,
+      "step": 9285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1606045714285713e-05,
+      "loss": 1.4406,
+      "step": 9286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1605777142857143e-05,
+      "loss": 1.3572,
+      "step": 9287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1605508571428572e-05,
+      "loss": 0.7901,
+      "step": 9288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.160524e-05,
+      "loss": 1.3873,
+      "step": 9289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.160497142857143e-05,
+      "loss": 1.4983,
+      "step": 9290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1604702857142858e-05,
+      "loss": 1.3889,
+      "step": 9291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1604434285714288e-05,
+      "loss": 1.5327,
+      "step": 9292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1604165714285714e-05,
+      "loss": 1.7879,
+      "step": 9293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1603897142857144e-05,
+      "loss": 1.4686,
+      "step": 9294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1603628571428572e-05,
+      "loss": 1.0437,
+      "step": 9295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.160336e-05,
+      "loss": 1.385,
+      "step": 9296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1603091428571428e-05,
+      "loss": 1.3504,
+      "step": 9297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1602822857142858e-05,
+      "loss": 1.4099,
+      "step": 9298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1602554285714285e-05,
+      "loss": 0.9177,
+      "step": 9299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1602285714285715e-05,
+      "loss": 1.309,
+      "step": 9300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1602017142857143e-05,
+      "loss": 0.9803,
+      "step": 9301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1601748571428573e-05,
+      "loss": 1.5564,
+      "step": 9302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.160148e-05,
+      "loss": 0.9911,
+      "step": 9303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1601211428571429e-05,
+      "loss": 0.7713,
+      "step": 9304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1600942857142859e-05,
+      "loss": 1.4062,
+      "step": 9305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1600674285714285e-05,
+      "loss": 1.3205,
+      "step": 9306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1600405714285715e-05,
+      "loss": 1.1808,
+      "step": 9307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1600137142857144e-05,
+      "loss": 1.1126,
+      "step": 9308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1599868571428572e-05,
+      "loss": 1.1105,
+      "step": 9309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.15996e-05,
+      "loss": 1.1455,
+      "step": 9310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.159933142857143e-05,
+      "loss": 1.253,
+      "step": 9311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1599062857142856e-05,
+      "loss": 1.2159,
+      "step": 9312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1598794285714286e-05,
+      "loss": 1.3954,
+      "step": 9313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1598525714285714e-05,
+      "loss": 0.9431,
+      "step": 9314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1598257142857144e-05,
+      "loss": 1.7848,
+      "step": 9315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.159798857142857e-05,
+      "loss": 1.0285,
+      "step": 9316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.159772e-05,
+      "loss": 1.2998,
+      "step": 9317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.159745142857143e-05,
+      "loss": 1.3801,
+      "step": 9318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1597182857142857e-05,
+      "loss": 1.0712,
+      "step": 9319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1596914285714287e-05,
+      "loss": 1.3752,
+      "step": 9320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1596645714285715e-05,
+      "loss": 1.3238,
+      "step": 9321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1596377142857143e-05,
+      "loss": 1.507,
+      "step": 9322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1596108571428571e-05,
+      "loss": 1.0678,
+      "step": 9323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1595840000000001e-05,
+      "loss": 1.3155,
+      "step": 9324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.159557142857143e-05,
+      "loss": 1.3273,
+      "step": 9325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1595302857142857e-05,
+      "loss": 1.0371,
+      "step": 9326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1595034285714286e-05,
+      "loss": 0.7976,
+      "step": 9327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1594765714285716e-05,
+      "loss": 1.3429,
+      "step": 9328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1594497142857142e-05,
+      "loss": 1.1112,
+      "step": 9329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1594228571428572e-05,
+      "loss": 1.0245,
+      "step": 9330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1593960000000002e-05,
+      "loss": 1.6003,
+      "step": 9331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1593691428571428e-05,
+      "loss": 1.2332,
+      "step": 9332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1593422857142858e-05,
+      "loss": 1.2919,
+      "step": 9333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1593154285714286e-05,
+      "loss": 1.2783,
+      "step": 9334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1592885714285714e-05,
+      "loss": 1.0313,
+      "step": 9335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1592617142857143e-05,
+      "loss": 1.0803,
+      "step": 9336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1592348571428572e-05,
+      "loss": 1.1126,
+      "step": 9337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.159208e-05,
+      "loss": 1.2125,
+      "step": 9338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1591811428571429e-05,
+      "loss": 0.9182,
+      "step": 9339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1591542857142857e-05,
+      "loss": 0.977,
+      "step": 9340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1591274285714287e-05,
+      "loss": 1.3105,
+      "step": 9341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1591005714285713e-05,
+      "loss": 1.4022,
+      "step": 9342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1590737142857143e-05,
+      "loss": 0.9742,
+      "step": 9343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1590468571428573e-05,
+      "loss": 1.5451,
+      "step": 9344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.15902e-05,
+      "loss": 1.4098,
+      "step": 9345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.158993142857143e-05,
+      "loss": 1.4103,
+      "step": 9346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1589662857142858e-05,
+      "loss": 1.2368,
+      "step": 9347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1589394285714288e-05,
+      "loss": 1.2371,
+      "step": 9348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1589125714285714e-05,
+      "loss": 1.0895,
+      "step": 9349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1588857142857144e-05,
+      "loss": 1.516,
+      "step": 9350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1588588571428572e-05,
+      "loss": 1.1048,
+      "step": 9351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.158832e-05,
+      "loss": 1.5407,
+      "step": 9352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1588051428571428e-05,
+      "loss": 1.2267,
+      "step": 9353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1587782857142858e-05,
+      "loss": 1.1251,
+      "step": 9354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1587514285714285e-05,
+      "loss": 1.7174,
+      "step": 9355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1587245714285715e-05,
+      "loss": 1.5504,
+      "step": 9356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1586977142857144e-05,
+      "loss": 1.1587,
+      "step": 9357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1586708571428571e-05,
+      "loss": 1.6303,
+      "step": 9358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.158644e-05,
+      "loss": 1.206,
+      "step": 9359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1586171428571429e-05,
+      "loss": 1.7038,
+      "step": 9360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1585902857142859e-05,
+      "loss": 1.4837,
+      "step": 9361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.1585634285714285e-05,
+      "loss": 1.2361,
+      "step": 9362
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1585365714285715e-05,
+      "loss": 1.1157,
+      "step": 9363
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1585097142857143e-05,
+      "loss": 1.0544,
+      "step": 9364
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1584828571428572e-05,
+      "loss": 1.5004,
+      "step": 9365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.158456e-05,
+      "loss": 1.6329,
+      "step": 9366
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.158429142857143e-05,
+      "loss": 1.3989,
+      "step": 9367
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1584022857142856e-05,
+      "loss": 1.7316,
+      "step": 9368
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1583754285714286e-05,
+      "loss": 1.2932,
+      "step": 9369
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1583485714285716e-05,
+      "loss": 1.489,
+      "step": 9370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1583217142857144e-05,
+      "loss": 1.2663,
+      "step": 9371
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1582948571428572e-05,
+      "loss": 1.37,
+      "step": 9372
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.158268e-05,
+      "loss": 1.4131,
+      "step": 9373
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.158241142857143e-05,
+      "loss": 1.0457,
+      "step": 9374
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1582142857142857e-05,
+      "loss": 1.3554,
+      "step": 9375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1581874285714287e-05,
+      "loss": 0.9441,
+      "step": 9376
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1581605714285715e-05,
+      "loss": 1.1802,
+      "step": 9377
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1581337142857143e-05,
+      "loss": 1.6271,
+      "step": 9378
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1581068571428571e-05,
+      "loss": 1.6215,
+      "step": 9379
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1580800000000001e-05,
+      "loss": 0.9338,
+      "step": 9380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1580531428571427e-05,
+      "loss": 1.0526,
+      "step": 9381
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1580262857142857e-05,
+      "loss": 1.6175,
+      "step": 9382
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1579994285714287e-05,
+      "loss": 1.4728,
+      "step": 9383
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1579725714285715e-05,
+      "loss": 1.0433,
+      "step": 9384
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1579457142857144e-05,
+      "loss": 1.1178,
+      "step": 9385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1579188571428572e-05,
+      "loss": 1.7538,
+      "step": 9386
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1578920000000002e-05,
+      "loss": 1.0801,
+      "step": 9387
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1578651428571428e-05,
+      "loss": 1.1319,
+      "step": 9388
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1578382857142858e-05,
+      "loss": 1.2517,
+      "step": 9389
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1578114285714286e-05,
+      "loss": 1.5805,
+      "step": 9390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1577845714285714e-05,
+      "loss": 1.6524,
+      "step": 9391
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1577577142857142e-05,
+      "loss": 1.1968,
+      "step": 9392
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1577308571428572e-05,
+      "loss": 0.9414,
+      "step": 9393
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1577039999999999e-05,
+      "loss": 1.0094,
+      "step": 9394
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1576771428571429e-05,
+      "loss": 0.9805,
+      "step": 9395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1576502857142859e-05,
+      "loss": 1.5263,
+      "step": 9396
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1576234285714287e-05,
+      "loss": 1.199,
+      "step": 9397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1575965714285715e-05,
+      "loss": 1.2734,
+      "step": 9398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1575697142857143e-05,
+      "loss": 1.3315,
+      "step": 9399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1575428571428573e-05,
+      "loss": 0.9816,
+      "step": 9400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.157516e-05,
+      "loss": 1.4991,
+      "step": 9401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.157489142857143e-05,
+      "loss": 1.3898,
+      "step": 9402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1574622857142858e-05,
+      "loss": 1.2344,
+      "step": 9403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1574354285714286e-05,
+      "loss": 1.2173,
+      "step": 9404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1574085714285714e-05,
+      "loss": 1.2834,
+      "step": 9405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1573817142857144e-05,
+      "loss": 1.394,
+      "step": 9406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1573548571428572e-05,
+      "loss": 0.898,
+      "step": 9407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.157328e-05,
+      "loss": 1.3509,
+      "step": 9408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1573011428571428e-05,
+      "loss": 0.8676,
+      "step": 9409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1572742857142858e-05,
+      "loss": 1.6309,
+      "step": 9410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1572474285714286e-05,
+      "loss": 1.3168,
+      "step": 9411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1572205714285714e-05,
+      "loss": 1.4921,
+      "step": 9412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1571937142857144e-05,
+      "loss": 1.0716,
+      "step": 9413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.157166857142857e-05,
+      "loss": 1.2507,
+      "step": 9414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.15714e-05,
+      "loss": 1.0273,
+      "step": 9415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1571131428571429e-05,
+      "loss": 1.1092,
+      "step": 9416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1570862857142857e-05,
+      "loss": 1.0182,
+      "step": 9417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1570594285714285e-05,
+      "loss": 1.191,
+      "step": 9418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1570325714285715e-05,
+      "loss": 1.2591,
+      "step": 9419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1570057142857143e-05,
+      "loss": 0.9761,
+      "step": 9420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1569788571428571e-05,
+      "loss": 1.5349,
+      "step": 9421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.156952e-05,
+      "loss": 1.4692,
+      "step": 9422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.156925142857143e-05,
+      "loss": 1.349,
+      "step": 9423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1568982857142858e-05,
+      "loss": 1.451,
+      "step": 9424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1568714285714286e-05,
+      "loss": 1.4852,
+      "step": 9425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1568445714285716e-05,
+      "loss": 0.9871,
+      "step": 9426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1568177142857142e-05,
+      "loss": 1.5623,
+      "step": 9427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1567908571428572e-05,
+      "loss": 1.2147,
+      "step": 9428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.156764e-05,
+      "loss": 1.6158,
+      "step": 9429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.156737142857143e-05,
+      "loss": 1.5495,
+      "step": 9430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1567102857142857e-05,
+      "loss": 1.606,
+      "step": 9431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1566834285714286e-05,
+      "loss": 1.3328,
+      "step": 9432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1566565714285715e-05,
+      "loss": 0.9563,
+      "step": 9433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1566297142857143e-05,
+      "loss": 1.0106,
+      "step": 9434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1566028571428571e-05,
+      "loss": 1.349,
+      "step": 9435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1565760000000001e-05,
+      "loss": 1.3768,
+      "step": 9436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1565491428571429e-05,
+      "loss": 1.2902,
+      "step": 9437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1565222857142857e-05,
+      "loss": 1.5642,
+      "step": 9438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1564954285714287e-05,
+      "loss": 1.3046,
+      "step": 9439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1564685714285714e-05,
+      "loss": 1.126,
+      "step": 9440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1564417142857143e-05,
+      "loss": 1.46,
+      "step": 9441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1564148571428572e-05,
+      "loss": 1.4106,
+      "step": 9442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1563880000000002e-05,
+      "loss": 1.1978,
+      "step": 9443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1563611428571428e-05,
+      "loss": 1.3288,
+      "step": 9444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1563342857142858e-05,
+      "loss": 1.785,
+      "step": 9445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1563074285714286e-05,
+      "loss": 1.1713,
+      "step": 9446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1562805714285714e-05,
+      "loss": 1.0304,
+      "step": 9447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1562537142857142e-05,
+      "loss": 1.6837,
+      "step": 9448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1562268571428572e-05,
+      "loss": 1.1168,
+      "step": 9449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1562e-05,
+      "loss": 1.1357,
+      "step": 9450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1561731428571429e-05,
+      "loss": 1.0689,
+      "step": 9451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1561462857142858e-05,
+      "loss": 1.1739,
+      "step": 9452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1561194285714287e-05,
+      "loss": 0.913,
+      "step": 9453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1560925714285715e-05,
+      "loss": 1.1025,
+      "step": 9454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1560657142857143e-05,
+      "loss": 1.3356,
+      "step": 9455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1560388571428573e-05,
+      "loss": 0.7866,
+      "step": 9456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.156012e-05,
+      "loss": 1.5531,
+      "step": 9457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.155985142857143e-05,
+      "loss": 0.8896,
+      "step": 9458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1559582857142857e-05,
+      "loss": 1.1861,
+      "step": 9459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1559314285714286e-05,
+      "loss": 1.1248,
+      "step": 9460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1559045714285714e-05,
+      "loss": 1.2925,
+      "step": 9461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1558777142857144e-05,
+      "loss": 1.5891,
+      "step": 9462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1558508571428572e-05,
+      "loss": 1.4471,
+      "step": 9463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.155824e-05,
+      "loss": 1.1353,
+      "step": 9464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.155797142857143e-05,
+      "loss": 1.0112,
+      "step": 9465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1557702857142858e-05,
+      "loss": 0.9982,
+      "step": 9466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1557434285714286e-05,
+      "loss": 1.5184,
+      "step": 9467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1557165714285714e-05,
+      "loss": 1.1481,
+      "step": 9468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1556897142857144e-05,
+      "loss": 1.2573,
+      "step": 9469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.155662857142857e-05,
+      "loss": 1.5107,
+      "step": 9470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.155636e-05,
+      "loss": 1.0715,
+      "step": 9471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1556091428571429e-05,
+      "loss": 1.0671,
+      "step": 9472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1555822857142857e-05,
+      "loss": 1.2624,
+      "step": 9473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1555554285714285e-05,
+      "loss": 1.1258,
+      "step": 9474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1555285714285715e-05,
+      "loss": 1.0705,
+      "step": 9475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1555017142857145e-05,
+      "loss": 1.3018,
+      "step": 9476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1554748571428571e-05,
+      "loss": 1.4047,
+      "step": 9477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1554480000000001e-05,
+      "loss": 1.2538,
+      "step": 9478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.155421142857143e-05,
+      "loss": 1.3437,
+      "step": 9479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1553942857142858e-05,
+      "loss": 1.2068,
+      "step": 9480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1553674285714286e-05,
+      "loss": 1.2249,
+      "step": 9481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1553405714285716e-05,
+      "loss": 1.3687,
+      "step": 9482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1553137142857142e-05,
+      "loss": 1.5477,
+      "step": 9483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1552868571428572e-05,
+      "loss": 1.1792,
+      "step": 9484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.15526e-05,
+      "loss": 1.3254,
+      "step": 9485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1552331428571428e-05,
+      "loss": 1.3919,
+      "step": 9486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1552062857142857e-05,
+      "loss": 1.028,
+      "step": 9487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1551794285714286e-05,
+      "loss": 0.9886,
+      "step": 9488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1551525714285716e-05,
+      "loss": 1.3325,
+      "step": 9489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1551257142857143e-05,
+      "loss": 0.9042,
+      "step": 9490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1550988571428573e-05,
+      "loss": 1.1241,
+      "step": 9491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.155072e-05,
+      "loss": 1.4274,
+      "step": 9492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1550451428571429e-05,
+      "loss": 1.506,
+      "step": 9493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1550182857142857e-05,
+      "loss": 1.4179,
+      "step": 9494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1549914285714287e-05,
+      "loss": 1.4014,
+      "step": 9495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1549645714285713e-05,
+      "loss": 1.35,
+      "step": 9496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1549377142857143e-05,
+      "loss": 0.5924,
+      "step": 9497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1549108571428572e-05,
+      "loss": 1.28,
+      "step": 9498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1548840000000001e-05,
+      "loss": 1.518,
+      "step": 9499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1548571428571428e-05,
+      "loss": 1.0152,
+      "step": 9500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1548302857142858e-05,
+      "loss": 1.0952,
+      "step": 9501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1548034285714288e-05,
+      "loss": 1.6827,
+      "step": 9502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1547765714285714e-05,
+      "loss": 1.2261,
+      "step": 9503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1547497142857144e-05,
+      "loss": 1.1607,
+      "step": 9504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1547228571428572e-05,
+      "loss": 1.232,
+      "step": 9505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.154696e-05,
+      "loss": 1.7076,
+      "step": 9506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1546691428571428e-05,
+      "loss": 1.2388,
+      "step": 9507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1546422857142858e-05,
+      "loss": 1.241,
+      "step": 9508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1546154285714285e-05,
+      "loss": 1.586,
+      "step": 9509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1545885714285715e-05,
+      "loss": 1.1891,
+      "step": 9510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1545617142857143e-05,
+      "loss": 1.4275,
+      "step": 9511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1545348571428573e-05,
+      "loss": 1.3161,
+      "step": 9512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.154508e-05,
+      "loss": 1.2746,
+      "step": 9513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1544811428571429e-05,
+      "loss": 1.3247,
+      "step": 9514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1544542857142857e-05,
+      "loss": 1.2662,
+      "step": 9515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1544274285714285e-05,
+      "loss": 1.4374,
+      "step": 9516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1544005714285715e-05,
+      "loss": 1.2879,
+      "step": 9517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1543737142857144e-05,
+      "loss": 1.067,
+      "step": 9518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1543468571428572e-05,
+      "loss": 1.2198,
+      "step": 9519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.15432e-05,
+      "loss": 1.2737,
+      "step": 9520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.154293142857143e-05,
+      "loss": 1.1233,
+      "step": 9521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1542662857142858e-05,
+      "loss": 1.2897,
+      "step": 9522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1542394285714286e-05,
+      "loss": 1.4766,
+      "step": 9523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1542125714285714e-05,
+      "loss": 1.022,
+      "step": 9524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1541857142857144e-05,
+      "loss": 1.318,
+      "step": 9525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.154158857142857e-05,
+      "loss": 1.0669,
+      "step": 9526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.154132e-05,
+      "loss": 0.9327,
+      "step": 9527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1541051428571429e-05,
+      "loss": 1.0324,
+      "step": 9528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1540782857142857e-05,
+      "loss": 1.312,
+      "step": 9529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1540514285714287e-05,
+      "loss": 1.6776,
+      "step": 9530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1540245714285715e-05,
+      "loss": 1.1422,
+      "step": 9531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1539977142857143e-05,
+      "loss": 1.4437,
+      "step": 9532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1539708571428571e-05,
+      "loss": 1.1038,
+      "step": 9533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1539440000000001e-05,
+      "loss": 1.1345,
+      "step": 9534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.153917142857143e-05,
+      "loss": 1.2259,
+      "step": 9535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1538902857142857e-05,
+      "loss": 0.7488,
+      "step": 9536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1538634285714286e-05,
+      "loss": 0.9583,
+      "step": 9537
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1538365714285716e-05,
+      "loss": 1.2701,
+      "step": 9538
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1538097142857142e-05,
+      "loss": 1.1561,
+      "step": 9539
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1537828571428572e-05,
+      "loss": 1.2022,
+      "step": 9540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.153756e-05,
+      "loss": 1.2176,
+      "step": 9541
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1537291428571428e-05,
+      "loss": 1.4196,
+      "step": 9542
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1537022857142858e-05,
+      "loss": 1.4335,
+      "step": 9543
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1536754285714286e-05,
+      "loss": 1.1678,
+      "step": 9544
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1536485714285716e-05,
+      "loss": 1.3705,
+      "step": 9545
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1536217142857143e-05,
+      "loss": 1.2156,
+      "step": 9546
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1535948571428572e-05,
+      "loss": 0.8173,
+      "step": 9547
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.153568e-05,
+      "loss": 1.2604,
+      "step": 9548
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1535411428571429e-05,
+      "loss": 0.9269,
+      "step": 9549
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1535142857142857e-05,
+      "loss": 1.4686,
+      "step": 9550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1534874285714287e-05,
+      "loss": 1.2521,
+      "step": 9551
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1534605714285713e-05,
+      "loss": 1.2989,
+      "step": 9552
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1534337142857143e-05,
+      "loss": 1.3527,
+      "step": 9553
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1534068571428571e-05,
+      "loss": 1.0638,
+      "step": 9554
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.15338e-05,
+      "loss": 1.4406,
+      "step": 9555
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.153353142857143e-05,
+      "loss": 1.459,
+      "step": 9556
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1533262857142858e-05,
+      "loss": 1.2156,
+      "step": 9557
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1532994285714288e-05,
+      "loss": 1.2831,
+      "step": 9558
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1532725714285714e-05,
+      "loss": 1.1348,
+      "step": 9559
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1532457142857144e-05,
+      "loss": 1.5045,
+      "step": 9560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1532188571428572e-05,
+      "loss": 1.0084,
+      "step": 9561
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.153192e-05,
+      "loss": 1.0593,
+      "step": 9562
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1531651428571428e-05,
+      "loss": 1.4581,
+      "step": 9563
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1531382857142858e-05,
+      "loss": 1.3865,
+      "step": 9564
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1531114285714285e-05,
+      "loss": 1.1065,
+      "step": 9565
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1530845714285715e-05,
+      "loss": 1.2533,
+      "step": 9566
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1530577142857143e-05,
+      "loss": 1.1084,
+      "step": 9567
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1530308571428571e-05,
+      "loss": 1.1066,
+      "step": 9568
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.153004e-05,
+      "loss": 1.3755,
+      "step": 9569
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1529771428571429e-05,
+      "loss": 1.1907,
+      "step": 9570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1529502857142859e-05,
+      "loss": 1.3469,
+      "step": 9571
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1529234285714285e-05,
+      "loss": 1.5409,
+      "step": 9572
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1528965714285715e-05,
+      "loss": 1.3687,
+      "step": 9573
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1528697142857143e-05,
+      "loss": 1.4898,
+      "step": 9574
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1528428571428572e-05,
+      "loss": 1.7311,
+      "step": 9575
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.152816e-05,
+      "loss": 1.3623,
+      "step": 9576
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.152789142857143e-05,
+      "loss": 1.12,
+      "step": 9577
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1527622857142856e-05,
+      "loss": 1.2675,
+      "step": 9578
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1527354285714286e-05,
+      "loss": 1.2312,
+      "step": 9579
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1527085714285714e-05,
+      "loss": 1.2934,
+      "step": 9580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1526817142857144e-05,
+      "loss": 1.75,
+      "step": 9581
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1526548571428572e-05,
+      "loss": 1.5126,
+      "step": 9582
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.152628e-05,
+      "loss": 1.2136,
+      "step": 9583
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.152601142857143e-05,
+      "loss": 1.3665,
+      "step": 9584
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1525742857142857e-05,
+      "loss": 1.2808,
+      "step": 9585
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1525474285714287e-05,
+      "loss": 0.9682,
+      "step": 9586
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1525205714285715e-05,
+      "loss": 1.1744,
+      "step": 9587
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1524937142857143e-05,
+      "loss": 1.2921,
+      "step": 9588
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1524668571428571e-05,
+      "loss": 1.2975,
+      "step": 9589
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1524400000000001e-05,
+      "loss": 1.169,
+      "step": 9590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1524131428571427e-05,
+      "loss": 0.9597,
+      "step": 9591
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1523862857142857e-05,
+      "loss": 1.3737,
+      "step": 9592
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1523594285714286e-05,
+      "loss": 1.1586,
+      "step": 9593
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1523325714285715e-05,
+      "loss": 1.2827,
+      "step": 9594
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1523057142857144e-05,
+      "loss": 1.6002,
+      "step": 9595
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1522788571428572e-05,
+      "loss": 1.2379,
+      "step": 9596
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1522520000000002e-05,
+      "loss": 1.2801,
+      "step": 9597
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1522251428571428e-05,
+      "loss": 1.2321,
+      "step": 9598
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1521982857142858e-05,
+      "loss": 1.6807,
+      "step": 9599
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1521714285714286e-05,
+      "loss": 1.2549,
+      "step": 9600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1521445714285714e-05,
+      "loss": 1.0523,
+      "step": 9601
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1521177142857142e-05,
+      "loss": 1.263,
+      "step": 9602
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1520908571428572e-05,
+      "loss": 1.5971,
+      "step": 9603
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.152064e-05,
+      "loss": 1.103,
+      "step": 9604
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1520371428571429e-05,
+      "loss": 1.6126,
+      "step": 9605
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1520102857142857e-05,
+      "loss": 1.4393,
+      "step": 9606
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1519834285714287e-05,
+      "loss": 1.4468,
+      "step": 9607
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1519565714285715e-05,
+      "loss": 1.2897,
+      "step": 9608
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1519297142857143e-05,
+      "loss": 1.2039,
+      "step": 9609
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1519028571428573e-05,
+      "loss": 0.8804,
+      "step": 9610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151876e-05,
+      "loss": 1.1096,
+      "step": 9611
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151849142857143e-05,
+      "loss": 1.1697,
+      "step": 9612
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1518222857142858e-05,
+      "loss": 0.9475,
+      "step": 9613
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1517954285714286e-05,
+      "loss": 1.351,
+      "step": 9614
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1517685714285714e-05,
+      "loss": 0.9795,
+      "step": 9615
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1517417142857144e-05,
+      "loss": 1.3169,
+      "step": 9616
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1517148571428572e-05,
+      "loss": 1.114,
+      "step": 9617
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151688e-05,
+      "loss": 1.249,
+      "step": 9618
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1516611428571428e-05,
+      "loss": 1.4892,
+      "step": 9619
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1516342857142858e-05,
+      "loss": 0.8048,
+      "step": 9620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1516074285714286e-05,
+      "loss": 1.1418,
+      "step": 9621
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1515805714285714e-05,
+      "loss": 1.0123,
+      "step": 9622
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1515537142857144e-05,
+      "loss": 1.287,
+      "step": 9623
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151526857142857e-05,
+      "loss": 1.5077,
+      "step": 9624
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1515e-05,
+      "loss": 0.9715,
+      "step": 9625
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1514731428571429e-05,
+      "loss": 1.0293,
+      "step": 9626
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1514462857142859e-05,
+      "loss": 1.2437,
+      "step": 9627
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1514194285714285e-05,
+      "loss": 1.2872,
+      "step": 9628
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1513925714285715e-05,
+      "loss": 1.2801,
+      "step": 9629
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1513657142857143e-05,
+      "loss": 1.608,
+      "step": 9630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1513388571428571e-05,
+      "loss": 1.1827,
+      "step": 9631
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151312e-05,
+      "loss": 0.8344,
+      "step": 9632
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151285142857143e-05,
+      "loss": 1.2071,
+      "step": 9633
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1512582857142858e-05,
+      "loss": 0.8969,
+      "step": 9634
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1512314285714286e-05,
+      "loss": 1.3709,
+      "step": 9635
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1512045714285716e-05,
+      "loss": 1.271,
+      "step": 9636
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1511777142857142e-05,
+      "loss": 0.9613,
+      "step": 9637
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1511508571428572e-05,
+      "loss": 1.289,
+      "step": 9638
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151124e-05,
+      "loss": 1.2711,
+      "step": 9639
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.151097142857143e-05,
+      "loss": 0.8426,
+      "step": 9640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1510702857142857e-05,
+      "loss": 1.0553,
+      "step": 9641
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1510434285714286e-05,
+      "loss": 1.1549,
+      "step": 9642
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1510165714285715e-05,
+      "loss": 1.1623,
+      "step": 9643
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1509897142857143e-05,
+      "loss": 1.1946,
+      "step": 9644
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1509628571428571e-05,
+      "loss": 1.2177,
+      "step": 9645
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1509360000000001e-05,
+      "loss": 1.1727,
+      "step": 9646
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1509091428571429e-05,
+      "loss": 1.3041,
+      "step": 9647
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1508822857142857e-05,
+      "loss": 1.6602,
+      "step": 9648
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1508554285714287e-05,
+      "loss": 1.1019,
+      "step": 9649
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1508285714285715e-05,
+      "loss": 1.2163,
+      "step": 9650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1508017142857143e-05,
+      "loss": 1.5254,
+      "step": 9651
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1507748571428572e-05,
+      "loss": 1.258,
+      "step": 9652
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1507480000000002e-05,
+      "loss": 1.5812,
+      "step": 9653
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1507211428571428e-05,
+      "loss": 0.9277,
+      "step": 9654
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1506942857142858e-05,
+      "loss": 1.445,
+      "step": 9655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1506674285714286e-05,
+      "loss": 1.4649,
+      "step": 9656
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1506405714285714e-05,
+      "loss": 1.2335,
+      "step": 9657
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1506137142857142e-05,
+      "loss": 1.2146,
+      "step": 9658
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1505868571428572e-05,
+      "loss": 1.1143,
+      "step": 9659
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.15056e-05,
+      "loss": 1.3553,
+      "step": 9660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1505331428571429e-05,
+      "loss": 1.4265,
+      "step": 9661
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1505062857142858e-05,
+      "loss": 1.1069,
+      "step": 9662
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1504794285714287e-05,
+      "loss": 1.3746,
+      "step": 9663
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1504525714285715e-05,
+      "loss": 0.9592,
+      "step": 9664
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1504257142857143e-05,
+      "loss": 1.1534,
+      "step": 9665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1503988571428573e-05,
+      "loss": 1.2743,
+      "step": 9666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.150372e-05,
+      "loss": 1.1253,
+      "step": 9667
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.150345142857143e-05,
+      "loss": 0.9849,
+      "step": 9668
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1503182857142857e-05,
+      "loss": 1.065,
+      "step": 9669
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1502914285714286e-05,
+      "loss": 1.1915,
+      "step": 9670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1502645714285714e-05,
+      "loss": 0.9894,
+      "step": 9671
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1502377142857144e-05,
+      "loss": 1.3631,
+      "step": 9672
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1502108571428572e-05,
+      "loss": 1.5896,
+      "step": 9673
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.150184e-05,
+      "loss": 1.1386,
+      "step": 9674
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.150157142857143e-05,
+      "loss": 1.6955,
+      "step": 9675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1501302857142858e-05,
+      "loss": 1.1124,
+      "step": 9676
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1501034285714286e-05,
+      "loss": 0.7866,
+      "step": 9677
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1500765714285714e-05,
+      "loss": 1.6154,
+      "step": 9678
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1500497142857144e-05,
+      "loss": 1.317,
+      "step": 9679
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.150022857142857e-05,
+      "loss": 1.2542,
+      "step": 9680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.149996e-05,
+      "loss": 1.0996,
+      "step": 9681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1499691428571429e-05,
+      "loss": 1.5826,
+      "step": 9682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1499422857142857e-05,
+      "loss": 1.0665,
+      "step": 9683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1499154285714285e-05,
+      "loss": 1.2096,
+      "step": 9684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1498885714285715e-05,
+      "loss": 1.2889,
+      "step": 9685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1498617142857143e-05,
+      "loss": 0.8944,
+      "step": 9686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1498348571428571e-05,
+      "loss": 1.1124,
+      "step": 9687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1498080000000001e-05,
+      "loss": 1.8256,
+      "step": 9688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.149781142857143e-05,
+      "loss": 0.8344,
+      "step": 9689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1497542857142858e-05,
+      "loss": 1.313,
+      "step": 9690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1497274285714286e-05,
+      "loss": 1.4862,
+      "step": 9691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1497005714285716e-05,
+      "loss": 1.1858,
+      "step": 9692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1496737142857142e-05,
+      "loss": 1.3633,
+      "step": 9693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1496468571428572e-05,
+      "loss": 0.8955,
+      "step": 9694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.14962e-05,
+      "loss": 1.0928,
+      "step": 9695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.149593142857143e-05,
+      "loss": 1.1285,
+      "step": 9696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1495662857142856e-05,
+      "loss": 1.0631,
+      "step": 9697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1495394285714286e-05,
+      "loss": 1.2911,
+      "step": 9698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1495125714285715e-05,
+      "loss": 1.0849,
+      "step": 9699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1494857142857143e-05,
+      "loss": 1.5117,
+      "step": 9700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1494588571428573e-05,
+      "loss": 1.274,
+      "step": 9701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.149432e-05,
+      "loss": 0.936,
+      "step": 9702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1494051428571429e-05,
+      "loss": 0.9595,
+      "step": 9703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1493782857142857e-05,
+      "loss": 1.3181,
+      "step": 9704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1493514285714287e-05,
+      "loss": 1.0358,
+      "step": 9705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1493245714285713e-05,
+      "loss": 1.5149,
+      "step": 9706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1492977142857143e-05,
+      "loss": 0.9733,
+      "step": 9707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1492708571428572e-05,
+      "loss": 1.4884,
+      "step": 9708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1492440000000001e-05,
+      "loss": 1.3589,
+      "step": 9709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1492171428571428e-05,
+      "loss": 1.4435,
+      "step": 9710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1491902857142858e-05,
+      "loss": 1.1483,
+      "step": 9711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.1491634285714286e-05,
+      "loss": 1.4389,
+      "step": 9712
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1491365714285714e-05,
+      "loss": 1.1003,
+      "step": 9713
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1491097142857144e-05,
+      "loss": 1.3321,
+      "step": 9714
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1490828571428572e-05,
+      "loss": 1.3883,
+      "step": 9715
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.149056e-05,
+      "loss": 1.1281,
+      "step": 9716
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1490291428571428e-05,
+      "loss": 1.5731,
+      "step": 9717
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1490022857142858e-05,
+      "loss": 1.3226,
+      "step": 9718
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1489754285714287e-05,
+      "loss": 1.4148,
+      "step": 9719
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1489485714285715e-05,
+      "loss": 1.5391,
+      "step": 9720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1489217142857143e-05,
+      "loss": 1.1543,
+      "step": 9721
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1488948571428573e-05,
+      "loss": 1.0691,
+      "step": 9722
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.148868e-05,
+      "loss": 1.0194,
+      "step": 9723
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1488411428571429e-05,
+      "loss": 1.2086,
+      "step": 9724
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1488142857142857e-05,
+      "loss": 1.1836,
+      "step": 9725
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1487874285714285e-05,
+      "loss": 1.1384,
+      "step": 9726
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1487605714285715e-05,
+      "loss": 1.2477,
+      "step": 9727
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1487337142857144e-05,
+      "loss": 1.4406,
+      "step": 9728
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1487068571428572e-05,
+      "loss": 1.3018,
+      "step": 9729
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.14868e-05,
+      "loss": 1.5104,
+      "step": 9730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.148653142857143e-05,
+      "loss": 1.3085,
+      "step": 9731
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1486262857142858e-05,
+      "loss": 1.1423,
+      "step": 9732
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1485994285714286e-05,
+      "loss": 1.8462,
+      "step": 9733
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1485725714285714e-05,
+      "loss": 1.2632,
+      "step": 9734
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1485457142857144e-05,
+      "loss": 1.1002,
+      "step": 9735
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.148518857142857e-05,
+      "loss": 1.1599,
+      "step": 9736
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.148492e-05,
+      "loss": 1.2434,
+      "step": 9737
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1484651428571429e-05,
+      "loss": 1.5394,
+      "step": 9738
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1484382857142857e-05,
+      "loss": 1.1922,
+      "step": 9739
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1484114285714287e-05,
+      "loss": 1.1608,
+      "step": 9740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1483845714285715e-05,
+      "loss": 1.3498,
+      "step": 9741
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1483577142857143e-05,
+      "loss": 1.3385,
+      "step": 9742
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1483308571428571e-05,
+      "loss": 1.2503,
+      "step": 9743
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1483040000000001e-05,
+      "loss": 1.3002,
+      "step": 9744
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.148277142857143e-05,
+      "loss": 1.1116,
+      "step": 9745
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1482502857142857e-05,
+      "loss": 1.6109,
+      "step": 9746
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1482234285714286e-05,
+      "loss": 1.4422,
+      "step": 9747
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1481965714285716e-05,
+      "loss": 0.9984,
+      "step": 9748
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1481697142857142e-05,
+      "loss": 1.7208,
+      "step": 9749
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1481428571428572e-05,
+      "loss": 1.223,
+      "step": 9750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.148116e-05,
+      "loss": 1.2505,
+      "step": 9751
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1480891428571428e-05,
+      "loss": 1.5014,
+      "step": 9752
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1480622857142858e-05,
+      "loss": 1.3596,
+      "step": 9753
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1480354285714286e-05,
+      "loss": 0.9759,
+      "step": 9754
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1480085714285716e-05,
+      "loss": 1.1064,
+      "step": 9755
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1479817142857143e-05,
+      "loss": 1.0027,
+      "step": 9756
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1479548571428572e-05,
+      "loss": 0.9439,
+      "step": 9757
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.147928e-05,
+      "loss": 1.4425,
+      "step": 9758
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1479011428571429e-05,
+      "loss": 1.5384,
+      "step": 9759
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1478742857142857e-05,
+      "loss": 1.1972,
+      "step": 9760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1478474285714287e-05,
+      "loss": 1.1538,
+      "step": 9761
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1478205714285713e-05,
+      "loss": 1.7088,
+      "step": 9762
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1477937142857143e-05,
+      "loss": 1.1883,
+      "step": 9763
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1477668571428571e-05,
+      "loss": 0.8924,
+      "step": 9764
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.14774e-05,
+      "loss": 1.1978,
+      "step": 9765
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.147713142857143e-05,
+      "loss": 1.7723,
+      "step": 9766
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1476862857142858e-05,
+      "loss": 0.8769,
+      "step": 9767
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1476594285714287e-05,
+      "loss": 1.5917,
+      "step": 9768
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1476325714285714e-05,
+      "loss": 1.5049,
+      "step": 9769
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1476057142857144e-05,
+      "loss": 1.5,
+      "step": 9770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1475788571428572e-05,
+      "loss": 1.7897,
+      "step": 9771
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.147552e-05,
+      "loss": 0.8035,
+      "step": 9772
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1475251428571428e-05,
+      "loss": 1.4081,
+      "step": 9773
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1474982857142858e-05,
+      "loss": 1.0976,
+      "step": 9774
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1474714285714285e-05,
+      "loss": 1.342,
+      "step": 9775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1474445714285715e-05,
+      "loss": 1.34,
+      "step": 9776
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1474177142857143e-05,
+      "loss": 1.6442,
+      "step": 9777
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1473908571428573e-05,
+      "loss": 1.1626,
+      "step": 9778
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.147364e-05,
+      "loss": 1.4275,
+      "step": 9779
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1473371428571429e-05,
+      "loss": 1.3478,
+      "step": 9780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1473102857142859e-05,
+      "loss": 0.9875,
+      "step": 9781
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1472834285714285e-05,
+      "loss": 1.4622,
+      "step": 9782
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1472565714285715e-05,
+      "loss": 1.1731,
+      "step": 9783
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1472297142857143e-05,
+      "loss": 0.8859,
+      "step": 9784
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1472028571428572e-05,
+      "loss": 1.2906,
+      "step": 9785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.147176e-05,
+      "loss": 1.3649,
+      "step": 9786
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.147149142857143e-05,
+      "loss": 1.593,
+      "step": 9787
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1471222857142856e-05,
+      "loss": 0.9494,
+      "step": 9788
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1470954285714286e-05,
+      "loss": 1.3775,
+      "step": 9789
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1470685714285714e-05,
+      "loss": 1.1697,
+      "step": 9790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1470417142857144e-05,
+      "loss": 1.0675,
+      "step": 9791
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1470148571428572e-05,
+      "loss": 1.5128,
+      "step": 9792
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.146988e-05,
+      "loss": 1.3803,
+      "step": 9793
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.146961142857143e-05,
+      "loss": 1.1719,
+      "step": 9794
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1469342857142857e-05,
+      "loss": 1.6896,
+      "step": 9795
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1469074285714287e-05,
+      "loss": 1.4348,
+      "step": 9796
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1468805714285715e-05,
+      "loss": 1.4115,
+      "step": 9797
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1468537142857143e-05,
+      "loss": 1.2235,
+      "step": 9798
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1468268571428571e-05,
+      "loss": 1.0198,
+      "step": 9799
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1468000000000001e-05,
+      "loss": 0.8611,
+      "step": 9800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.146773142857143e-05,
+      "loss": 1.1566,
+      "step": 9801
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1467462857142857e-05,
+      "loss": 1.2095,
+      "step": 9802
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1467194285714286e-05,
+      "loss": 0.9516,
+      "step": 9803
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1466925714285715e-05,
+      "loss": 1.2661,
+      "step": 9804
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1466657142857144e-05,
+      "loss": 1.1865,
+      "step": 9805
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1466388571428572e-05,
+      "loss": 1.1219,
+      "step": 9806
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1466120000000002e-05,
+      "loss": 1.3064,
+      "step": 9807
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1465851428571428e-05,
+      "loss": 1.5902,
+      "step": 9808
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1465582857142858e-05,
+      "loss": 1.8614,
+      "step": 9809
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1465314285714286e-05,
+      "loss": 1.5321,
+      "step": 9810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1465045714285714e-05,
+      "loss": 1.6342,
+      "step": 9811
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1464777142857142e-05,
+      "loss": 1.1728,
+      "step": 9812
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1464508571428572e-05,
+      "loss": 1.0945,
+      "step": 9813
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.146424e-05,
+      "loss": 1.0691,
+      "step": 9814
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1463971428571429e-05,
+      "loss": 1.4174,
+      "step": 9815
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1463702857142857e-05,
+      "loss": 1.0345,
+      "step": 9816
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1463434285714287e-05,
+      "loss": 1.2673,
+      "step": 9817
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1463165714285715e-05,
+      "loss": 1.2059,
+      "step": 9818
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1462897142857143e-05,
+      "loss": 1.1285,
+      "step": 9819
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1462628571428573e-05,
+      "loss": 1.1778,
+      "step": 9820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.146236e-05,
+      "loss": 1.2695,
+      "step": 9821
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.146209142857143e-05,
+      "loss": 1.3475,
+      "step": 9822
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1461822857142858e-05,
+      "loss": 1.5968,
+      "step": 9823
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1461554285714287e-05,
+      "loss": 1.1972,
+      "step": 9824
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1461285714285714e-05,
+      "loss": 1.4187,
+      "step": 9825
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1461017142857144e-05,
+      "loss": 1.4465,
+      "step": 9826
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1460748571428572e-05,
+      "loss": 1.3781,
+      "step": 9827
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.146048e-05,
+      "loss": 1.2694,
+      "step": 9828
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1460211428571428e-05,
+      "loss": 1.4687,
+      "step": 9829
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1459942857142858e-05,
+      "loss": 1.0254,
+      "step": 9830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1459674285714286e-05,
+      "loss": 1.2734,
+      "step": 9831
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1459405714285714e-05,
+      "loss": 1.4311,
+      "step": 9832
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1459137142857144e-05,
+      "loss": 1.2566,
+      "step": 9833
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.145886857142857e-05,
+      "loss": 1.0528,
+      "step": 9834
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.14586e-05,
+      "loss": 1.4564,
+      "step": 9835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1458331428571429e-05,
+      "loss": 1.2991,
+      "step": 9836
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1458062857142859e-05,
+      "loss": 1.0384,
+      "step": 9837
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1457794285714285e-05,
+      "loss": 1.1823,
+      "step": 9838
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1457525714285715e-05,
+      "loss": 0.8585,
+      "step": 9839
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1457257142857143e-05,
+      "loss": 1.4841,
+      "step": 9840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1456988571428571e-05,
+      "loss": 1.1966,
+      "step": 9841
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.145672e-05,
+      "loss": 1.323,
+      "step": 9842
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.145645142857143e-05,
+      "loss": 0.8402,
+      "step": 9843
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1456182857142858e-05,
+      "loss": 1.4594,
+      "step": 9844
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1455914285714286e-05,
+      "loss": 1.3652,
+      "step": 9845
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1455645714285716e-05,
+      "loss": 1.1895,
+      "step": 9846
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1455377142857144e-05,
+      "loss": 1.1266,
+      "step": 9847
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1455108571428572e-05,
+      "loss": 1.2465,
+      "step": 9848
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.145484e-05,
+      "loss": 1.0239,
+      "step": 9849
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.145457142857143e-05,
+      "loss": 1.4811,
+      "step": 9850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1454302857142857e-05,
+      "loss": 0.8344,
+      "step": 9851
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1454034285714286e-05,
+      "loss": 1.3166,
+      "step": 9852
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1453765714285715e-05,
+      "loss": 1.2458,
+      "step": 9853
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1453497142857143e-05,
+      "loss": 1.5208,
+      "step": 9854
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1453228571428571e-05,
+      "loss": 1.2077,
+      "step": 9855
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1452960000000001e-05,
+      "loss": 1.305,
+      "step": 9856
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1452691428571429e-05,
+      "loss": 1.2368,
+      "step": 9857
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1452422857142857e-05,
+      "loss": 1.3519,
+      "step": 9858
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1452154285714287e-05,
+      "loss": 1.4446,
+      "step": 9859
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1451885714285715e-05,
+      "loss": 1.5264,
+      "step": 9860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1451617142857143e-05,
+      "loss": 1.1809,
+      "step": 9861
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1451348571428572e-05,
+      "loss": 1.2164,
+      "step": 9862
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1451080000000002e-05,
+      "loss": 0.9485,
+      "step": 9863
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1450811428571428e-05,
+      "loss": 1.2378,
+      "step": 9864
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1450542857142858e-05,
+      "loss": 1.1064,
+      "step": 9865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1450274285714286e-05,
+      "loss": 1.1452,
+      "step": 9866
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1450005714285714e-05,
+      "loss": 1.2678,
+      "step": 9867
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1449737142857142e-05,
+      "loss": 1.1119,
+      "step": 9868
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1449468571428572e-05,
+      "loss": 1.4031,
+      "step": 9869
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.14492e-05,
+      "loss": 1.4805,
+      "step": 9870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1448931428571429e-05,
+      "loss": 1.8311,
+      "step": 9871
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1448662857142858e-05,
+      "loss": 1.8096,
+      "step": 9872
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1448394285714287e-05,
+      "loss": 1.5022,
+      "step": 9873
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1448125714285715e-05,
+      "loss": 1.2025,
+      "step": 9874
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1447857142857143e-05,
+      "loss": 1.0153,
+      "step": 9875
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1447588571428573e-05,
+      "loss": 1.0575,
+      "step": 9876
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.144732e-05,
+      "loss": 1.2529,
+      "step": 9877
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.144705142857143e-05,
+      "loss": 0.9599,
+      "step": 9878
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1446782857142857e-05,
+      "loss": 1.1571,
+      "step": 9879
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1446514285714286e-05,
+      "loss": 1.0851,
+      "step": 9880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1446245714285714e-05,
+      "loss": 1.462,
+      "step": 9881
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1445977142857144e-05,
+      "loss": 1.2006,
+      "step": 9882
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1445708571428572e-05,
+      "loss": 1.3031,
+      "step": 9883
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.144544e-05,
+      "loss": 1.1642,
+      "step": 9884
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.144517142857143e-05,
+      "loss": 1.618,
+      "step": 9885
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1444902857142858e-05,
+      "loss": 1.2502,
+      "step": 9886
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1444634285714286e-05,
+      "loss": 1.0722,
+      "step": 9887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1444365714285714e-05,
+      "loss": 1.0444,
+      "step": 9888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1444097142857144e-05,
+      "loss": 1.2409,
+      "step": 9889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.144382857142857e-05,
+      "loss": 1.8181,
+      "step": 9890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.144356e-05,
+      "loss": 1.2533,
+      "step": 9891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1443291428571429e-05,
+      "loss": 1.6638,
+      "step": 9892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1443022857142857e-05,
+      "loss": 1.3783,
+      "step": 9893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1442754285714285e-05,
+      "loss": 1.4629,
+      "step": 9894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1442485714285715e-05,
+      "loss": 1.6564,
+      "step": 9895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1442217142857143e-05,
+      "loss": 1.391,
+      "step": 9896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1441948571428571e-05,
+      "loss": 1.0937,
+      "step": 9897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1441680000000001e-05,
+      "loss": 0.8789,
+      "step": 9898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.144141142857143e-05,
+      "loss": 1.2042,
+      "step": 9899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1441142857142858e-05,
+      "loss": 1.2282,
+      "step": 9900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1440874285714286e-05,
+      "loss": 0.9373,
+      "step": 9901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1440605714285716e-05,
+      "loss": 0.8074,
+      "step": 9902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1440337142857142e-05,
+      "loss": 1.3687,
+      "step": 9903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1440068571428572e-05,
+      "loss": 1.2623,
+      "step": 9904
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.14398e-05,
+      "loss": 1.7105,
+      "step": 9905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.143953142857143e-05,
+      "loss": 0.9517,
+      "step": 9906
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1439262857142856e-05,
+      "loss": 1.0121,
+      "step": 9907
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1438994285714286e-05,
+      "loss": 1.3785,
+      "step": 9908
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1438725714285715e-05,
+      "loss": 1.3685,
+      "step": 9909
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1438457142857143e-05,
+      "loss": 1.0817,
+      "step": 9910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1438188571428573e-05,
+      "loss": 1.4243,
+      "step": 9911
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.143792e-05,
+      "loss": 1.4268,
+      "step": 9912
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1437651428571429e-05,
+      "loss": 1.0816,
+      "step": 9913
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1437382857142857e-05,
+      "loss": 1.6544,
+      "step": 9914
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1437114285714287e-05,
+      "loss": 1.5908,
+      "step": 9915
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1436845714285713e-05,
+      "loss": 1.072,
+      "step": 9916
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1436577142857143e-05,
+      "loss": 1.0128,
+      "step": 9917
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1436308571428572e-05,
+      "loss": 1.2374,
+      "step": 9918
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1436040000000001e-05,
+      "loss": 1.3808,
+      "step": 9919
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1435771428571428e-05,
+      "loss": 1.4266,
+      "step": 9920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1435502857142858e-05,
+      "loss": 0.8341,
+      "step": 9921
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1435234285714286e-05,
+      "loss": 1.2304,
+      "step": 9922
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1434965714285714e-05,
+      "loss": 1.1597,
+      "step": 9923
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1434697142857144e-05,
+      "loss": 1.0003,
+      "step": 9924
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1434428571428572e-05,
+      "loss": 1.1017,
+      "step": 9925
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.143416e-05,
+      "loss": 1.121,
+      "step": 9926
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1433891428571428e-05,
+      "loss": 1.7909,
+      "step": 9927
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1433622857142858e-05,
+      "loss": 1.2418,
+      "step": 9928
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1433354285714287e-05,
+      "loss": 1.2948,
+      "step": 9929
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1433085714285715e-05,
+      "loss": 1.5209,
+      "step": 9930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1432817142857143e-05,
+      "loss": 1.181,
+      "step": 9931
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1432548571428573e-05,
+      "loss": 1.1161,
+      "step": 9932
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.143228e-05,
+      "loss": 1.2315,
+      "step": 9933
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1432011428571429e-05,
+      "loss": 0.8908,
+      "step": 9934
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1431742857142857e-05,
+      "loss": 1.3572,
+      "step": 9935
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1431474285714285e-05,
+      "loss": 1.1956,
+      "step": 9936
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1431205714285715e-05,
+      "loss": 1.4706,
+      "step": 9937
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1430937142857144e-05,
+      "loss": 1.4246,
+      "step": 9938
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1430668571428572e-05,
+      "loss": 1.1238,
+      "step": 9939
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.14304e-05,
+      "loss": 0.9083,
+      "step": 9940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.143013142857143e-05,
+      "loss": 1.3312,
+      "step": 9941
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1429862857142858e-05,
+      "loss": 1.1265,
+      "step": 9942
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1429594285714286e-05,
+      "loss": 1.2994,
+      "step": 9943
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1429325714285714e-05,
+      "loss": 1.3609,
+      "step": 9944
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1429057142857144e-05,
+      "loss": 1.1884,
+      "step": 9945
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.142878857142857e-05,
+      "loss": 1.1704,
+      "step": 9946
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.142852e-05,
+      "loss": 1.2565,
+      "step": 9947
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1428251428571429e-05,
+      "loss": 1.1977,
+      "step": 9948
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1427982857142857e-05,
+      "loss": 1.4656,
+      "step": 9949
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1427714285714287e-05,
+      "loss": 1.334,
+      "step": 9950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1427445714285715e-05,
+      "loss": 1.5837,
+      "step": 9951
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1427177142857145e-05,
+      "loss": 0.9633,
+      "step": 9952
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1426908571428571e-05,
+      "loss": 1.1895,
+      "step": 9953
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1426640000000001e-05,
+      "loss": 1.0482,
+      "step": 9954
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.142637142857143e-05,
+      "loss": 0.9303,
+      "step": 9955
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1426102857142857e-05,
+      "loss": 1.6567,
+      "step": 9956
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1425834285714286e-05,
+      "loss": 1.0621,
+      "step": 9957
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1425565714285716e-05,
+      "loss": 1.6659,
+      "step": 9958
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1425297142857142e-05,
+      "loss": 1.7175,
+      "step": 9959
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1425028571428572e-05,
+      "loss": 1.3803,
+      "step": 9960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.142476e-05,
+      "loss": 1.2208,
+      "step": 9961
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1424491428571428e-05,
+      "loss": 1.4403,
+      "step": 9962
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1424222857142858e-05,
+      "loss": 1.6419,
+      "step": 9963
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1423954285714286e-05,
+      "loss": 1.3323,
+      "step": 9964
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1423685714285716e-05,
+      "loss": 1.4404,
+      "step": 9965
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1423417142857143e-05,
+      "loss": 1.4812,
+      "step": 9966
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1423148571428572e-05,
+      "loss": 0.8756,
+      "step": 9967
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.142288e-05,
+      "loss": 1.1789,
+      "step": 9968
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1422611428571429e-05,
+      "loss": 0.9601,
+      "step": 9969
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1422342857142857e-05,
+      "loss": 1.4174,
+      "step": 9970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1422074285714287e-05,
+      "loss": 1.3316,
+      "step": 9971
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1421805714285713e-05,
+      "loss": 1.0345,
+      "step": 9972
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1421537142857143e-05,
+      "loss": 1.4385,
+      "step": 9973
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1421268571428571e-05,
+      "loss": 1.3089,
+      "step": 9974
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1421000000000001e-05,
+      "loss": 1.1732,
+      "step": 9975
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.142073142857143e-05,
+      "loss": 1.4229,
+      "step": 9976
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1420462857142858e-05,
+      "loss": 0.799,
+      "step": 9977
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1420194285714287e-05,
+      "loss": 1.4879,
+      "step": 9978
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1419925714285714e-05,
+      "loss": 1.2134,
+      "step": 9979
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1419657142857144e-05,
+      "loss": 1.3607,
+      "step": 9980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1419388571428572e-05,
+      "loss": 1.4448,
+      "step": 9981
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.141912e-05,
+      "loss": 1.2628,
+      "step": 9982
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1418851428571428e-05,
+      "loss": 1.2767,
+      "step": 9983
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1418582857142858e-05,
+      "loss": 1.0437,
+      "step": 9984
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1418314285714285e-05,
+      "loss": 0.7709,
+      "step": 9985
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1418045714285715e-05,
+      "loss": 1.1315,
+      "step": 9986
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1417777142857143e-05,
+      "loss": 1.5928,
+      "step": 9987
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1417508571428573e-05,
+      "loss": 1.0679,
+      "step": 9988
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.141724e-05,
+      "loss": 1.0003,
+      "step": 9989
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1416971428571429e-05,
+      "loss": 1.3759,
+      "step": 9990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1416702857142859e-05,
+      "loss": 1.36,
+      "step": 9991
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1416434285714285e-05,
+      "loss": 1.3426,
+      "step": 9992
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1416165714285715e-05,
+      "loss": 1.242,
+      "step": 9993
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1415897142857143e-05,
+      "loss": 1.1067,
+      "step": 9994
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1415628571428572e-05,
+      "loss": 1.2806,
+      "step": 9995
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.141536e-05,
+      "loss": 1.1003,
+      "step": 9996
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.141509142857143e-05,
+      "loss": 1.4542,
+      "step": 9997
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1414822857142858e-05,
+      "loss": 1.1427,
+      "step": 9998
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1414554285714286e-05,
+      "loss": 1.8779,
+      "step": 9999
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1414285714285714e-05,
+      "loss": 1.2404,
+      "step": 10000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1414017142857144e-05,
+      "loss": 1.041,
+      "step": 10001
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1413748571428572e-05,
+      "loss": 0.9161,
+      "step": 10002
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.141348e-05,
+      "loss": 1.1565,
+      "step": 10003
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.141321142857143e-05,
+      "loss": 1.4142,
+      "step": 10004
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1412942857142857e-05,
+      "loss": 1.1273,
+      "step": 10005
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1412674285714287e-05,
+      "loss": 1.3634,
+      "step": 10006
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1412405714285715e-05,
+      "loss": 1.5246,
+      "step": 10007
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1412137142857143e-05,
+      "loss": 1.3418,
+      "step": 10008
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1411868571428571e-05,
+      "loss": 1.762,
+      "step": 10009
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1411600000000001e-05,
+      "loss": 1.3087,
+      "step": 10010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.141133142857143e-05,
+      "loss": 1.0743,
+      "step": 10011
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1411062857142857e-05,
+      "loss": 1.3304,
+      "step": 10012
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1410794285714286e-05,
+      "loss": 1.0335,
+      "step": 10013
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1410525714285715e-05,
+      "loss": 1.529,
+      "step": 10014
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1410257142857144e-05,
+      "loss": 1.3752,
+      "step": 10015
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1409988571428572e-05,
+      "loss": 1.6126,
+      "step": 10016
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1409720000000002e-05,
+      "loss": 1.3665,
+      "step": 10017
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1409451428571428e-05,
+      "loss": 0.826,
+      "step": 10018
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1409182857142858e-05,
+      "loss": 1.4429,
+      "step": 10019
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1408914285714286e-05,
+      "loss": 0.9042,
+      "step": 10020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1408645714285716e-05,
+      "loss": 1.223,
+      "step": 10021
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1408377142857142e-05,
+      "loss": 1.1757,
+      "step": 10022
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1408108571428572e-05,
+      "loss": 1.3417,
+      "step": 10023
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.140784e-05,
+      "loss": 1.4261,
+      "step": 10024
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1407571428571429e-05,
+      "loss": 1.2317,
+      "step": 10025
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1407302857142857e-05,
+      "loss": 1.0503,
+      "step": 10026
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1407034285714287e-05,
+      "loss": 1.3095,
+      "step": 10027
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1406765714285715e-05,
+      "loss": 1.0677,
+      "step": 10028
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1406497142857143e-05,
+      "loss": 1.9256,
+      "step": 10029
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1406228571428573e-05,
+      "loss": 1.1965,
+      "step": 10030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.140596e-05,
+      "loss": 1.8918,
+      "step": 10031
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.140569142857143e-05,
+      "loss": 1.5715,
+      "step": 10032
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1405422857142858e-05,
+      "loss": 1.4874,
+      "step": 10033
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1405154285714287e-05,
+      "loss": 1.5708,
+      "step": 10034
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1404885714285714e-05,
+      "loss": 1.2002,
+      "step": 10035
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1404617142857144e-05,
+      "loss": 1.5122,
+      "step": 10036
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1404348571428572e-05,
+      "loss": 1.5094,
+      "step": 10037
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.140408e-05,
+      "loss": 1.356,
+      "step": 10038
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1403811428571428e-05,
+      "loss": 1.0278,
+      "step": 10039
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1403542857142858e-05,
+      "loss": 1.338,
+      "step": 10040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1403274285714286e-05,
+      "loss": 1.1198,
+      "step": 10041
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1403005714285714e-05,
+      "loss": 1.0642,
+      "step": 10042
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1402737142857144e-05,
+      "loss": 1.2618,
+      "step": 10043
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1402468571428573e-05,
+      "loss": 0.8593,
+      "step": 10044
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.14022e-05,
+      "loss": 1.2299,
+      "step": 10045
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1401931428571429e-05,
+      "loss": 1.0813,
+      "step": 10046
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1401662857142859e-05,
+      "loss": 1.0211,
+      "step": 10047
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1401394285714285e-05,
+      "loss": 1.1644,
+      "step": 10048
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1401125714285715e-05,
+      "loss": 1.4209,
+      "step": 10049
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1400857142857143e-05,
+      "loss": 1.3143,
+      "step": 10050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1400588571428571e-05,
+      "loss": 1.0647,
+      "step": 10051
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.140032e-05,
+      "loss": 1.3374,
+      "step": 10052
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.140005142857143e-05,
+      "loss": 0.8894,
+      "step": 10053
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1399782857142858e-05,
+      "loss": 1.3228,
+      "step": 10054
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1399514285714286e-05,
+      "loss": 1.1062,
+      "step": 10055
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1399245714285716e-05,
+      "loss": 1.5102,
+      "step": 10056
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1398977142857144e-05,
+      "loss": 1.2141,
+      "step": 10057
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1398708571428572e-05,
+      "loss": 0.9556,
+      "step": 10058
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.139844e-05,
+      "loss": 1.0761,
+      "step": 10059
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.139817142857143e-05,
+      "loss": 1.1191,
+      "step": 10060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1397902857142857e-05,
+      "loss": 0.9319,
+      "step": 10061
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1397634285714286e-05,
+      "loss": 1.4064,
+      "step": 10062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1397365714285715e-05,
+      "loss": 0.9247,
+      "step": 10063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1397097142857143e-05,
+      "loss": 1.5475,
+      "step": 10064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1396828571428571e-05,
+      "loss": 1.2512,
+      "step": 10065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1396560000000001e-05,
+      "loss": 1.2034,
+      "step": 10066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1396291428571429e-05,
+      "loss": 1.4348,
+      "step": 10067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1396022857142857e-05,
+      "loss": 1.306,
+      "step": 10068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1395754285714287e-05,
+      "loss": 1.2518,
+      "step": 10069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1395485714285715e-05,
+      "loss": 1.23,
+      "step": 10070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1395217142857143e-05,
+      "loss": 1.2755,
+      "step": 10071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1394948571428572e-05,
+      "loss": 0.9691,
+      "step": 10072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1394680000000001e-05,
+      "loss": 1.2811,
+      "step": 10073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1394411428571428e-05,
+      "loss": 1.4328,
+      "step": 10074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1394142857142858e-05,
+      "loss": 1.2344,
+      "step": 10075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1393874285714286e-05,
+      "loss": 1.5588,
+      "step": 10076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1393605714285714e-05,
+      "loss": 1.4004,
+      "step": 10077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1393337142857142e-05,
+      "loss": 1.1083,
+      "step": 10078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1393068571428572e-05,
+      "loss": 1.0475,
+      "step": 10079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.13928e-05,
+      "loss": 1.241,
+      "step": 10080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1392531428571429e-05,
+      "loss": 1.134,
+      "step": 10081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1392262857142858e-05,
+      "loss": 1.3173,
+      "step": 10082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1391994285714287e-05,
+      "loss": 1.5445,
+      "step": 10083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1391725714285715e-05,
+      "loss": 1.1901,
+      "step": 10084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1391457142857143e-05,
+      "loss": 1.0752,
+      "step": 10085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1391188571428573e-05,
+      "loss": 1.0687,
+      "step": 10086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.139092e-05,
+      "loss": 1.4686,
+      "step": 10087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.139065142857143e-05,
+      "loss": 0.8121,
+      "step": 10088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1390382857142857e-05,
+      "loss": 0.9171,
+      "step": 10089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1390114285714286e-05,
+      "loss": 1.4409,
+      "step": 10090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1389845714285714e-05,
+      "loss": 1.1763,
+      "step": 10091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1389577142857144e-05,
+      "loss": 1.7842,
+      "step": 10092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1389308571428572e-05,
+      "loss": 1.123,
+      "step": 10093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.138904e-05,
+      "loss": 1.4261,
+      "step": 10094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.138877142857143e-05,
+      "loss": 1.0854,
+      "step": 10095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1388502857142858e-05,
+      "loss": 1.5233,
+      "step": 10096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1388234285714286e-05,
+      "loss": 1.0272,
+      "step": 10097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1387965714285714e-05,
+      "loss": 1.0529,
+      "step": 10098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1387697142857144e-05,
+      "loss": 1.1,
+      "step": 10099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.138742857142857e-05,
+      "loss": 1.0788,
+      "step": 10100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.138716e-05,
+      "loss": 1.1429,
+      "step": 10101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1386891428571429e-05,
+      "loss": 1.4529,
+      "step": 10102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1386622857142859e-05,
+      "loss": 1.3165,
+      "step": 10103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1386354285714285e-05,
+      "loss": 1.1426,
+      "step": 10104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1386085714285715e-05,
+      "loss": 1.0919,
+      "step": 10105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1385817142857143e-05,
+      "loss": 1.2678,
+      "step": 10106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1385548571428571e-05,
+      "loss": 1.2114,
+      "step": 10107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1385280000000001e-05,
+      "loss": 1.5829,
+      "step": 10108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.138501142857143e-05,
+      "loss": 1.2026,
+      "step": 10109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1384742857142858e-05,
+      "loss": 1.1188,
+      "step": 10110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1384474285714286e-05,
+      "loss": 1.2559,
+      "step": 10111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1384205714285716e-05,
+      "loss": 1.2616,
+      "step": 10112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1383937142857142e-05,
+      "loss": 1.0478,
+      "step": 10113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1383668571428572e-05,
+      "loss": 1.6688,
+      "step": 10114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.13834e-05,
+      "loss": 0.9574,
+      "step": 10115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.138313142857143e-05,
+      "loss": 1.1709,
+      "step": 10116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1382862857142856e-05,
+      "loss": 0.8918,
+      "step": 10117
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1382594285714286e-05,
+      "loss": 1.2339,
+      "step": 10118
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1382325714285715e-05,
+      "loss": 1.5095,
+      "step": 10119
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1382057142857143e-05,
+      "loss": 1.444,
+      "step": 10120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1381788571428573e-05,
+      "loss": 1.359,
+      "step": 10121
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.138152e-05,
+      "loss": 1.2282,
+      "step": 10122
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1381251428571429e-05,
+      "loss": 0.9022,
+      "step": 10123
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1380982857142857e-05,
+      "loss": 1.1801,
+      "step": 10124
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1380714285714287e-05,
+      "loss": 1.0156,
+      "step": 10125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1380445714285715e-05,
+      "loss": 1.4465,
+      "step": 10126
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1380177142857143e-05,
+      "loss": 1.3176,
+      "step": 10127
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1379908571428572e-05,
+      "loss": 1.3252,
+      "step": 10128
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1379640000000001e-05,
+      "loss": 1.2511,
+      "step": 10129
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1379371428571428e-05,
+      "loss": 1.3033,
+      "step": 10130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1379102857142858e-05,
+      "loss": 1.6662,
+      "step": 10131
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1378834285714286e-05,
+      "loss": 1.3173,
+      "step": 10132
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1378565714285714e-05,
+      "loss": 1.3682,
+      "step": 10133
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1378297142857144e-05,
+      "loss": 1.3007,
+      "step": 10134
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1378028571428572e-05,
+      "loss": 1.3201,
+      "step": 10135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.137776e-05,
+      "loss": 1.1408,
+      "step": 10136
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1377491428571428e-05,
+      "loss": 0.8536,
+      "step": 10137
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1377222857142858e-05,
+      "loss": 1.1834,
+      "step": 10138
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1376954285714287e-05,
+      "loss": 1.3071,
+      "step": 10139
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1376685714285715e-05,
+      "loss": 1.0833,
+      "step": 10140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1376417142857143e-05,
+      "loss": 0.9809,
+      "step": 10141
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1376148571428573e-05,
+      "loss": 0.8275,
+      "step": 10142
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.137588e-05,
+      "loss": 1.6185,
+      "step": 10143
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1375611428571429e-05,
+      "loss": 1.16,
+      "step": 10144
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1375342857142857e-05,
+      "loss": 1.271,
+      "step": 10145
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1375074285714285e-05,
+      "loss": 0.9332,
+      "step": 10146
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1374805714285715e-05,
+      "loss": 1.3763,
+      "step": 10147
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1374537142857144e-05,
+      "loss": 1.4027,
+      "step": 10148
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1374268571428573e-05,
+      "loss": 0.9742,
+      "step": 10149
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1374e-05,
+      "loss": 1.3016,
+      "step": 10150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.137373142857143e-05,
+      "loss": 1.1469,
+      "step": 10151
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1373462857142858e-05,
+      "loss": 1.1465,
+      "step": 10152
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1373194285714286e-05,
+      "loss": 1.4894,
+      "step": 10153
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1372925714285714e-05,
+      "loss": 0.9388,
+      "step": 10154
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1372657142857144e-05,
+      "loss": 1.0281,
+      "step": 10155
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.137238857142857e-05,
+      "loss": 1.3418,
+      "step": 10156
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.137212e-05,
+      "loss": 1.393,
+      "step": 10157
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1371851428571429e-05,
+      "loss": 0.9738,
+      "step": 10158
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1371582857142857e-05,
+      "loss": 1.2305,
+      "step": 10159
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1371314285714287e-05,
+      "loss": 0.9978,
+      "step": 10160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1371045714285715e-05,
+      "loss": 1.5179,
+      "step": 10161
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1370777142857145e-05,
+      "loss": 1.2559,
+      "step": 10162
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1370508571428571e-05,
+      "loss": 1.3931,
+      "step": 10163
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1370240000000001e-05,
+      "loss": 0.959,
+      "step": 10164
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.136997142857143e-05,
+      "loss": 1.292,
+      "step": 10165
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1369702857142857e-05,
+      "loss": 1.0109,
+      "step": 10166
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1369434285714286e-05,
+      "loss": 1.3006,
+      "step": 10167
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1369165714285716e-05,
+      "loss": 1.2812,
+      "step": 10168
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1368897142857142e-05,
+      "loss": 1.1429,
+      "step": 10169
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1368628571428572e-05,
+      "loss": 1.4311,
+      "step": 10170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.136836e-05,
+      "loss": 1.125,
+      "step": 10171
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.136809142857143e-05,
+      "loss": 1.1641,
+      "step": 10172
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1367822857142858e-05,
+      "loss": 1.3672,
+      "step": 10173
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1367554285714286e-05,
+      "loss": 0.5837,
+      "step": 10174
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1367285714285716e-05,
+      "loss": 1.1489,
+      "step": 10175
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1367017142857143e-05,
+      "loss": 1.5536,
+      "step": 10176
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1366748571428572e-05,
+      "loss": 1.1008,
+      "step": 10177
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.136648e-05,
+      "loss": 1.5114,
+      "step": 10178
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1366211428571429e-05,
+      "loss": 1.2954,
+      "step": 10179
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1365942857142857e-05,
+      "loss": 1.5899,
+      "step": 10180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1365674285714287e-05,
+      "loss": 1.2808,
+      "step": 10181
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1365405714285713e-05,
+      "loss": 1.146,
+      "step": 10182
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1365137142857143e-05,
+      "loss": 1.362,
+      "step": 10183
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1364868571428571e-05,
+      "loss": 1.2584,
+      "step": 10184
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1364600000000001e-05,
+      "loss": 1.5709,
+      "step": 10185
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.136433142857143e-05,
+      "loss": 1.2545,
+      "step": 10186
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1364062857142858e-05,
+      "loss": 1.2895,
+      "step": 10187
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1363794285714287e-05,
+      "loss": 1.0221,
+      "step": 10188
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1363525714285714e-05,
+      "loss": 1.4004,
+      "step": 10189
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1363257142857144e-05,
+      "loss": 1.3239,
+      "step": 10190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1362988571428572e-05,
+      "loss": 1.4242,
+      "step": 10191
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.136272e-05,
+      "loss": 1.452,
+      "step": 10192
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1362451428571428e-05,
+      "loss": 1.4256,
+      "step": 10193
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1362182857142858e-05,
+      "loss": 1.1348,
+      "step": 10194
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1361914285714286e-05,
+      "loss": 1.2173,
+      "step": 10195
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1361645714285715e-05,
+      "loss": 1.3286,
+      "step": 10196
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1361377142857143e-05,
+      "loss": 1.1425,
+      "step": 10197
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1361108571428573e-05,
+      "loss": 1.0289,
+      "step": 10198
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.136084e-05,
+      "loss": 1.1993,
+      "step": 10199
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1360571428571429e-05,
+      "loss": 1.1942,
+      "step": 10200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1360302857142859e-05,
+      "loss": 1.0369,
+      "step": 10201
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1360034285714285e-05,
+      "loss": 1.4578,
+      "step": 10202
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1359765714285715e-05,
+      "loss": 1.263,
+      "step": 10203
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1359497142857143e-05,
+      "loss": 1.6292,
+      "step": 10204
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1359228571428572e-05,
+      "loss": 0.8562,
+      "step": 10205
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.135896e-05,
+      "loss": 1.215,
+      "step": 10206
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.135869142857143e-05,
+      "loss": 1.358,
+      "step": 10207
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1358422857142858e-05,
+      "loss": 1.0893,
+      "step": 10208
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1358154285714286e-05,
+      "loss": 1.589,
+      "step": 10209
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1357885714285714e-05,
+      "loss": 1.0061,
+      "step": 10210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1357617142857144e-05,
+      "loss": 1.2583,
+      "step": 10211
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.135734857142857e-05,
+      "loss": 1.7516,
+      "step": 10212
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.135708e-05,
+      "loss": 1.4945,
+      "step": 10213
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.135681142857143e-05,
+      "loss": 1.0587,
+      "step": 10214
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1356542857142857e-05,
+      "loss": 1.6511,
+      "step": 10215
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1356274285714287e-05,
+      "loss": 1.4206,
+      "step": 10216
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1356005714285715e-05,
+      "loss": 1.2547,
+      "step": 10217
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1355737142857143e-05,
+      "loss": 1.3002,
+      "step": 10218
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1355468571428571e-05,
+      "loss": 1.6332,
+      "step": 10219
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1355200000000001e-05,
+      "loss": 1.0832,
+      "step": 10220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.135493142857143e-05,
+      "loss": 1.423,
+      "step": 10221
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1354662857142857e-05,
+      "loss": 1.3275,
+      "step": 10222
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1354394285714286e-05,
+      "loss": 1.2241,
+      "step": 10223
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1354125714285715e-05,
+      "loss": 1.2381,
+      "step": 10224
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1353857142857142e-05,
+      "loss": 1.291,
+      "step": 10225
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1353588571428572e-05,
+      "loss": 1.0233,
+      "step": 10226
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1353320000000002e-05,
+      "loss": 1.5806,
+      "step": 10227
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1353051428571428e-05,
+      "loss": 1.116,
+      "step": 10228
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1352782857142858e-05,
+      "loss": 1.8154,
+      "step": 10229
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1352514285714286e-05,
+      "loss": 0.8851,
+      "step": 10230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1352245714285716e-05,
+      "loss": 1.3656,
+      "step": 10231
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1351977142857142e-05,
+      "loss": 1.355,
+      "step": 10232
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1351708571428572e-05,
+      "loss": 1.6146,
+      "step": 10233
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.135144e-05,
+      "loss": 1.5199,
+      "step": 10234
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1351171428571429e-05,
+      "loss": 1.2075,
+      "step": 10235
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1350902857142857e-05,
+      "loss": 1.4069,
+      "step": 10236
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1350634285714287e-05,
+      "loss": 1.4657,
+      "step": 10237
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1350365714285713e-05,
+      "loss": 1.1476,
+      "step": 10238
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1350097142857143e-05,
+      "loss": 1.0519,
+      "step": 10239
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1349828571428573e-05,
+      "loss": 1.0039,
+      "step": 10240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.134956e-05,
+      "loss": 1.0339,
+      "step": 10241
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.134929142857143e-05,
+      "loss": 1.4907,
+      "step": 10242
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1349022857142858e-05,
+      "loss": 1.185,
+      "step": 10243
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1348754285714287e-05,
+      "loss": 1.0191,
+      "step": 10244
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1348485714285714e-05,
+      "loss": 1.041,
+      "step": 10245
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1348217142857144e-05,
+      "loss": 1.2023,
+      "step": 10246
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1347948571428572e-05,
+      "loss": 1.5551,
+      "step": 10247
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.134768e-05,
+      "loss": 1.3115,
+      "step": 10248
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1347411428571428e-05,
+      "loss": 1.0107,
+      "step": 10249
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1347142857142858e-05,
+      "loss": 1.0029,
+      "step": 10250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1346874285714285e-05,
+      "loss": 1.072,
+      "step": 10251
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1346605714285714e-05,
+      "loss": 1.32,
+      "step": 10252
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1346337142857144e-05,
+      "loss": 1.5012,
+      "step": 10253
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1346068571428573e-05,
+      "loss": 1.2435,
+      "step": 10254
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.13458e-05,
+      "loss": 1.5108,
+      "step": 10255
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1345531428571429e-05,
+      "loss": 1.2083,
+      "step": 10256
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1345262857142859e-05,
+      "loss": 0.9339,
+      "step": 10257
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1344994285714285e-05,
+      "loss": 1.2985,
+      "step": 10258
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1344725714285715e-05,
+      "loss": 1.3325,
+      "step": 10259
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1344457142857143e-05,
+      "loss": 1.3657,
+      "step": 10260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1344188571428571e-05,
+      "loss": 1.4402,
+      "step": 10261
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.134392e-05,
+      "loss": 1.4284,
+      "step": 10262
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.134365142857143e-05,
+      "loss": 1.1628,
+      "step": 10263
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1343382857142856e-05,
+      "loss": 1.3054,
+      "step": 10264
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1343114285714286e-05,
+      "loss": 1.3911,
+      "step": 10265
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1342845714285716e-05,
+      "loss": 1.727,
+      "step": 10266
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1342577142857144e-05,
+      "loss": 1.1872,
+      "step": 10267
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1342308571428572e-05,
+      "loss": 1.3542,
+      "step": 10268
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.134204e-05,
+      "loss": 0.7406,
+      "step": 10269
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.134177142857143e-05,
+      "loss": 1.1065,
+      "step": 10270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1341502857142857e-05,
+      "loss": 1.5959,
+      "step": 10271
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1341234285714286e-05,
+      "loss": 1.5065,
+      "step": 10272
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1340965714285715e-05,
+      "loss": 1.1088,
+      "step": 10273
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1340697142857143e-05,
+      "loss": 1.3712,
+      "step": 10274
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1340428571428571e-05,
+      "loss": 1.3666,
+      "step": 10275
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1340160000000001e-05,
+      "loss": 1.2324,
+      "step": 10276
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1339891428571429e-05,
+      "loss": 1.4469,
+      "step": 10277
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1339622857142857e-05,
+      "loss": 1.1928,
+      "step": 10278
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1339354285714287e-05,
+      "loss": 1.1087,
+      "step": 10279
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1339085714285715e-05,
+      "loss": 1.4018,
+      "step": 10280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1338817142857143e-05,
+      "loss": 1.2003,
+      "step": 10281
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1338548571428572e-05,
+      "loss": 1.1992,
+      "step": 10282
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1338280000000001e-05,
+      "loss": 1.4149,
+      "step": 10283
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1338011428571428e-05,
+      "loss": 1.5135,
+      "step": 10284
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1337742857142858e-05,
+      "loss": 1.5634,
+      "step": 10285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1337474285714286e-05,
+      "loss": 1.5887,
+      "step": 10286
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1337205714285714e-05,
+      "loss": 0.8726,
+      "step": 10287
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1336937142857142e-05,
+      "loss": 1.286,
+      "step": 10288
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1336668571428572e-05,
+      "loss": 1.4743,
+      "step": 10289
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.13364e-05,
+      "loss": 1.383,
+      "step": 10290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1336131428571429e-05,
+      "loss": 1.5889,
+      "step": 10291
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1335862857142858e-05,
+      "loss": 0.8957,
+      "step": 10292
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1335594285714287e-05,
+      "loss": 1.4336,
+      "step": 10293
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1335325714285715e-05,
+      "loss": 1.3783,
+      "step": 10294
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1335057142857143e-05,
+      "loss": 0.9636,
+      "step": 10295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1334788571428573e-05,
+      "loss": 1.2172,
+      "step": 10296
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.133452e-05,
+      "loss": 1.6418,
+      "step": 10297
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.133425142857143e-05,
+      "loss": 1.0934,
+      "step": 10298
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1333982857142857e-05,
+      "loss": 1.0519,
+      "step": 10299
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1333714285714287e-05,
+      "loss": 1.3363,
+      "step": 10300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1333445714285714e-05,
+      "loss": 1.4218,
+      "step": 10301
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1333177142857144e-05,
+      "loss": 1.2215,
+      "step": 10302
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1332908571428572e-05,
+      "loss": 1.4583,
+      "step": 10303
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.133264e-05,
+      "loss": 1.4558,
+      "step": 10304
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.133237142857143e-05,
+      "loss": 1.3037,
+      "step": 10305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1332102857142858e-05,
+      "loss": 1.2229,
+      "step": 10306
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1331834285714286e-05,
+      "loss": 1.4515,
+      "step": 10307
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1331565714285714e-05,
+      "loss": 1.0864,
+      "step": 10308
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1331297142857144e-05,
+      "loss": 1.3346,
+      "step": 10309
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.133102857142857e-05,
+      "loss": 1.5561,
+      "step": 10310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.133076e-05,
+      "loss": 1.1297,
+      "step": 10311
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1330491428571429e-05,
+      "loss": 1.5866,
+      "step": 10312
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1330222857142859e-05,
+      "loss": 1.2217,
+      "step": 10313
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1329954285714285e-05,
+      "loss": 1.5993,
+      "step": 10314
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1329685714285715e-05,
+      "loss": 1.1711,
+      "step": 10315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1329417142857143e-05,
+      "loss": 1.5456,
+      "step": 10316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1329148571428571e-05,
+      "loss": 1.394,
+      "step": 10317
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.132888e-05,
+      "loss": 1.1542,
+      "step": 10318
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.132861142857143e-05,
+      "loss": 1.3275,
+      "step": 10319
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1328342857142858e-05,
+      "loss": 1.0122,
+      "step": 10320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1328074285714286e-05,
+      "loss": 0.8047,
+      "step": 10321
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1327805714285716e-05,
+      "loss": 1.4611,
+      "step": 10322
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1327537142857144e-05,
+      "loss": 1.0447,
+      "step": 10323
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1327268571428572e-05,
+      "loss": 1.1822,
+      "step": 10324
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1327e-05,
+      "loss": 1.2603,
+      "step": 10325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.132673142857143e-05,
+      "loss": 1.2858,
+      "step": 10326
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1326462857142856e-05,
+      "loss": 0.9137,
+      "step": 10327
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1326194285714286e-05,
+      "loss": 0.6535,
+      "step": 10328
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1325925714285715e-05,
+      "loss": 1.1679,
+      "step": 10329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1325657142857143e-05,
+      "loss": 0.9383,
+      "step": 10330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1325388571428571e-05,
+      "loss": 1.438,
+      "step": 10331
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.132512e-05,
+      "loss": 1.328,
+      "step": 10332
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1324851428571429e-05,
+      "loss": 1.5963,
+      "step": 10333
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1324582857142857e-05,
+      "loss": 1.2898,
+      "step": 10334
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1324314285714287e-05,
+      "loss": 0.9651,
+      "step": 10335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1324045714285715e-05,
+      "loss": 0.8961,
+      "step": 10336
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1323777142857143e-05,
+      "loss": 1.2649,
+      "step": 10337
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1323508571428572e-05,
+      "loss": 1.5486,
+      "step": 10338
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1323240000000001e-05,
+      "loss": 1.4218,
+      "step": 10339
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1322971428571428e-05,
+      "loss": 1.3517,
+      "step": 10340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1322702857142858e-05,
+      "loss": 0.8901,
+      "step": 10341
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1322434285714286e-05,
+      "loss": 0.9541,
+      "step": 10342
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1322165714285714e-05,
+      "loss": 1.1648,
+      "step": 10343
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1321897142857142e-05,
+      "loss": 0.9935,
+      "step": 10344
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1321628571428572e-05,
+      "loss": 1.2824,
+      "step": 10345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1321360000000002e-05,
+      "loss": 1.2734,
+      "step": 10346
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1321091428571428e-05,
+      "loss": 1.103,
+      "step": 10347
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1320822857142858e-05,
+      "loss": 1.2025,
+      "step": 10348
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1320554285714287e-05,
+      "loss": 1.6266,
+      "step": 10349
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1320285714285715e-05,
+      "loss": 1.2192,
+      "step": 10350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1320017142857143e-05,
+      "loss": 1.3853,
+      "step": 10351
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1319748571428573e-05,
+      "loss": 1.4297,
+      "step": 10352
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131948e-05,
+      "loss": 1.4012,
+      "step": 10353
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1319211428571429e-05,
+      "loss": 1.3743,
+      "step": 10354
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1318942857142857e-05,
+      "loss": 1.7217,
+      "step": 10355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1318674285714285e-05,
+      "loss": 1.1989,
+      "step": 10356
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1318405714285714e-05,
+      "loss": 1.1336,
+      "step": 10357
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1318137142857144e-05,
+      "loss": 1.5469,
+      "step": 10358
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1317868571428573e-05,
+      "loss": 1.4375,
+      "step": 10359
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.13176e-05,
+      "loss": 1.5317,
+      "step": 10360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131733142857143e-05,
+      "loss": 0.9709,
+      "step": 10361
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1317062857142858e-05,
+      "loss": 1.349,
+      "step": 10362
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1316794285714286e-05,
+      "loss": 1.4947,
+      "step": 10363
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1316525714285714e-05,
+      "loss": 1.0673,
+      "step": 10364
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1316257142857144e-05,
+      "loss": 1.4995,
+      "step": 10365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131598857142857e-05,
+      "loss": 1.4581,
+      "step": 10366
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131572e-05,
+      "loss": 1.5074,
+      "step": 10367
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1315451428571429e-05,
+      "loss": 1.1907,
+      "step": 10368
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1315182857142859e-05,
+      "loss": 1.1037,
+      "step": 10369
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1314914285714285e-05,
+      "loss": 1.4819,
+      "step": 10370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1314645714285715e-05,
+      "loss": 0.9233,
+      "step": 10371
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1314377142857145e-05,
+      "loss": 0.9606,
+      "step": 10372
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1314108571428571e-05,
+      "loss": 1.4418,
+      "step": 10373
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1313840000000001e-05,
+      "loss": 1.2795,
+      "step": 10374
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131357142857143e-05,
+      "loss": 1.2143,
+      "step": 10375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1313302857142857e-05,
+      "loss": 1.2537,
+      "step": 10376
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1313034285714286e-05,
+      "loss": 1.1656,
+      "step": 10377
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1312765714285715e-05,
+      "loss": 1.1507,
+      "step": 10378
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1312497142857142e-05,
+      "loss": 1.2832,
+      "step": 10379
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1312228571428572e-05,
+      "loss": 1.4075,
+      "step": 10380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131196e-05,
+      "loss": 1.0353,
+      "step": 10381
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131169142857143e-05,
+      "loss": 1.0912,
+      "step": 10382
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1311422857142856e-05,
+      "loss": 0.9816,
+      "step": 10383
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1311154285714286e-05,
+      "loss": 1.0503,
+      "step": 10384
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1310885714285716e-05,
+      "loss": 1.1869,
+      "step": 10385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1310617142857143e-05,
+      "loss": 0.9851,
+      "step": 10386
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1310348571428572e-05,
+      "loss": 1.373,
+      "step": 10387
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.131008e-05,
+      "loss": 1.1836,
+      "step": 10388
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1309811428571429e-05,
+      "loss": 1.2163,
+      "step": 10389
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1309542857142857e-05,
+      "loss": 1.0947,
+      "step": 10390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1309274285714287e-05,
+      "loss": 1.4841,
+      "step": 10391
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1309005714285713e-05,
+      "loss": 1.1013,
+      "step": 10392
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1308737142857143e-05,
+      "loss": 1.5952,
+      "step": 10393
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1308468571428571e-05,
+      "loss": 1.2891,
+      "step": 10394
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1308200000000001e-05,
+      "loss": 1.4751,
+      "step": 10395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1307931428571428e-05,
+      "loss": 1.5905,
+      "step": 10396
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1307662857142858e-05,
+      "loss": 1.6241,
+      "step": 10397
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1307394285714287e-05,
+      "loss": 0.939,
+      "step": 10398
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1307125714285714e-05,
+      "loss": 1.0707,
+      "step": 10399
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1306857142857144e-05,
+      "loss": 1.4102,
+      "step": 10400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1306588571428572e-05,
+      "loss": 0.9511,
+      "step": 10401
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.130632e-05,
+      "loss": 1.282,
+      "step": 10402
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1306051428571428e-05,
+      "loss": 1.4367,
+      "step": 10403
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1305782857142858e-05,
+      "loss": 1.4047,
+      "step": 10404
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1305514285714286e-05,
+      "loss": 1.3404,
+      "step": 10405
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1305245714285715e-05,
+      "loss": 0.7816,
+      "step": 10406
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1304977142857143e-05,
+      "loss": 1.4023,
+      "step": 10407
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1304708571428573e-05,
+      "loss": 1.2371,
+      "step": 10408
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1304439999999999e-05,
+      "loss": 1.326,
+      "step": 10409
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1304171428571429e-05,
+      "loss": 1.2012,
+      "step": 10410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1303902857142859e-05,
+      "loss": 1.2728,
+      "step": 10411
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1303634285714285e-05,
+      "loss": 1.2676,
+      "step": 10412
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1303365714285715e-05,
+      "loss": 1.8547,
+      "step": 10413
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1303097142857143e-05,
+      "loss": 1.3359,
+      "step": 10414
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1302828571428572e-05,
+      "loss": 1.3126,
+      "step": 10415
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.130256e-05,
+      "loss": 1.4569,
+      "step": 10416
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.130229142857143e-05,
+      "loss": 1.6968,
+      "step": 10417
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1302022857142858e-05,
+      "loss": 1.0601,
+      "step": 10418
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1301754285714286e-05,
+      "loss": 1.091,
+      "step": 10419
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1301485714285714e-05,
+      "loss": 1.0454,
+      "step": 10420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1301217142857144e-05,
+      "loss": 1.054,
+      "step": 10421
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.130094857142857e-05,
+      "loss": 0.892,
+      "step": 10422
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.130068e-05,
+      "loss": 1.0081,
+      "step": 10423
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1300411428571429e-05,
+      "loss": 0.749,
+      "step": 10424
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1300142857142857e-05,
+      "loss": 1.677,
+      "step": 10425
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1299874285714287e-05,
+      "loss": 1.259,
+      "step": 10426
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1299605714285715e-05,
+      "loss": 1.449,
+      "step": 10427
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1299337142857145e-05,
+      "loss": 0.9949,
+      "step": 10428
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1299068571428571e-05,
+      "loss": 1.5165,
+      "step": 10429
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1298800000000001e-05,
+      "loss": 1.1546,
+      "step": 10430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.129853142857143e-05,
+      "loss": 1.3645,
+      "step": 10431
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1298262857142857e-05,
+      "loss": 1.1243,
+      "step": 10432
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1297994285714286e-05,
+      "loss": 1.031,
+      "step": 10433
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1297725714285715e-05,
+      "loss": 0.9352,
+      "step": 10434
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1297457142857142e-05,
+      "loss": 1.1747,
+      "step": 10435
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1297188571428572e-05,
+      "loss": 1.5997,
+      "step": 10436
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.129692e-05,
+      "loss": 0.7401,
+      "step": 10437
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1296651428571428e-05,
+      "loss": 1.2999,
+      "step": 10438
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1296382857142858e-05,
+      "loss": 1.2508,
+      "step": 10439
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1296114285714286e-05,
+      "loss": 1.0567,
+      "step": 10440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1295845714285716e-05,
+      "loss": 1.5298,
+      "step": 10441
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1295577142857142e-05,
+      "loss": 1.4236,
+      "step": 10442
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1295308571428572e-05,
+      "loss": 1.4642,
+      "step": 10443
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.129504e-05,
+      "loss": 1.5473,
+      "step": 10444
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1294771428571429e-05,
+      "loss": 1.2313,
+      "step": 10445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1294502857142857e-05,
+      "loss": 1.8289,
+      "step": 10446
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1294234285714287e-05,
+      "loss": 1.1205,
+      "step": 10447
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1293965714285713e-05,
+      "loss": 1.6581,
+      "step": 10448
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1293697142857143e-05,
+      "loss": 1.3414,
+      "step": 10449
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1293428571428571e-05,
+      "loss": 1.1717,
+      "step": 10450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1293160000000001e-05,
+      "loss": 0.9222,
+      "step": 10451
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.129289142857143e-05,
+      "loss": 1.2209,
+      "step": 10452
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1292622857142858e-05,
+      "loss": 0.9339,
+      "step": 10453
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1292354285714287e-05,
+      "loss": 1.2169,
+      "step": 10454
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1292085714285714e-05,
+      "loss": 1.4411,
+      "step": 10455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1291817142857144e-05,
+      "loss": 1.3083,
+      "step": 10456
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1291548571428572e-05,
+      "loss": 1.1056,
+      "step": 10457
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.129128e-05,
+      "loss": 0.7345,
+      "step": 10458
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1291011428571428e-05,
+      "loss": 1.4304,
+      "step": 10459
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1290742857142858e-05,
+      "loss": 0.9539,
+      "step": 10460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1290474285714285e-05,
+      "loss": 1.1119,
+      "step": 10461
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1290205714285714e-05,
+      "loss": 0.8806,
+      "step": 10462
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1289937142857143e-05,
+      "loss": 1.0163,
+      "step": 10463
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1289668571428573e-05,
+      "loss": 1.0808,
+      "step": 10464
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.12894e-05,
+      "loss": 1.3451,
+      "step": 10465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1289131428571429e-05,
+      "loss": 1.5551,
+      "step": 10466
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1288862857142859e-05,
+      "loss": 1.1889,
+      "step": 10467
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1288594285714285e-05,
+      "loss": 0.9092,
+      "step": 10468
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1288325714285715e-05,
+      "loss": 1.2049,
+      "step": 10469
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1288057142857143e-05,
+      "loss": 1.6751,
+      "step": 10470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1287788571428571e-05,
+      "loss": 1.2298,
+      "step": 10471
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.128752e-05,
+      "loss": 1.1556,
+      "step": 10472
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.128725142857143e-05,
+      "loss": 0.9541,
+      "step": 10473
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1286982857142858e-05,
+      "loss": 0.981,
+      "step": 10474
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1286714285714286e-05,
+      "loss": 1.4848,
+      "step": 10475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1286445714285714e-05,
+      "loss": 1.4792,
+      "step": 10476
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1286177142857144e-05,
+      "loss": 1.4779,
+      "step": 10477
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1285908571428572e-05,
+      "loss": 1.1131,
+      "step": 10478
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.128564e-05,
+      "loss": 1.0698,
+      "step": 10479
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.128537142857143e-05,
+      "loss": 1.0635,
+      "step": 10480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1285102857142857e-05,
+      "loss": 1.3877,
+      "step": 10481
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1284834285714286e-05,
+      "loss": 1.0162,
+      "step": 10482
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1284565714285715e-05,
+      "loss": 1.0366,
+      "step": 10483
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1284297142857143e-05,
+      "loss": 0.8717,
+      "step": 10484
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1284028571428571e-05,
+      "loss": 1.2726,
+      "step": 10485
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1283760000000001e-05,
+      "loss": 0.8591,
+      "step": 10486
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1283491428571429e-05,
+      "loss": 1.1885,
+      "step": 10487
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1283222857142857e-05,
+      "loss": 1.196,
+      "step": 10488
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1282954285714285e-05,
+      "loss": 1.399,
+      "step": 10489
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1282685714285715e-05,
+      "loss": 0.8578,
+      "step": 10490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1282417142857143e-05,
+      "loss": 1.378,
+      "step": 10491
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1282148571428572e-05,
+      "loss": 0.7787,
+      "step": 10492
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1281880000000001e-05,
+      "loss": 0.9165,
+      "step": 10493
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1281611428571428e-05,
+      "loss": 1.1066,
+      "step": 10494
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1281342857142858e-05,
+      "loss": 1.0763,
+      "step": 10495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1281074285714286e-05,
+      "loss": 1.5204,
+      "step": 10496
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1280805714285716e-05,
+      "loss": 0.9972,
+      "step": 10497
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1280537142857142e-05,
+      "loss": 1.1758,
+      "step": 10498
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1280268571428572e-05,
+      "loss": 1.2729,
+      "step": 10499
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.128e-05,
+      "loss": 1.366,
+      "step": 10500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1279731428571429e-05,
+      "loss": 1.5757,
+      "step": 10501
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1279462857142857e-05,
+      "loss": 1.0518,
+      "step": 10502
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1279194285714287e-05,
+      "loss": 1.2413,
+      "step": 10503
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1278925714285715e-05,
+      "loss": 1.142,
+      "step": 10504
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1278657142857143e-05,
+      "loss": 1.2727,
+      "step": 10505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1278388571428573e-05,
+      "loss": 1.3146,
+      "step": 10506
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.127812e-05,
+      "loss": 1.34,
+      "step": 10507
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.127785142857143e-05,
+      "loss": 1.1841,
+      "step": 10508
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1277582857142857e-05,
+      "loss": 1.1501,
+      "step": 10509
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1277314285714287e-05,
+      "loss": 1.1275,
+      "step": 10510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1277045714285714e-05,
+      "loss": 1.0702,
+      "step": 10511
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1276777142857144e-05,
+      "loss": 1.0454,
+      "step": 10512
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1276508571428572e-05,
+      "loss": 1.0921,
+      "step": 10513
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.127624e-05,
+      "loss": 1.2313,
+      "step": 10514
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1275971428571428e-05,
+      "loss": 1.5493,
+      "step": 10515
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1275702857142858e-05,
+      "loss": 1.5341,
+      "step": 10516
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1275434285714286e-05,
+      "loss": 1.1608,
+      "step": 10517
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1275165714285714e-05,
+      "loss": 1.5805,
+      "step": 10518
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1274897142857144e-05,
+      "loss": 1.4529,
+      "step": 10519
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1274628571428572e-05,
+      "loss": 1.0474,
+      "step": 10520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.127436e-05,
+      "loss": 1.4149,
+      "step": 10521
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1274091428571429e-05,
+      "loss": 1.3923,
+      "step": 10522
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1273822857142859e-05,
+      "loss": 1.1387,
+      "step": 10523
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1273554285714285e-05,
+      "loss": 1.3318,
+      "step": 10524
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1273285714285715e-05,
+      "loss": 1.2056,
+      "step": 10525
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1273017142857143e-05,
+      "loss": 1.4285,
+      "step": 10526
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1272748571428571e-05,
+      "loss": 1.4234,
+      "step": 10527
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.127248e-05,
+      "loss": 1.6582,
+      "step": 10528
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.127221142857143e-05,
+      "loss": 1.3481,
+      "step": 10529
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1271942857142858e-05,
+      "loss": 1.011,
+      "step": 10530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1271674285714286e-05,
+      "loss": 1.3917,
+      "step": 10531
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1271405714285716e-05,
+      "loss": 1.3492,
+      "step": 10532
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1271137142857144e-05,
+      "loss": 1.4277,
+      "step": 10533
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1270868571428572e-05,
+      "loss": 1.5182,
+      "step": 10534
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.12706e-05,
+      "loss": 1.1616,
+      "step": 10535
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.127033142857143e-05,
+      "loss": 1.1741,
+      "step": 10536
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1270062857142856e-05,
+      "loss": 1.1482,
+      "step": 10537
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1269794285714286e-05,
+      "loss": 1.312,
+      "step": 10538
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1269525714285715e-05,
+      "loss": 1.3042,
+      "step": 10539
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1269257142857143e-05,
+      "loss": 1.3005,
+      "step": 10540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1268988571428571e-05,
+      "loss": 1.1712,
+      "step": 10541
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.126872e-05,
+      "loss": 1.1938,
+      "step": 10542
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1268451428571429e-05,
+      "loss": 1.4758,
+      "step": 10543
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1268182857142857e-05,
+      "loss": 1.2482,
+      "step": 10544
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1267914285714287e-05,
+      "loss": 1.5771,
+      "step": 10545
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1267645714285715e-05,
+      "loss": 1.4245,
+      "step": 10546
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1267377142857143e-05,
+      "loss": 1.2889,
+      "step": 10547
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1267108571428572e-05,
+      "loss": 1.3442,
+      "step": 10548
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1266840000000001e-05,
+      "loss": 0.9079,
+      "step": 10549
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1266571428571428e-05,
+      "loss": 1.3243,
+      "step": 10550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1266302857142858e-05,
+      "loss": 1.3357,
+      "step": 10551
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1266034285714286e-05,
+      "loss": 1.1547,
+      "step": 10552
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1265765714285714e-05,
+      "loss": 1.4876,
+      "step": 10553
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1265497142857142e-05,
+      "loss": 1.4142,
+      "step": 10554
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1265228571428572e-05,
+      "loss": 1.2709,
+      "step": 10555
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.126496e-05,
+      "loss": 1.1029,
+      "step": 10556
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1264691428571428e-05,
+      "loss": 1.2482,
+      "step": 10557
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1264422857142858e-05,
+      "loss": 1.3457,
+      "step": 10558
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1264154285714287e-05,
+      "loss": 1.0763,
+      "step": 10559
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1263885714285715e-05,
+      "loss": 1.52,
+      "step": 10560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1263617142857143e-05,
+      "loss": 1.4795,
+      "step": 10561
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1263348571428573e-05,
+      "loss": 1.4381,
+      "step": 10562
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.126308e-05,
+      "loss": 1.1884,
+      "step": 10563
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1262811428571429e-05,
+      "loss": 1.1048,
+      "step": 10564
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1262542857142857e-05,
+      "loss": 1.7287,
+      "step": 10565
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1262274285714285e-05,
+      "loss": 1.0827,
+      "step": 10566
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1262005714285714e-05,
+      "loss": 1.0521,
+      "step": 10567
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1261737142857144e-05,
+      "loss": 1.0265,
+      "step": 10568
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1261468571428572e-05,
+      "loss": 1.4224,
+      "step": 10569
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.12612e-05,
+      "loss": 1.669,
+      "step": 10570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.126093142857143e-05,
+      "loss": 1.3906,
+      "step": 10571
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1260662857142858e-05,
+      "loss": 1.3737,
+      "step": 10572
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1260394285714286e-05,
+      "loss": 1.1672,
+      "step": 10573
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1260125714285714e-05,
+      "loss": 1.2526,
+      "step": 10574
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1259857142857144e-05,
+      "loss": 1.3086,
+      "step": 10575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.125958857142857e-05,
+      "loss": 1.5542,
+      "step": 10576
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.125932e-05,
+      "loss": 1.4996,
+      "step": 10577
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1259051428571429e-05,
+      "loss": 1.5655,
+      "step": 10578
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1258782857142859e-05,
+      "loss": 1.379,
+      "step": 10579
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1258514285714285e-05,
+      "loss": 1.037,
+      "step": 10580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1258245714285715e-05,
+      "loss": 1.3053,
+      "step": 10581
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1257977142857143e-05,
+      "loss": 1.3693,
+      "step": 10582
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1257708571428571e-05,
+      "loss": 1.2203,
+      "step": 10583
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1257440000000001e-05,
+      "loss": 0.9391,
+      "step": 10584
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.125717142857143e-05,
+      "loss": 1.6458,
+      "step": 10585
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1256902857142857e-05,
+      "loss": 1.5399,
+      "step": 10586
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1256634285714286e-05,
+      "loss": 0.8904,
+      "step": 10587
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1256365714285715e-05,
+      "loss": 0.9766,
+      "step": 10588
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1256097142857142e-05,
+      "loss": 1.2264,
+      "step": 10589
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1255828571428572e-05,
+      "loss": 1.3404,
+      "step": 10590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.125556e-05,
+      "loss": 1.0856,
+      "step": 10591
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.125529142857143e-05,
+      "loss": 1.1579,
+      "step": 10592
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1255022857142856e-05,
+      "loss": 0.8727,
+      "step": 10593
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1254754285714286e-05,
+      "loss": 1.3557,
+      "step": 10594
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1254485714285714e-05,
+      "loss": 1.8197,
+      "step": 10595
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1254217142857143e-05,
+      "loss": 1.1686,
+      "step": 10596
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1253948571428572e-05,
+      "loss": 1.4358,
+      "step": 10597
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.125368e-05,
+      "loss": 1.5924,
+      "step": 10598
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1253411428571429e-05,
+      "loss": 1.1782,
+      "step": 10599
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1253142857142857e-05,
+      "loss": 1.3448,
+      "step": 10600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1252874285714287e-05,
+      "loss": 0.7687,
+      "step": 10601
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1252605714285715e-05,
+      "loss": 1.2653,
+      "step": 10602
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1252337142857143e-05,
+      "loss": 1.2986,
+      "step": 10603
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1252068571428571e-05,
+      "loss": 1.4762,
+      "step": 10604
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1251800000000001e-05,
+      "loss": 1.0791,
+      "step": 10605
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1251531428571428e-05,
+      "loss": 1.2538,
+      "step": 10606
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1251262857142858e-05,
+      "loss": 1.6076,
+      "step": 10607
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1250994285714286e-05,
+      "loss": 1.2162,
+      "step": 10608
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1250725714285714e-05,
+      "loss": 1.1104,
+      "step": 10609
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1250457142857144e-05,
+      "loss": 1.5417,
+      "step": 10610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1250188571428572e-05,
+      "loss": 1.1969,
+      "step": 10611
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124992e-05,
+      "loss": 1.5424,
+      "step": 10612
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1249651428571428e-05,
+      "loss": 1.354,
+      "step": 10613
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1249382857142858e-05,
+      "loss": 1.4303,
+      "step": 10614
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1249114285714286e-05,
+      "loss": 1.4727,
+      "step": 10615
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1248845714285715e-05,
+      "loss": 1.282,
+      "step": 10616
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1248577142857143e-05,
+      "loss": 1.4935,
+      "step": 10617
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1248308571428573e-05,
+      "loss": 1.114,
+      "step": 10618
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1248039999999999e-05,
+      "loss": 1.3027,
+      "step": 10619
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1247771428571429e-05,
+      "loss": 1.0121,
+      "step": 10620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1247502857142857e-05,
+      "loss": 1.391,
+      "step": 10621
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1247234285714285e-05,
+      "loss": 1.0754,
+      "step": 10622
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1246965714285715e-05,
+      "loss": 1.2844,
+      "step": 10623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1246697142857143e-05,
+      "loss": 0.969,
+      "step": 10624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1246428571428573e-05,
+      "loss": 1.3759,
+      "step": 10625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124616e-05,
+      "loss": 1.2595,
+      "step": 10626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124589142857143e-05,
+      "loss": 1.1607,
+      "step": 10627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1245622857142858e-05,
+      "loss": 1.6098,
+      "step": 10628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1245354285714286e-05,
+      "loss": 1.4232,
+      "step": 10629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1245085714285714e-05,
+      "loss": 1.2185,
+      "step": 10630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1244817142857144e-05,
+      "loss": 1.1829,
+      "step": 10631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124454857142857e-05,
+      "loss": 1.1006,
+      "step": 10632
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124428e-05,
+      "loss": 1.0166,
+      "step": 10633
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1244011428571429e-05,
+      "loss": 1.2638,
+      "step": 10634
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1243742857142857e-05,
+      "loss": 1.2759,
+      "step": 10635
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1243474285714287e-05,
+      "loss": 1.1282,
+      "step": 10636
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1243205714285715e-05,
+      "loss": 1.363,
+      "step": 10637
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1242937142857145e-05,
+      "loss": 1.1216,
+      "step": 10638
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1242668571428571e-05,
+      "loss": 1.3539,
+      "step": 10639
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1242400000000001e-05,
+      "loss": 0.9453,
+      "step": 10640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124213142857143e-05,
+      "loss": 1.3252,
+      "step": 10641
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1241862857142857e-05,
+      "loss": 1.2822,
+      "step": 10642
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1241594285714286e-05,
+      "loss": 1.2434,
+      "step": 10643
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1241325714285715e-05,
+      "loss": 0.9938,
+      "step": 10644
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1241057142857142e-05,
+      "loss": 0.8393,
+      "step": 10645
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1240788571428572e-05,
+      "loss": 1.1515,
+      "step": 10646
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124052e-05,
+      "loss": 1.3173,
+      "step": 10647
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124025142857143e-05,
+      "loss": 1.2015,
+      "step": 10648
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1239982857142858e-05,
+      "loss": 1.444,
+      "step": 10649
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1239714285714286e-05,
+      "loss": 1.1418,
+      "step": 10650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1239445714285716e-05,
+      "loss": 1.3238,
+      "step": 10651
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1239177142857142e-05,
+      "loss": 1.397,
+      "step": 10652
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1238908571428572e-05,
+      "loss": 1.4398,
+      "step": 10653
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.123864e-05,
+      "loss": 1.2517,
+      "step": 10654
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1238371428571429e-05,
+      "loss": 0.827,
+      "step": 10655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1238102857142857e-05,
+      "loss": 1.8905,
+      "step": 10656
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1237834285714287e-05,
+      "loss": 0.8907,
+      "step": 10657
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1237565714285713e-05,
+      "loss": 1.6249,
+      "step": 10658
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1237297142857143e-05,
+      "loss": 1.4509,
+      "step": 10659
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1237028571428571e-05,
+      "loss": 1.4309,
+      "step": 10660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1236760000000001e-05,
+      "loss": 1.0983,
+      "step": 10661
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.123649142857143e-05,
+      "loss": 1.6042,
+      "step": 10662
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1236222857142858e-05,
+      "loss": 0.9362,
+      "step": 10663
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1235954285714287e-05,
+      "loss": 1.5149,
+      "step": 10664
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1235685714285714e-05,
+      "loss": 1.1624,
+      "step": 10665
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1235417142857144e-05,
+      "loss": 1.0001,
+      "step": 10666
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1235148571428572e-05,
+      "loss": 1.0424,
+      "step": 10667
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.123488e-05,
+      "loss": 1.2227,
+      "step": 10668
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1234611428571428e-05,
+      "loss": 1.2972,
+      "step": 10669
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1234342857142858e-05,
+      "loss": 1.0375,
+      "step": 10670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1234074285714286e-05,
+      "loss": 1.1559,
+      "step": 10671
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1233805714285714e-05,
+      "loss": 1.2442,
+      "step": 10672
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1233537142857143e-05,
+      "loss": 1.1002,
+      "step": 10673
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1233268571428573e-05,
+      "loss": 1.2569,
+      "step": 10674
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1233e-05,
+      "loss": 1.3015,
+      "step": 10675
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1232731428571429e-05,
+      "loss": 1.2506,
+      "step": 10676
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1232462857142859e-05,
+      "loss": 1.2432,
+      "step": 10677
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1232194285714285e-05,
+      "loss": 1.5123,
+      "step": 10678
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1231925714285715e-05,
+      "loss": 1.3884,
+      "step": 10679
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1231657142857143e-05,
+      "loss": 1.0836,
+      "step": 10680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1231388571428571e-05,
+      "loss": 1.4227,
+      "step": 10681
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.123112e-05,
+      "loss": 1.3673,
+      "step": 10682
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.123085142857143e-05,
+      "loss": 1.203,
+      "step": 10683
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1230582857142858e-05,
+      "loss": 1.5028,
+      "step": 10684
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1230314285714286e-05,
+      "loss": 1.1771,
+      "step": 10685
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1230045714285714e-05,
+      "loss": 1.2253,
+      "step": 10686
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1229777142857144e-05,
+      "loss": 1.1416,
+      "step": 10687
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1229508571428572e-05,
+      "loss": 1.2342,
+      "step": 10688
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.122924e-05,
+      "loss": 1.3142,
+      "step": 10689
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.122897142857143e-05,
+      "loss": 1.1389,
+      "step": 10690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1228702857142857e-05,
+      "loss": 1.2397,
+      "step": 10691
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1228434285714286e-05,
+      "loss": 1.4368,
+      "step": 10692
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1228165714285715e-05,
+      "loss": 1.1786,
+      "step": 10693
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1227897142857145e-05,
+      "loss": 1.4245,
+      "step": 10694
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1227628571428571e-05,
+      "loss": 1.0467,
+      "step": 10695
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1227360000000001e-05,
+      "loss": 0.8273,
+      "step": 10696
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1227091428571429e-05,
+      "loss": 1.6397,
+      "step": 10697
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1226822857142857e-05,
+      "loss": 1.0699,
+      "step": 10698
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1226554285714285e-05,
+      "loss": 1.2949,
+      "step": 10699
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1226285714285715e-05,
+      "loss": 1.6671,
+      "step": 10700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1226017142857143e-05,
+      "loss": 1.2122,
+      "step": 10701
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1225748571428572e-05,
+      "loss": 1.2602,
+      "step": 10702
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1225480000000001e-05,
+      "loss": 1.4489,
+      "step": 10703
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1225211428571428e-05,
+      "loss": 1.3868,
+      "step": 10704
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1224942857142858e-05,
+      "loss": 1.5019,
+      "step": 10705
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1224674285714286e-05,
+      "loss": 1.5796,
+      "step": 10706
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1224405714285716e-05,
+      "loss": 1.4023,
+      "step": 10707
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1224137142857142e-05,
+      "loss": 1.1832,
+      "step": 10708
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1223868571428572e-05,
+      "loss": 1.2001,
+      "step": 10709
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.12236e-05,
+      "loss": 1.4106,
+      "step": 10710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1223331428571429e-05,
+      "loss": 1.3641,
+      "step": 10711
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1223062857142857e-05,
+      "loss": 1.0974,
+      "step": 10712
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1222794285714287e-05,
+      "loss": 1.4253,
+      "step": 10713
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1222525714285715e-05,
+      "loss": 1.246,
+      "step": 10714
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1222257142857143e-05,
+      "loss": 1.4365,
+      "step": 10715
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1221988571428573e-05,
+      "loss": 1.3624,
+      "step": 10716
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.122172e-05,
+      "loss": 1.1491,
+      "step": 10717
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.122145142857143e-05,
+      "loss": 1.2657,
+      "step": 10718
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1221182857142857e-05,
+      "loss": 0.6595,
+      "step": 10719
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1220914285714287e-05,
+      "loss": 1.4621,
+      "step": 10720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1220645714285714e-05,
+      "loss": 0.973,
+      "step": 10721
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1220377142857144e-05,
+      "loss": 0.9967,
+      "step": 10722
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1220108571428572e-05,
+      "loss": 1.1384,
+      "step": 10723
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.121984e-05,
+      "loss": 1.1554,
+      "step": 10724
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1219571428571428e-05,
+      "loss": 1.5548,
+      "step": 10725
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1219302857142858e-05,
+      "loss": 0.8353,
+      "step": 10726
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1219034285714286e-05,
+      "loss": 1.1121,
+      "step": 10727
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1218765714285714e-05,
+      "loss": 1.2908,
+      "step": 10728
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1218497142857144e-05,
+      "loss": 0.8169,
+      "step": 10729
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1218228571428572e-05,
+      "loss": 1.09,
+      "step": 10730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.121796e-05,
+      "loss": 0.9076,
+      "step": 10731
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1217691428571429e-05,
+      "loss": 1.0465,
+      "step": 10732
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1217422857142859e-05,
+      "loss": 1.3274,
+      "step": 10733
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1217154285714285e-05,
+      "loss": 1.2122,
+      "step": 10734
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1216885714285715e-05,
+      "loss": 0.9148,
+      "step": 10735
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1216617142857143e-05,
+      "loss": 0.8617,
+      "step": 10736
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1216348571428571e-05,
+      "loss": 1.3274,
+      "step": 10737
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.121608e-05,
+      "loss": 1.1909,
+      "step": 10738
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.121581142857143e-05,
+      "loss": 1.1859,
+      "step": 10739
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1215542857142858e-05,
+      "loss": 1.7218,
+      "step": 10740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1215274285714286e-05,
+      "loss": 1.1805,
+      "step": 10741
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1215005714285716e-05,
+      "loss": 1.0633,
+      "step": 10742
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1214737142857144e-05,
+      "loss": 1.02,
+      "step": 10743
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1214468571428572e-05,
+      "loss": 0.9607,
+      "step": 10744
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.12142e-05,
+      "loss": 1.06,
+      "step": 10745
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.121393142857143e-05,
+      "loss": 1.5124,
+      "step": 10746
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1213662857142856e-05,
+      "loss": 1.3679,
+      "step": 10747
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1213394285714286e-05,
+      "loss": 1.2613,
+      "step": 10748
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1213125714285715e-05,
+      "loss": 1.5024,
+      "step": 10749
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1212857142857143e-05,
+      "loss": 1.4354,
+      "step": 10750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1212588571428571e-05,
+      "loss": 1.2057,
+      "step": 10751
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.121232e-05,
+      "loss": 1.4491,
+      "step": 10752
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1212051428571429e-05,
+      "loss": 1.2188,
+      "step": 10753
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1211782857142857e-05,
+      "loss": 0.9408,
+      "step": 10754
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1211514285714287e-05,
+      "loss": 1.0902,
+      "step": 10755
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1211245714285715e-05,
+      "loss": 1.3363,
+      "step": 10756
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1210977142857143e-05,
+      "loss": 1.4674,
+      "step": 10757
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1210708571428572e-05,
+      "loss": 1.0135,
+      "step": 10758
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1210440000000001e-05,
+      "loss": 1.2532,
+      "step": 10759
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1210171428571428e-05,
+      "loss": 1.0582,
+      "step": 10760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1209902857142858e-05,
+      "loss": 1.414,
+      "step": 10761
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1209634285714286e-05,
+      "loss": 1.0855,
+      "step": 10762
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1209365714285714e-05,
+      "loss": 1.173,
+      "step": 10763
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1209097142857142e-05,
+      "loss": 1.0096,
+      "step": 10764
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1208828571428572e-05,
+      "loss": 1.2515,
+      "step": 10765
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.120856e-05,
+      "loss": 1.0831,
+      "step": 10766
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1208291428571428e-05,
+      "loss": 1.4116,
+      "step": 10767
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1208022857142858e-05,
+      "loss": 0.9846,
+      "step": 10768
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1207754285714287e-05,
+      "loss": 1.3205,
+      "step": 10769
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1207485714285715e-05,
+      "loss": 0.8004,
+      "step": 10770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1207217142857143e-05,
+      "loss": 1.3705,
+      "step": 10771
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1206948571428573e-05,
+      "loss": 0.9048,
+      "step": 10772
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.120668e-05,
+      "loss": 0.9594,
+      "step": 10773
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1206411428571429e-05,
+      "loss": 1.1997,
+      "step": 10774
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1206142857142857e-05,
+      "loss": 1.1024,
+      "step": 10775
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1205874285714287e-05,
+      "loss": 1.4115,
+      "step": 10776
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1205605714285714e-05,
+      "loss": 1.4687,
+      "step": 10777
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1205337142857144e-05,
+      "loss": 1.3661,
+      "step": 10778
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1205068571428572e-05,
+      "loss": 1.2043,
+      "step": 10779
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.12048e-05,
+      "loss": 1.4103,
+      "step": 10780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.120453142857143e-05,
+      "loss": 1.1294,
+      "step": 10781
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1204262857142858e-05,
+      "loss": 1.7555,
+      "step": 10782
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1203994285714286e-05,
+      "loss": 1.1407,
+      "step": 10783
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1203725714285714e-05,
+      "loss": 1.3356,
+      "step": 10784
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1203457142857144e-05,
+      "loss": 1.3043,
+      "step": 10785
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.120318857142857e-05,
+      "loss": 1.3592,
+      "step": 10786
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.120292e-05,
+      "loss": 1.5137,
+      "step": 10787
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1202651428571429e-05,
+      "loss": 1.1023,
+      "step": 10788
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1202382857142859e-05,
+      "loss": 1.5447,
+      "step": 10789
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1202114285714285e-05,
+      "loss": 1.3781,
+      "step": 10790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1201845714285715e-05,
+      "loss": 1.2051,
+      "step": 10791
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1201577142857143e-05,
+      "loss": 1.1301,
+      "step": 10792
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1201308571428571e-05,
+      "loss": 0.996,
+      "step": 10793
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1201040000000001e-05,
+      "loss": 1.4269,
+      "step": 10794
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.120077142857143e-05,
+      "loss": 1.1393,
+      "step": 10795
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1200502857142857e-05,
+      "loss": 1.1904,
+      "step": 10796
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1200234285714286e-05,
+      "loss": 1.5044,
+      "step": 10797
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1199965714285715e-05,
+      "loss": 1.5281,
+      "step": 10798
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1199697142857144e-05,
+      "loss": 1.1685,
+      "step": 10799
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1199428571428572e-05,
+      "loss": 1.6975,
+      "step": 10800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.119916e-05,
+      "loss": 1.1085,
+      "step": 10801
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.119889142857143e-05,
+      "loss": 1.3022,
+      "step": 10802
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1198622857142856e-05,
+      "loss": 1.1029,
+      "step": 10803
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1198354285714286e-05,
+      "loss": 1.2671,
+      "step": 10804
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1198085714285714e-05,
+      "loss": 1.0862,
+      "step": 10805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1197817142857143e-05,
+      "loss": 1.4807,
+      "step": 10806
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1197548571428572e-05,
+      "loss": 0.8124,
+      "step": 10807
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.119728e-05,
+      "loss": 1.3059,
+      "step": 10808
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1197011428571429e-05,
+      "loss": 1.1393,
+      "step": 10809
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1196742857142857e-05,
+      "loss": 1.3632,
+      "step": 10810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1196474285714287e-05,
+      "loss": 1.0756,
+      "step": 10811
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1196205714285715e-05,
+      "loss": 1.5874,
+      "step": 10812
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1195937142857143e-05,
+      "loss": 0.9401,
+      "step": 10813
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1195668571428571e-05,
+      "loss": 1.6417,
+      "step": 10814
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1195400000000001e-05,
+      "loss": 1.3169,
+      "step": 10815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1195131428571428e-05,
+      "loss": 1.4502,
+      "step": 10816
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1194862857142858e-05,
+      "loss": 1.4774,
+      "step": 10817
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1194594285714286e-05,
+      "loss": 1.0494,
+      "step": 10818
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1194325714285714e-05,
+      "loss": 1.3617,
+      "step": 10819
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1194057142857144e-05,
+      "loss": 1.0477,
+      "step": 10820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1193788571428572e-05,
+      "loss": 1.4199,
+      "step": 10821
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1193520000000002e-05,
+      "loss": 1.3243,
+      "step": 10822
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1193251428571428e-05,
+      "loss": 1.3067,
+      "step": 10823
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1192982857142858e-05,
+      "loss": 1.1583,
+      "step": 10824
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1192714285714286e-05,
+      "loss": 1.4003,
+      "step": 10825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1192445714285715e-05,
+      "loss": 0.9687,
+      "step": 10826
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1192177142857143e-05,
+      "loss": 1.1739,
+      "step": 10827
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1191908571428573e-05,
+      "loss": 1.4179,
+      "step": 10828
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1191639999999999e-05,
+      "loss": 1.377,
+      "step": 10829
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1191371428571429e-05,
+      "loss": 1.2465,
+      "step": 10830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1191102857142857e-05,
+      "loss": 1.471,
+      "step": 10831
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1190834285714285e-05,
+      "loss": 1.1838,
+      "step": 10832
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1190565714285715e-05,
+      "loss": 1.0503,
+      "step": 10833
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1190297142857143e-05,
+      "loss": 1.2955,
+      "step": 10834
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1190028571428573e-05,
+      "loss": 1.3834,
+      "step": 10835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118976e-05,
+      "loss": 1.3166,
+      "step": 10836
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118949142857143e-05,
+      "loss": 0.889,
+      "step": 10837
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1189222857142858e-05,
+      "loss": 1.4016,
+      "step": 10838
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1188954285714286e-05,
+      "loss": 1.3775,
+      "step": 10839
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1188685714285714e-05,
+      "loss": 1.0665,
+      "step": 10840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1188417142857144e-05,
+      "loss": 1.3386,
+      "step": 10841
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118814857142857e-05,
+      "loss": 1.4311,
+      "step": 10842
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118788e-05,
+      "loss": 1.8103,
+      "step": 10843
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1187611428571429e-05,
+      "loss": 1.486,
+      "step": 10844
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1187342857142858e-05,
+      "loss": 1.1136,
+      "step": 10845
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1187074285714287e-05,
+      "loss": 1.0316,
+      "step": 10846
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1186805714285715e-05,
+      "loss": 1.1367,
+      "step": 10847
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1186537142857145e-05,
+      "loss": 1.1377,
+      "step": 10848
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1186268571428571e-05,
+      "loss": 1.2492,
+      "step": 10849
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1186000000000001e-05,
+      "loss": 1.4028,
+      "step": 10850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118573142857143e-05,
+      "loss": 0.9755,
+      "step": 10851
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1185462857142857e-05,
+      "loss": 1.3739,
+      "step": 10852
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1185194285714286e-05,
+      "loss": 1.2171,
+      "step": 10853
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1184925714285715e-05,
+      "loss": 1.0897,
+      "step": 10854
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1184657142857142e-05,
+      "loss": 1.7265,
+      "step": 10855
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1184388571428572e-05,
+      "loss": 1.4015,
+      "step": 10856
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118412e-05,
+      "loss": 1.6722,
+      "step": 10857
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118385142857143e-05,
+      "loss": 1.4415,
+      "step": 10858
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1183582857142858e-05,
+      "loss": 1.3342,
+      "step": 10859
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1183314285714286e-05,
+      "loss": 1.2407,
+      "step": 10860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1183045714285716e-05,
+      "loss": 1.0374,
+      "step": 10861
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1182777142857142e-05,
+      "loss": 1.4435,
+      "step": 10862
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1182508571428572e-05,
+      "loss": 0.7461,
+      "step": 10863
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118224e-05,
+      "loss": 0.7052,
+      "step": 10864
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1181971428571429e-05,
+      "loss": 1.1351,
+      "step": 10865
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1181702857142857e-05,
+      "loss": 1.413,
+      "step": 10866
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1181434285714287e-05,
+      "loss": 1.2409,
+      "step": 10867
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1181165714285715e-05,
+      "loss": 1.1395,
+      "step": 10868
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1180897142857143e-05,
+      "loss": 1.2268,
+      "step": 10869
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1180628571428571e-05,
+      "loss": 1.4991,
+      "step": 10870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1180360000000001e-05,
+      "loss": 1.0726,
+      "step": 10871
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.118009142857143e-05,
+      "loss": 1.2537,
+      "step": 10872
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1179822857142858e-05,
+      "loss": 1.3076,
+      "step": 10873
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1179554285714287e-05,
+      "loss": 0.9956,
+      "step": 10874
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1179285714285714e-05,
+      "loss": 1.1408,
+      "step": 10875
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1179017142857144e-05,
+      "loss": 0.9713,
+      "step": 10876
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1178748571428572e-05,
+      "loss": 1.1804,
+      "step": 10877
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.117848e-05,
+      "loss": 1.16,
+      "step": 10878
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1178211428571428e-05,
+      "loss": 1.2668,
+      "step": 10879
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1177942857142858e-05,
+      "loss": 1.3711,
+      "step": 10880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1177674285714286e-05,
+      "loss": 1.2653,
+      "step": 10881
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1177405714285714e-05,
+      "loss": 1.2382,
+      "step": 10882
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1177137142857143e-05,
+      "loss": 1.3406,
+      "step": 10883
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1176868571428573e-05,
+      "loss": 1.2417,
+      "step": 10884
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.11766e-05,
+      "loss": 1.6681,
+      "step": 10885
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1176331428571429e-05,
+      "loss": 1.5458,
+      "step": 10886
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1176062857142859e-05,
+      "loss": 0.9109,
+      "step": 10887
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1175794285714285e-05,
+      "loss": 1.281,
+      "step": 10888
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1175525714285715e-05,
+      "loss": 1.2051,
+      "step": 10889
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1175257142857143e-05,
+      "loss": 1.1709,
+      "step": 10890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1174988571428571e-05,
+      "loss": 1.0789,
+      "step": 10891
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.117472e-05,
+      "loss": 1.6447,
+      "step": 10892
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.117445142857143e-05,
+      "loss": 1.0187,
+      "step": 10893
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1174182857142858e-05,
+      "loss": 0.9132,
+      "step": 10894
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1173914285714286e-05,
+      "loss": 1.6209,
+      "step": 10895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1173645714285714e-05,
+      "loss": 1.2003,
+      "step": 10896
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1173377142857144e-05,
+      "loss": 1.2161,
+      "step": 10897
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1173108571428572e-05,
+      "loss": 1.2783,
+      "step": 10898
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.117284e-05,
+      "loss": 1.5456,
+      "step": 10899
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.117257142857143e-05,
+      "loss": 1.0052,
+      "step": 10900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1172302857142857e-05,
+      "loss": 1.0086,
+      "step": 10901
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1172034285714286e-05,
+      "loss": 0.8007,
+      "step": 10902
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1171765714285715e-05,
+      "loss": 0.8693,
+      "step": 10903
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1171497142857145e-05,
+      "loss": 1.3148,
+      "step": 10904
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1171228571428571e-05,
+      "loss": 1.4188,
+      "step": 10905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1170960000000001e-05,
+      "loss": 0.8909,
+      "step": 10906
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1170691428571429e-05,
+      "loss": 1.3353,
+      "step": 10907
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1170422857142857e-05,
+      "loss": 1.2932,
+      "step": 10908
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1170154285714285e-05,
+      "loss": 1.1941,
+      "step": 10909
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1169885714285715e-05,
+      "loss": 1.2804,
+      "step": 10910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1169617142857143e-05,
+      "loss": 1.3465,
+      "step": 10911
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1169348571428572e-05,
+      "loss": 1.0902,
+      "step": 10912
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1169080000000001e-05,
+      "loss": 0.9803,
+      "step": 10913
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1168811428571428e-05,
+      "loss": 1.4902,
+      "step": 10914
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1168542857142858e-05,
+      "loss": 1.2692,
+      "step": 10915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1168274285714286e-05,
+      "loss": 1.2486,
+      "step": 10916
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1168005714285716e-05,
+      "loss": 1.2245,
+      "step": 10917
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1167737142857142e-05,
+      "loss": 1.0104,
+      "step": 10918
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1167468571428572e-05,
+      "loss": 1.0503,
+      "step": 10919
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.11672e-05,
+      "loss": 1.4256,
+      "step": 10920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1166931428571429e-05,
+      "loss": 1.2336,
+      "step": 10921
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1166662857142857e-05,
+      "loss": 1.0725,
+      "step": 10922
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1166394285714287e-05,
+      "loss": 1.1549,
+      "step": 10923
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1166125714285715e-05,
+      "loss": 1.3511,
+      "step": 10924
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1165857142857143e-05,
+      "loss": 0.9213,
+      "step": 10925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1165588571428573e-05,
+      "loss": 1.4948,
+      "step": 10926
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1165320000000001e-05,
+      "loss": 0.7954,
+      "step": 10927
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.116505142857143e-05,
+      "loss": 1.526,
+      "step": 10928
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1164782857142857e-05,
+      "loss": 1.3636,
+      "step": 10929
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1164514285714287e-05,
+      "loss": 1.0102,
+      "step": 10930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1164245714285714e-05,
+      "loss": 1.4358,
+      "step": 10931
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1163977142857144e-05,
+      "loss": 1.6341,
+      "step": 10932
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1163708571428572e-05,
+      "loss": 1.0049,
+      "step": 10933
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.116344e-05,
+      "loss": 1.2032,
+      "step": 10934
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1163171428571428e-05,
+      "loss": 1.0722,
+      "step": 10935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1162902857142858e-05,
+      "loss": 1.156,
+      "step": 10936
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1162634285714286e-05,
+      "loss": 1.2257,
+      "step": 10937
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1162365714285714e-05,
+      "loss": 0.8763,
+      "step": 10938
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1162097142857144e-05,
+      "loss": 1.1368,
+      "step": 10939
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1161828571428572e-05,
+      "loss": 1.1903,
+      "step": 10940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.116156e-05,
+      "loss": 1.2458,
+      "step": 10941
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1161291428571429e-05,
+      "loss": 1.1787,
+      "step": 10942
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1161022857142859e-05,
+      "loss": 1.3429,
+      "step": 10943
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1160754285714285e-05,
+      "loss": 1.3932,
+      "step": 10944
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1160485714285715e-05,
+      "loss": 1.4238,
+      "step": 10945
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1160217142857143e-05,
+      "loss": 1.5566,
+      "step": 10946
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1159948571428571e-05,
+      "loss": 1.0838,
+      "step": 10947
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.115968e-05,
+      "loss": 1.4598,
+      "step": 10948
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.115941142857143e-05,
+      "loss": 1.1965,
+      "step": 10949
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1159142857142858e-05,
+      "loss": 1.6367,
+      "step": 10950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1158874285714286e-05,
+      "loss": 0.896,
+      "step": 10951
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1158605714285716e-05,
+      "loss": 0.8503,
+      "step": 10952
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1158337142857144e-05,
+      "loss": 1.5742,
+      "step": 10953
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1158068571428572e-05,
+      "loss": 1.2287,
+      "step": 10954
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.11578e-05,
+      "loss": 1.5825,
+      "step": 10955
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.115753142857143e-05,
+      "loss": 1.1426,
+      "step": 10956
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1157262857142856e-05,
+      "loss": 1.7018,
+      "step": 10957
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1156994285714286e-05,
+      "loss": 1.6448,
+      "step": 10958
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1156725714285715e-05,
+      "loss": 1.1864,
+      "step": 10959
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1156457142857143e-05,
+      "loss": 1.2782,
+      "step": 10960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1156188571428571e-05,
+      "loss": 0.9509,
+      "step": 10961
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.115592e-05,
+      "loss": 1.4977,
+      "step": 10962
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1155651428571429e-05,
+      "loss": 1.158,
+      "step": 10963
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1155382857142857e-05,
+      "loss": 1.2603,
+      "step": 10964
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1155114285714287e-05,
+      "loss": 1.358,
+      "step": 10965
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1154845714285715e-05,
+      "loss": 0.8294,
+      "step": 10966
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1154577142857143e-05,
+      "loss": 1.0886,
+      "step": 10967
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1154308571428572e-05,
+      "loss": 1.116,
+      "step": 10968
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1154040000000001e-05,
+      "loss": 1.6204,
+      "step": 10969
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1153771428571428e-05,
+      "loss": 1.4561,
+      "step": 10970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1153502857142858e-05,
+      "loss": 0.9873,
+      "step": 10971
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1153234285714286e-05,
+      "loss": 1.1036,
+      "step": 10972
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1152965714285716e-05,
+      "loss": 1.0777,
+      "step": 10973
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1152697142857142e-05,
+      "loss": 1.1317,
+      "step": 10974
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1152428571428572e-05,
+      "loss": 1.5388,
+      "step": 10975
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.115216e-05,
+      "loss": 1.293,
+      "step": 10976
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1151891428571428e-05,
+      "loss": 1.2778,
+      "step": 10977
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1151622857142858e-05,
+      "loss": 0.9529,
+      "step": 10978
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1151354285714287e-05,
+      "loss": 1.3147,
+      "step": 10979
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1151085714285715e-05,
+      "loss": 1.3098,
+      "step": 10980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1150817142857143e-05,
+      "loss": 1.2979,
+      "step": 10981
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1150548571428573e-05,
+      "loss": 1.262,
+      "step": 10982
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.115028e-05,
+      "loss": 1.4993,
+      "step": 10983
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1150011428571429e-05,
+      "loss": 0.9885,
+      "step": 10984
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1149742857142857e-05,
+      "loss": 1.0408,
+      "step": 10985
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1149474285714287e-05,
+      "loss": 1.3103,
+      "step": 10986
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1149205714285714e-05,
+      "loss": 1.4681,
+      "step": 10987
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1148937142857143e-05,
+      "loss": 1.298,
+      "step": 10988
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1148668571428572e-05,
+      "loss": 1.2402,
+      "step": 10989
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.11484e-05,
+      "loss": 0.8019,
+      "step": 10990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.114813142857143e-05,
+      "loss": 1.4184,
+      "step": 10991
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1147862857142858e-05,
+      "loss": 1.13,
+      "step": 10992
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1147594285714286e-05,
+      "loss": 1.2553,
+      "step": 10993
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1147325714285714e-05,
+      "loss": 1.8403,
+      "step": 10994
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1147057142857144e-05,
+      "loss": 1.2146,
+      "step": 10995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1146788571428572e-05,
+      "loss": 1.2478,
+      "step": 10996
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.114652e-05,
+      "loss": 1.3938,
+      "step": 10997
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1146251428571429e-05,
+      "loss": 1.4149,
+      "step": 10998
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1145982857142859e-05,
+      "loss": 1.1686,
+      "step": 10999
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1145714285714285e-05,
+      "loss": 1.1759,
+      "step": 11000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1145445714285715e-05,
+      "loss": 0.8635,
+      "step": 11001
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1145177142857143e-05,
+      "loss": 1.4357,
+      "step": 11002
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1144908571428571e-05,
+      "loss": 1.3398,
+      "step": 11003
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1144640000000001e-05,
+      "loss": 1.4065,
+      "step": 11004
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.114437142857143e-05,
+      "loss": 1.1727,
+      "step": 11005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1144102857142857e-05,
+      "loss": 1.6684,
+      "step": 11006
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1143834285714286e-05,
+      "loss": 1.2753,
+      "step": 11007
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1143565714285715e-05,
+      "loss": 1.2975,
+      "step": 11008
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1143297142857144e-05,
+      "loss": 1.5203,
+      "step": 11009
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1143028571428572e-05,
+      "loss": 1.0686,
+      "step": 11010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.114276e-05,
+      "loss": 1.0441,
+      "step": 11011
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.114249142857143e-05,
+      "loss": 1.5385,
+      "step": 11012
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1142222857142856e-05,
+      "loss": 1.2907,
+      "step": 11013
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1141954285714286e-05,
+      "loss": 1.2197,
+      "step": 11014
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1141685714285714e-05,
+      "loss": 1.4012,
+      "step": 11015
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1141417142857143e-05,
+      "loss": 1.2256,
+      "step": 11016
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1141148571428572e-05,
+      "loss": 1.29,
+      "step": 11017
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.114088e-05,
+      "loss": 1.3013,
+      "step": 11018
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.114061142857143e-05,
+      "loss": 0.9349,
+      "step": 11019
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1140342857142857e-05,
+      "loss": 1.0598,
+      "step": 11020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1140074285714287e-05,
+      "loss": 1.5907,
+      "step": 11021
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1139805714285715e-05,
+      "loss": 1.3231,
+      "step": 11022
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1139537142857143e-05,
+      "loss": 1.2085,
+      "step": 11023
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1139268571428571e-05,
+      "loss": 1.1797,
+      "step": 11024
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1139000000000001e-05,
+      "loss": 1.1995,
+      "step": 11025
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1138731428571428e-05,
+      "loss": 1.3362,
+      "step": 11026
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1138462857142858e-05,
+      "loss": 1.2069,
+      "step": 11027
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1138194285714286e-05,
+      "loss": 1.4395,
+      "step": 11028
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1137925714285714e-05,
+      "loss": 1.2033,
+      "step": 11029
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1137657142857144e-05,
+      "loss": 0.9144,
+      "step": 11030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1137388571428572e-05,
+      "loss": 1.291,
+      "step": 11031
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1137120000000002e-05,
+      "loss": 1.3298,
+      "step": 11032
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1136851428571428e-05,
+      "loss": 1.1007,
+      "step": 11033
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1136582857142858e-05,
+      "loss": 1.1948,
+      "step": 11034
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1136314285714286e-05,
+      "loss": 1.4822,
+      "step": 11035
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1136045714285715e-05,
+      "loss": 0.9873,
+      "step": 11036
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1135777142857143e-05,
+      "loss": 1.1507,
+      "step": 11037
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1135508571428573e-05,
+      "loss": 1.7115,
+      "step": 11038
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1135239999999999e-05,
+      "loss": 1.1425,
+      "step": 11039
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1134971428571429e-05,
+      "loss": 1.1163,
+      "step": 11040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1134702857142857e-05,
+      "loss": 1.4129,
+      "step": 11041
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1134434285714287e-05,
+      "loss": 1.1844,
+      "step": 11042
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1134165714285715e-05,
+      "loss": 1.0076,
+      "step": 11043
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1133897142857143e-05,
+      "loss": 0.9964,
+      "step": 11044
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1133628571428573e-05,
+      "loss": 1.0476,
+      "step": 11045
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.113336e-05,
+      "loss": 0.9357,
+      "step": 11046
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.113309142857143e-05,
+      "loss": 1.0718,
+      "step": 11047
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1132822857142858e-05,
+      "loss": 1.6405,
+      "step": 11048
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1132554285714286e-05,
+      "loss": 1.0469,
+      "step": 11049
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1132285714285714e-05,
+      "loss": 1.7294,
+      "step": 11050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1132017142857144e-05,
+      "loss": 0.9319,
+      "step": 11051
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.113174857142857e-05,
+      "loss": 1.6776,
+      "step": 11052
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.113148e-05,
+      "loss": 1.0106,
+      "step": 11053
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1131211428571429e-05,
+      "loss": 1.6363,
+      "step": 11054
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1130942857142858e-05,
+      "loss": 1.4358,
+      "step": 11055
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1130674285714287e-05,
+      "loss": 1.2364,
+      "step": 11056
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1130405714285715e-05,
+      "loss": 1.4042,
+      "step": 11057
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1130137142857145e-05,
+      "loss": 1.5006,
+      "step": 11058
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1129868571428571e-05,
+      "loss": 1.108,
+      "step": 11059
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1129600000000001e-05,
+      "loss": 1.6522,
+      "step": 11060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.112933142857143e-05,
+      "loss": 1.2123,
+      "step": 11061
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1129062857142857e-05,
+      "loss": 1.246,
+      "step": 11062
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1128794285714286e-05,
+      "loss": 1.2852,
+      "step": 11063
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1128525714285715e-05,
+      "loss": 1.4788,
+      "step": 11064
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1128257142857142e-05,
+      "loss": 1.3249,
+      "step": 11065
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1127988571428572e-05,
+      "loss": 1.1151,
+      "step": 11066
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.112772e-05,
+      "loss": 1.1503,
+      "step": 11067
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.112745142857143e-05,
+      "loss": 0.9734,
+      "step": 11068
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1127182857142858e-05,
+      "loss": 1.2239,
+      "step": 11069
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1126914285714286e-05,
+      "loss": 1.4221,
+      "step": 11070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1126645714285716e-05,
+      "loss": 1.2974,
+      "step": 11071
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1126377142857142e-05,
+      "loss": 1.0995,
+      "step": 11072
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1126108571428572e-05,
+      "loss": 1.0394,
+      "step": 11073
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.112584e-05,
+      "loss": 1.4853,
+      "step": 11074
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1125571428571429e-05,
+      "loss": 1.6154,
+      "step": 11075
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1125302857142857e-05,
+      "loss": 1.2243,
+      "step": 11076
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1125034285714287e-05,
+      "loss": 1.3961,
+      "step": 11077
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1124765714285715e-05,
+      "loss": 1.4803,
+      "step": 11078
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1124497142857143e-05,
+      "loss": 1.3527,
+      "step": 11079
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1124228571428571e-05,
+      "loss": 1.4956,
+      "step": 11080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1123960000000001e-05,
+      "loss": 1.2254,
+      "step": 11081
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.112369142857143e-05,
+      "loss": 1.3835,
+      "step": 11082
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1123422857142858e-05,
+      "loss": 1.2946,
+      "step": 11083
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1123154285714287e-05,
+      "loss": 0.8919,
+      "step": 11084
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1122885714285714e-05,
+      "loss": 1.5265,
+      "step": 11085
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1122617142857144e-05,
+      "loss": 0.9621,
+      "step": 11086
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1122348571428572e-05,
+      "loss": 1.5731,
+      "step": 11087
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.112208e-05,
+      "loss": 1.3367,
+      "step": 11088
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1121811428571428e-05,
+      "loss": 0.8788,
+      "step": 11089
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1121542857142858e-05,
+      "loss": 1.2546,
+      "step": 11090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1121274285714286e-05,
+      "loss": 1.4833,
+      "step": 11091
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1121005714285714e-05,
+      "loss": 1.3905,
+      "step": 11092
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1120737142857143e-05,
+      "loss": 1.1612,
+      "step": 11093
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1120468571428573e-05,
+      "loss": 1.2995,
+      "step": 11094
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.11202e-05,
+      "loss": 1.1569,
+      "step": 11095
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1119931428571429e-05,
+      "loss": 0.7372,
+      "step": 11096
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1119662857142859e-05,
+      "loss": 1.5705,
+      "step": 11097
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1119394285714285e-05,
+      "loss": 1.0267,
+      "step": 11098
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1119125714285715e-05,
+      "loss": 1.2216,
+      "step": 11099
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1118857142857143e-05,
+      "loss": 1.057,
+      "step": 11100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1118588571428573e-05,
+      "loss": 1.342,
+      "step": 11101
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.111832e-05,
+      "loss": 1.1119,
+      "step": 11102
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.111805142857143e-05,
+      "loss": 1.1795,
+      "step": 11103
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1117782857142858e-05,
+      "loss": 0.9011,
+      "step": 11104
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1117514285714286e-05,
+      "loss": 1.4295,
+      "step": 11105
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1117245714285714e-05,
+      "loss": 1.0276,
+      "step": 11106
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1116977142857144e-05,
+      "loss": 1.4115,
+      "step": 11107
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.111670857142857e-05,
+      "loss": 1.0722,
+      "step": 11108
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.111644e-05,
+      "loss": 1.2412,
+      "step": 11109
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.111617142857143e-05,
+      "loss": 1.5677,
+      "step": 11110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1115902857142857e-05,
+      "loss": 1.3526,
+      "step": 11111
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.1115634285714286e-05,
+      "loss": 1.5595,
+      "step": 11112
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1115365714285715e-05,
+      "loss": 1.4631,
+      "step": 11113
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1115097142857145e-05,
+      "loss": 1.1875,
+      "step": 11114
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1114828571428571e-05,
+      "loss": 1.4116,
+      "step": 11115
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1114560000000001e-05,
+      "loss": 0.946,
+      "step": 11116
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1114291428571429e-05,
+      "loss": 1.1728,
+      "step": 11117
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1114022857142857e-05,
+      "loss": 1.1925,
+      "step": 11118
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1113754285714285e-05,
+      "loss": 0.7466,
+      "step": 11119
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1113485714285715e-05,
+      "loss": 1.3434,
+      "step": 11120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1113217142857142e-05,
+      "loss": 1.5253,
+      "step": 11121
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1112948571428572e-05,
+      "loss": 1.3696,
+      "step": 11122
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1112680000000001e-05,
+      "loss": 1.33,
+      "step": 11123
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.111241142857143e-05,
+      "loss": 0.9491,
+      "step": 11124
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1112142857142858e-05,
+      "loss": 1.5316,
+      "step": 11125
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1111874285714286e-05,
+      "loss": 1.0805,
+      "step": 11126
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1111605714285716e-05,
+      "loss": 1.6262,
+      "step": 11127
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1111337142857142e-05,
+      "loss": 1.2818,
+      "step": 11128
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1111068571428572e-05,
+      "loss": 1.6124,
+      "step": 11129
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.11108e-05,
+      "loss": 1.1806,
+      "step": 11130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1110531428571429e-05,
+      "loss": 1.026,
+      "step": 11131
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1110262857142857e-05,
+      "loss": 1.3227,
+      "step": 11132
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1109994285714287e-05,
+      "loss": 1.2825,
+      "step": 11133
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1109725714285713e-05,
+      "loss": 1.5258,
+      "step": 11134
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1109457142857143e-05,
+      "loss": 1.758,
+      "step": 11135
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1109188571428573e-05,
+      "loss": 1.0602,
+      "step": 11136
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1108920000000001e-05,
+      "loss": 1.3489,
+      "step": 11137
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.110865142857143e-05,
+      "loss": 1.214,
+      "step": 11138
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1108382857142857e-05,
+      "loss": 1.1372,
+      "step": 11139
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1108114285714287e-05,
+      "loss": 1.2252,
+      "step": 11140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1107845714285714e-05,
+      "loss": 1.4535,
+      "step": 11141
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1107577142857144e-05,
+      "loss": 0.9634,
+      "step": 11142
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1107308571428572e-05,
+      "loss": 1.2243,
+      "step": 11143
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.110704e-05,
+      "loss": 1.2873,
+      "step": 11144
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1106771428571428e-05,
+      "loss": 1.0191,
+      "step": 11145
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1106502857142858e-05,
+      "loss": 1.0797,
+      "step": 11146
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1106234285714286e-05,
+      "loss": 1.5315,
+      "step": 11147
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1105965714285714e-05,
+      "loss": 1.0476,
+      "step": 11148
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1105697142857144e-05,
+      "loss": 1.3117,
+      "step": 11149
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1105428571428572e-05,
+      "loss": 1.2762,
+      "step": 11150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.110516e-05,
+      "loss": 1.1048,
+      "step": 11151
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1104891428571429e-05,
+      "loss": 1.4132,
+      "step": 11152
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1104622857142859e-05,
+      "loss": 1.365,
+      "step": 11153
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1104354285714285e-05,
+      "loss": 1.3491,
+      "step": 11154
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1104085714285715e-05,
+      "loss": 1.659,
+      "step": 11155
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1103817142857143e-05,
+      "loss": 1.4186,
+      "step": 11156
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1103548571428571e-05,
+      "loss": 1.0475,
+      "step": 11157
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.110328e-05,
+      "loss": 1.1375,
+      "step": 11158
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.110301142857143e-05,
+      "loss": 1.3535,
+      "step": 11159
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1102742857142858e-05,
+      "loss": 1.2858,
+      "step": 11160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1102474285714286e-05,
+      "loss": 1.1379,
+      "step": 11161
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1102205714285716e-05,
+      "loss": 1.0516,
+      "step": 11162
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1101937142857144e-05,
+      "loss": 1.1573,
+      "step": 11163
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1101668571428572e-05,
+      "loss": 0.7478,
+      "step": 11164
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.11014e-05,
+      "loss": 1.2588,
+      "step": 11165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.110113142857143e-05,
+      "loss": 1.0364,
+      "step": 11166
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1100862857142856e-05,
+      "loss": 1.6063,
+      "step": 11167
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1100594285714286e-05,
+      "loss": 1.3663,
+      "step": 11168
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1100325714285715e-05,
+      "loss": 0.9893,
+      "step": 11169
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1100057142857144e-05,
+      "loss": 1.1067,
+      "step": 11170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1099788571428571e-05,
+      "loss": 1.3798,
+      "step": 11171
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.109952e-05,
+      "loss": 1.6824,
+      "step": 11172
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1099251428571429e-05,
+      "loss": 1.4714,
+      "step": 11173
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1098982857142857e-05,
+      "loss": 1.4848,
+      "step": 11174
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1098714285714287e-05,
+      "loss": 1.4505,
+      "step": 11175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1098445714285715e-05,
+      "loss": 1.0116,
+      "step": 11176
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1098177142857143e-05,
+      "loss": 1.5166,
+      "step": 11177
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1097908571428572e-05,
+      "loss": 1.5747,
+      "step": 11178
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1097640000000001e-05,
+      "loss": 1.358,
+      "step": 11179
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1097371428571428e-05,
+      "loss": 1.6086,
+      "step": 11180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1097102857142858e-05,
+      "loss": 1.2065,
+      "step": 11181
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1096834285714286e-05,
+      "loss": 1.3109,
+      "step": 11182
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1096565714285716e-05,
+      "loss": 1.1724,
+      "step": 11183
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1096297142857142e-05,
+      "loss": 1.439,
+      "step": 11184
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1096028571428572e-05,
+      "loss": 1.1468,
+      "step": 11185
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.109576e-05,
+      "loss": 1.0045,
+      "step": 11186
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1095491428571428e-05,
+      "loss": 0.7787,
+      "step": 11187
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1095222857142858e-05,
+      "loss": 1.3836,
+      "step": 11188
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1094954285714287e-05,
+      "loss": 1.2664,
+      "step": 11189
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1094685714285715e-05,
+      "loss": 1.138,
+      "step": 11190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1094417142857143e-05,
+      "loss": 1.2673,
+      "step": 11191
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1094148571428573e-05,
+      "loss": 1.372,
+      "step": 11192
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1093880000000001e-05,
+      "loss": 1.3392,
+      "step": 11193
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1093611428571429e-05,
+      "loss": 1.2558,
+      "step": 11194
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1093342857142857e-05,
+      "loss": 1.3278,
+      "step": 11195
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1093074285714287e-05,
+      "loss": 1.9086,
+      "step": 11196
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1092805714285714e-05,
+      "loss": 0.7576,
+      "step": 11197
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1092537142857143e-05,
+      "loss": 1.2703,
+      "step": 11198
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1092268571428572e-05,
+      "loss": 1.5202,
+      "step": 11199
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1092e-05,
+      "loss": 1.4677,
+      "step": 11200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.109173142857143e-05,
+      "loss": 1.309,
+      "step": 11201
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1091462857142858e-05,
+      "loss": 1.2231,
+      "step": 11202
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1091194285714286e-05,
+      "loss": 1.4381,
+      "step": 11203
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1090925714285714e-05,
+      "loss": 1.1012,
+      "step": 11204
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1090657142857144e-05,
+      "loss": 1.0981,
+      "step": 11205
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1090388571428572e-05,
+      "loss": 1.0904,
+      "step": 11206
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.109012e-05,
+      "loss": 1.2768,
+      "step": 11207
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1089851428571429e-05,
+      "loss": 0.9827,
+      "step": 11208
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1089582857142859e-05,
+      "loss": 0.8126,
+      "step": 11209
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1089314285714285e-05,
+      "loss": 1.1136,
+      "step": 11210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1089045714285715e-05,
+      "loss": 1.1618,
+      "step": 11211
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1088777142857143e-05,
+      "loss": 1.3083,
+      "step": 11212
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1088508571428571e-05,
+      "loss": 1.491,
+      "step": 11213
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.108824e-05,
+      "loss": 1.0167,
+      "step": 11214
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.108797142857143e-05,
+      "loss": 1.3092,
+      "step": 11215
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1087702857142857e-05,
+      "loss": 1.4966,
+      "step": 11216
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1087434285714286e-05,
+      "loss": 1.0828,
+      "step": 11217
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1087165714285715e-05,
+      "loss": 1.6495,
+      "step": 11218
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1086897142857144e-05,
+      "loss": 1.3535,
+      "step": 11219
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1086628571428572e-05,
+      "loss": 1.192,
+      "step": 11220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.108636e-05,
+      "loss": 0.978,
+      "step": 11221
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.108609142857143e-05,
+      "loss": 1.4007,
+      "step": 11222
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1085822857142856e-05,
+      "loss": 1.6698,
+      "step": 11223
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1085554285714286e-05,
+      "loss": 1.1163,
+      "step": 11224
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1085285714285714e-05,
+      "loss": 0.9961,
+      "step": 11225
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1085017142857143e-05,
+      "loss": 1.2978,
+      "step": 11226
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.108474857142857e-05,
+      "loss": 0.9523,
+      "step": 11227
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.108448e-05,
+      "loss": 1.0697,
+      "step": 11228
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.108421142857143e-05,
+      "loss": 1.6208,
+      "step": 11229
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1083942857142857e-05,
+      "loss": 1.1003,
+      "step": 11230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1083674285714287e-05,
+      "loss": 1.2457,
+      "step": 11231
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1083405714285715e-05,
+      "loss": 1.1334,
+      "step": 11232
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1083137142857143e-05,
+      "loss": 1.4142,
+      "step": 11233
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1082868571428571e-05,
+      "loss": 1.278,
+      "step": 11234
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1082600000000001e-05,
+      "loss": 1.3255,
+      "step": 11235
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1082331428571428e-05,
+      "loss": 1.0835,
+      "step": 11236
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1082062857142858e-05,
+      "loss": 1.1713,
+      "step": 11237
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1081794285714286e-05,
+      "loss": 1.2708,
+      "step": 11238
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1081525714285714e-05,
+      "loss": 1.8231,
+      "step": 11239
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1081257142857142e-05,
+      "loss": 1.4706,
+      "step": 11240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1080988571428572e-05,
+      "loss": 1.2764,
+      "step": 11241
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1080720000000002e-05,
+      "loss": 1.4255,
+      "step": 11242
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1080451428571428e-05,
+      "loss": 1.2718,
+      "step": 11243
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1080182857142858e-05,
+      "loss": 1.3835,
+      "step": 11244
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1079914285714286e-05,
+      "loss": 1.2168,
+      "step": 11245
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1079645714285715e-05,
+      "loss": 1.0616,
+      "step": 11246
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1079377142857143e-05,
+      "loss": 1.1948,
+      "step": 11247
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1079108571428573e-05,
+      "loss": 1.0706,
+      "step": 11248
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1078839999999999e-05,
+      "loss": 1.6649,
+      "step": 11249
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1078571428571429e-05,
+      "loss": 1.3504,
+      "step": 11250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1078302857142857e-05,
+      "loss": 0.9992,
+      "step": 11251
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1078034285714287e-05,
+      "loss": 1.1309,
+      "step": 11252
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1077765714285714e-05,
+      "loss": 1.3555,
+      "step": 11253
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1077497142857143e-05,
+      "loss": 0.883,
+      "step": 11254
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1077228571428573e-05,
+      "loss": 1.4886,
+      "step": 11255
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.107696e-05,
+      "loss": 1.1694,
+      "step": 11256
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.107669142857143e-05,
+      "loss": 1.0546,
+      "step": 11257
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1076422857142858e-05,
+      "loss": 0.7588,
+      "step": 11258
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1076154285714286e-05,
+      "loss": 1.5533,
+      "step": 11259
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1075885714285714e-05,
+      "loss": 1.2601,
+      "step": 11260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1075617142857144e-05,
+      "loss": 1.2705,
+      "step": 11261
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.107534857142857e-05,
+      "loss": 1.6904,
+      "step": 11262
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.107508e-05,
+      "loss": 1.4228,
+      "step": 11263
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1074811428571429e-05,
+      "loss": 1.2272,
+      "step": 11264
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1074542857142858e-05,
+      "loss": 1.164,
+      "step": 11265
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1074274285714285e-05,
+      "loss": 1.3249,
+      "step": 11266
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1074005714285715e-05,
+      "loss": 0.9668,
+      "step": 11267
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1073737142857145e-05,
+      "loss": 1.3419,
+      "step": 11268
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1073468571428571e-05,
+      "loss": 1.1944,
+      "step": 11269
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1073200000000001e-05,
+      "loss": 1.2743,
+      "step": 11270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.107293142857143e-05,
+      "loss": 1.0884,
+      "step": 11271
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1072662857142857e-05,
+      "loss": 1.6357,
+      "step": 11272
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1072394285714286e-05,
+      "loss": 1.4169,
+      "step": 11273
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1072125714285715e-05,
+      "loss": 1.7084,
+      "step": 11274
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1071857142857144e-05,
+      "loss": 1.547,
+      "step": 11275
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1071588571428572e-05,
+      "loss": 0.8702,
+      "step": 11276
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.107132e-05,
+      "loss": 1.581,
+      "step": 11277
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.107105142857143e-05,
+      "loss": 1.2859,
+      "step": 11278
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1070782857142856e-05,
+      "loss": 1.1935,
+      "step": 11279
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1070514285714286e-05,
+      "loss": 1.3071,
+      "step": 11280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1070245714285716e-05,
+      "loss": 0.9181,
+      "step": 11281
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1069977142857142e-05,
+      "loss": 1.1982,
+      "step": 11282
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1069708571428572e-05,
+      "loss": 1.2769,
+      "step": 11283
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.106944e-05,
+      "loss": 1.2556,
+      "step": 11284
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1069171428571429e-05,
+      "loss": 0.8558,
+      "step": 11285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1068902857142857e-05,
+      "loss": 1.0161,
+      "step": 11286
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.1068634285714287e-05,
+      "loss": 0.9083,
+      "step": 11287
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1068365714285715e-05,
+      "loss": 1.203,
+      "step": 11288
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1068097142857143e-05,
+      "loss": 1.3321,
+      "step": 11289
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1067828571428571e-05,
+      "loss": 1.3634,
+      "step": 11290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1067560000000001e-05,
+      "loss": 1.1484,
+      "step": 11291
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1067291428571428e-05,
+      "loss": 1.0006,
+      "step": 11292
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1067022857142857e-05,
+      "loss": 1.139,
+      "step": 11293
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1066754285714287e-05,
+      "loss": 1.2188,
+      "step": 11294
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1066485714285714e-05,
+      "loss": 1.3201,
+      "step": 11295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1066217142857144e-05,
+      "loss": 1.3154,
+      "step": 11296
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1065948571428572e-05,
+      "loss": 1.2448,
+      "step": 11297
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1065680000000002e-05,
+      "loss": 1.1061,
+      "step": 11298
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1065411428571428e-05,
+      "loss": 1.0695,
+      "step": 11299
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1065142857142858e-05,
+      "loss": 1.5197,
+      "step": 11300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1064874285714286e-05,
+      "loss": 1.3576,
+      "step": 11301
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1064605714285714e-05,
+      "loss": 1.2408,
+      "step": 11302
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1064337142857143e-05,
+      "loss": 1.139,
+      "step": 11303
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1064068571428573e-05,
+      "loss": 1.0896,
+      "step": 11304
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1063799999999999e-05,
+      "loss": 0.9984,
+      "step": 11305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1063531428571429e-05,
+      "loss": 0.8513,
+      "step": 11306
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1063262857142859e-05,
+      "loss": 1.1368,
+      "step": 11307
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1062994285714285e-05,
+      "loss": 0.9654,
+      "step": 11308
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1062725714285715e-05,
+      "loss": 1.0295,
+      "step": 11309
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1062457142857143e-05,
+      "loss": 1.3048,
+      "step": 11310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1062188571428573e-05,
+      "loss": 1.0849,
+      "step": 11311
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.106192e-05,
+      "loss": 1.0916,
+      "step": 11312
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.106165142857143e-05,
+      "loss": 0.957,
+      "step": 11313
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1061382857142858e-05,
+      "loss": 0.8324,
+      "step": 11314
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1061114285714286e-05,
+      "loss": 1.002,
+      "step": 11315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1060845714285714e-05,
+      "loss": 1.6066,
+      "step": 11316
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1060577142857144e-05,
+      "loss": 1.058,
+      "step": 11317
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.106030857142857e-05,
+      "loss": 0.9866,
+      "step": 11318
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.106004e-05,
+      "loss": 1.283,
+      "step": 11319
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1059771428571428e-05,
+      "loss": 1.0225,
+      "step": 11320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1059502857142858e-05,
+      "loss": 1.3069,
+      "step": 11321
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1059234285714286e-05,
+      "loss": 0.9175,
+      "step": 11322
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1058965714285715e-05,
+      "loss": 1.361,
+      "step": 11323
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1058697142857145e-05,
+      "loss": 1.0613,
+      "step": 11324
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1058428571428571e-05,
+      "loss": 1.3365,
+      "step": 11325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1058160000000001e-05,
+      "loss": 0.9702,
+      "step": 11326
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1057891428571429e-05,
+      "loss": 1.2065,
+      "step": 11327
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1057622857142857e-05,
+      "loss": 1.3573,
+      "step": 11328
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1057354285714285e-05,
+      "loss": 0.965,
+      "step": 11329
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1057085714285715e-05,
+      "loss": 1.3459,
+      "step": 11330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1056817142857142e-05,
+      "loss": 1.4214,
+      "step": 11331
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1056548571428572e-05,
+      "loss": 1.6792,
+      "step": 11332
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.105628e-05,
+      "loss": 0.8376,
+      "step": 11333
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.105601142857143e-05,
+      "loss": 1.6353,
+      "step": 11334
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1055742857142858e-05,
+      "loss": 1.21,
+      "step": 11335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1055474285714286e-05,
+      "loss": 1.6629,
+      "step": 11336
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1055205714285716e-05,
+      "loss": 1.5003,
+      "step": 11337
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1054937142857142e-05,
+      "loss": 1.0928,
+      "step": 11338
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1054668571428572e-05,
+      "loss": 1.4362,
+      "step": 11339
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.10544e-05,
+      "loss": 1.5083,
+      "step": 11340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1054131428571429e-05,
+      "loss": 1.3602,
+      "step": 11341
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1053862857142857e-05,
+      "loss": 1.5775,
+      "step": 11342
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1053594285714287e-05,
+      "loss": 1.2345,
+      "step": 11343
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1053325714285715e-05,
+      "loss": 1.2492,
+      "step": 11344
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1053057142857143e-05,
+      "loss": 1.3323,
+      "step": 11345
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1052788571428571e-05,
+      "loss": 1.2862,
+      "step": 11346
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1052520000000001e-05,
+      "loss": 1.3621,
+      "step": 11347
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.105225142857143e-05,
+      "loss": 1.3969,
+      "step": 11348
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1051982857142857e-05,
+      "loss": 1.0337,
+      "step": 11349
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1051714285714287e-05,
+      "loss": 1.3061,
+      "step": 11350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1051445714285714e-05,
+      "loss": 1.184,
+      "step": 11351
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1051177142857144e-05,
+      "loss": 0.9044,
+      "step": 11352
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1050908571428572e-05,
+      "loss": 1.6188,
+      "step": 11353
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.105064e-05,
+      "loss": 1.3899,
+      "step": 11354
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1050371428571428e-05,
+      "loss": 1.4205,
+      "step": 11355
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1050102857142858e-05,
+      "loss": 1.4203,
+      "step": 11356
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1049834285714286e-05,
+      "loss": 0.9637,
+      "step": 11357
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1049565714285714e-05,
+      "loss": 0.864,
+      "step": 11358
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1049297142857143e-05,
+      "loss": 1.1887,
+      "step": 11359
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1049028571428572e-05,
+      "loss": 1.2765,
+      "step": 11360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.104876e-05,
+      "loss": 1.2063,
+      "step": 11361
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1048491428571429e-05,
+      "loss": 1.2197,
+      "step": 11362
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1048222857142859e-05,
+      "loss": 0.8904,
+      "step": 11363
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1047954285714285e-05,
+      "loss": 1.0225,
+      "step": 11364
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1047685714285715e-05,
+      "loss": 1.0991,
+      "step": 11365
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1047417142857143e-05,
+      "loss": 1.1512,
+      "step": 11366
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1047148571428573e-05,
+      "loss": 0.9879,
+      "step": 11367
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.104688e-05,
+      "loss": 1.3386,
+      "step": 11368
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.104661142857143e-05,
+      "loss": 0.9419,
+      "step": 11369
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1046342857142858e-05,
+      "loss": 1.0954,
+      "step": 11370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1046074285714286e-05,
+      "loss": 1.385,
+      "step": 11371
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1045805714285714e-05,
+      "loss": 1.6862,
+      "step": 11372
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1045537142857144e-05,
+      "loss": 1.6513,
+      "step": 11373
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1045268571428572e-05,
+      "loss": 1.3761,
+      "step": 11374
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1045e-05,
+      "loss": 1.1745,
+      "step": 11375
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.104473142857143e-05,
+      "loss": 1.3948,
+      "step": 11376
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1044462857142856e-05,
+      "loss": 1.0078,
+      "step": 11377
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1044194285714286e-05,
+      "loss": 1.4059,
+      "step": 11378
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1043925714285715e-05,
+      "loss": 1.324,
+      "step": 11379
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1043657142857144e-05,
+      "loss": 1.5019,
+      "step": 11380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1043388571428571e-05,
+      "loss": 1.1016,
+      "step": 11381
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.104312e-05,
+      "loss": 1.3805,
+      "step": 11382
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1042851428571429e-05,
+      "loss": 1.0435,
+      "step": 11383
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1042582857142857e-05,
+      "loss": 1.1938,
+      "step": 11384
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1042314285714285e-05,
+      "loss": 1.6946,
+      "step": 11385
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1042045714285715e-05,
+      "loss": 0.7923,
+      "step": 11386
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1041777142857143e-05,
+      "loss": 1.4804,
+      "step": 11387
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1041508571428572e-05,
+      "loss": 1.574,
+      "step": 11388
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1041240000000001e-05,
+      "loss": 1.2595,
+      "step": 11389
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1040971428571428e-05,
+      "loss": 1.4194,
+      "step": 11390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1040702857142858e-05,
+      "loss": 1.6414,
+      "step": 11391
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1040434285714286e-05,
+      "loss": 1.5564,
+      "step": 11392
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1040165714285716e-05,
+      "loss": 1.3317,
+      "step": 11393
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1039897142857142e-05,
+      "loss": 1.2162,
+      "step": 11394
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1039628571428572e-05,
+      "loss": 1.1599,
+      "step": 11395
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.103936e-05,
+      "loss": 0.8231,
+      "step": 11396
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1039091428571428e-05,
+      "loss": 0.8616,
+      "step": 11397
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1038822857142857e-05,
+      "loss": 1.8002,
+      "step": 11398
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1038554285714287e-05,
+      "loss": 1.252,
+      "step": 11399
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1038285714285715e-05,
+      "loss": 0.8845,
+      "step": 11400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1038017142857143e-05,
+      "loss": 1.3078,
+      "step": 11401
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1037748571428573e-05,
+      "loss": 1.5613,
+      "step": 11402
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1037480000000001e-05,
+      "loss": 1.1095,
+      "step": 11403
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1037211428571429e-05,
+      "loss": 1.5626,
+      "step": 11404
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1036942857142857e-05,
+      "loss": 1.3029,
+      "step": 11405
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1036674285714287e-05,
+      "loss": 1.2495,
+      "step": 11406
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1036405714285714e-05,
+      "loss": 1.0287,
+      "step": 11407
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1036137142857143e-05,
+      "loss": 1.0789,
+      "step": 11408
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1035868571428572e-05,
+      "loss": 1.3907,
+      "step": 11409
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.10356e-05,
+      "loss": 1.2598,
+      "step": 11410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1035331428571428e-05,
+      "loss": 1.6943,
+      "step": 11411
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1035062857142858e-05,
+      "loss": 1.3808,
+      "step": 11412
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1034794285714286e-05,
+      "loss": 1.3205,
+      "step": 11413
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1034525714285714e-05,
+      "loss": 1.1998,
+      "step": 11414
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1034257142857144e-05,
+      "loss": 1.4473,
+      "step": 11415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1033988571428572e-05,
+      "loss": 2.002,
+      "step": 11416
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.103372e-05,
+      "loss": 1.5758,
+      "step": 11417
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1033451428571429e-05,
+      "loss": 1.5562,
+      "step": 11418
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1033182857142859e-05,
+      "loss": 0.7549,
+      "step": 11419
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1032914285714285e-05,
+      "loss": 1.365,
+      "step": 11420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1032645714285715e-05,
+      "loss": 1.2663,
+      "step": 11421
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1032377142857143e-05,
+      "loss": 1.5205,
+      "step": 11422
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1032108571428571e-05,
+      "loss": 1.5107,
+      "step": 11423
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.103184e-05,
+      "loss": 1.4285,
+      "step": 11424
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.103157142857143e-05,
+      "loss": 1.2602,
+      "step": 11425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1031302857142857e-05,
+      "loss": 1.2611,
+      "step": 11426
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1031034285714286e-05,
+      "loss": 1.1276,
+      "step": 11427
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1030765714285715e-05,
+      "loss": 1.1633,
+      "step": 11428
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1030497142857144e-05,
+      "loss": 1.4412,
+      "step": 11429
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1030228571428572e-05,
+      "loss": 1.5441,
+      "step": 11430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.102996e-05,
+      "loss": 1.3153,
+      "step": 11431
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.102969142857143e-05,
+      "loss": 1.2818,
+      "step": 11432
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1029422857142856e-05,
+      "loss": 1.4969,
+      "step": 11433
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1029154285714286e-05,
+      "loss": 1.1692,
+      "step": 11434
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1028885714285714e-05,
+      "loss": 1.5372,
+      "step": 11435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1028617142857143e-05,
+      "loss": 1.47,
+      "step": 11436
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.102834857142857e-05,
+      "loss": 0.999,
+      "step": 11437
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.102808e-05,
+      "loss": 1.281,
+      "step": 11438
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1027811428571429e-05,
+      "loss": 1.5389,
+      "step": 11439
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1027542857142857e-05,
+      "loss": 0.9354,
+      "step": 11440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1027274285714287e-05,
+      "loss": 1.255,
+      "step": 11441
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1027005714285715e-05,
+      "loss": 1.1847,
+      "step": 11442
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1026737142857143e-05,
+      "loss": 1.1438,
+      "step": 11443
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1026468571428571e-05,
+      "loss": 1.2808,
+      "step": 11444
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1026200000000001e-05,
+      "loss": 1.0372,
+      "step": 11445
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1025931428571428e-05,
+      "loss": 1.4535,
+      "step": 11446
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1025662857142858e-05,
+      "loss": 1.1716,
+      "step": 11447
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1025394285714286e-05,
+      "loss": 0.8948,
+      "step": 11448
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1025125714285716e-05,
+      "loss": 1.294,
+      "step": 11449
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1024857142857142e-05,
+      "loss": 1.2823,
+      "step": 11450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1024588571428572e-05,
+      "loss": 1.1224,
+      "step": 11451
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.102432e-05,
+      "loss": 1.2363,
+      "step": 11452
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1024051428571428e-05,
+      "loss": 1.8432,
+      "step": 11453
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1023782857142858e-05,
+      "loss": 1.0481,
+      "step": 11454
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1023514285714286e-05,
+      "loss": 1.05,
+      "step": 11455
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1023245714285715e-05,
+      "loss": 0.8901,
+      "step": 11456
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1022977142857143e-05,
+      "loss": 1.2987,
+      "step": 11457
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1022708571428573e-05,
+      "loss": 1.2489,
+      "step": 11458
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1022439999999999e-05,
+      "loss": 0.59,
+      "step": 11459
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1022171428571429e-05,
+      "loss": 1.3235,
+      "step": 11460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1021902857142857e-05,
+      "loss": 1.2824,
+      "step": 11461
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.1021634285714287e-05,
+      "loss": 1.3279,
+      "step": 11462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1021365714285714e-05,
+      "loss": 1.3014,
+      "step": 11463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1021097142857143e-05,
+      "loss": 1.6099,
+      "step": 11464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1020828571428572e-05,
+      "loss": 1.3841,
+      "step": 11465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.102056e-05,
+      "loss": 1.4599,
+      "step": 11466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.102029142857143e-05,
+      "loss": 1.5046,
+      "step": 11467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1020022857142858e-05,
+      "loss": 1.4543,
+      "step": 11468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1019754285714286e-05,
+      "loss": 0.9263,
+      "step": 11469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1019485714285714e-05,
+      "loss": 0.9674,
+      "step": 11470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1019217142857144e-05,
+      "loss": 1.0156,
+      "step": 11471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1018948571428572e-05,
+      "loss": 1.124,
+      "step": 11472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.101868e-05,
+      "loss": 1.262,
+      "step": 11473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1018411428571429e-05,
+      "loss": 1.082,
+      "step": 11474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1018142857142858e-05,
+      "loss": 0.9159,
+      "step": 11475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1017874285714285e-05,
+      "loss": 1.1839,
+      "step": 11476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1017605714285715e-05,
+      "loss": 1.2344,
+      "step": 11477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1017337142857143e-05,
+      "loss": 1.2966,
+      "step": 11478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1017068571428571e-05,
+      "loss": 1.4874,
+      "step": 11479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1016800000000001e-05,
+      "loss": 1.1952,
+      "step": 11480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1016531428571429e-05,
+      "loss": 1.1108,
+      "step": 11481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1016262857142857e-05,
+      "loss": 1.3355,
+      "step": 11482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1015994285714286e-05,
+      "loss": 1.5453,
+      "step": 11483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1015725714285715e-05,
+      "loss": 1.6554,
+      "step": 11484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1015457142857144e-05,
+      "loss": 0.8568,
+      "step": 11485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1015188571428572e-05,
+      "loss": 1.1283,
+      "step": 11486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.101492e-05,
+      "loss": 1.2389,
+      "step": 11487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.101465142857143e-05,
+      "loss": 1.288,
+      "step": 11488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1014382857142856e-05,
+      "loss": 1.2801,
+      "step": 11489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1014114285714286e-05,
+      "loss": 1.8213,
+      "step": 11490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1013845714285714e-05,
+      "loss": 1.4763,
+      "step": 11491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1013577142857142e-05,
+      "loss": 1.462,
+      "step": 11492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1013308571428572e-05,
+      "loss": 1.7335,
+      "step": 11493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.101304e-05,
+      "loss": 1.4022,
+      "step": 11494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.101277142857143e-05,
+      "loss": 0.9877,
+      "step": 11495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1012502857142857e-05,
+      "loss": 1.2709,
+      "step": 11496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1012234285714287e-05,
+      "loss": 0.8683,
+      "step": 11497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1011965714285715e-05,
+      "loss": 1.1001,
+      "step": 11498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1011697142857143e-05,
+      "loss": 0.9158,
+      "step": 11499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1011428571428571e-05,
+      "loss": 1.0698,
+      "step": 11500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1011160000000001e-05,
+      "loss": 1.4139,
+      "step": 11501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1010891428571428e-05,
+      "loss": 1.1366,
+      "step": 11502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1010622857142857e-05,
+      "loss": 1.2781,
+      "step": 11503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1010354285714286e-05,
+      "loss": 1.0567,
+      "step": 11504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1010085714285714e-05,
+      "loss": 1.4221,
+      "step": 11505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1009817142857144e-05,
+      "loss": 1.2786,
+      "step": 11506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1009548571428572e-05,
+      "loss": 1.4118,
+      "step": 11507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1009280000000002e-05,
+      "loss": 1.424,
+      "step": 11508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1009011428571428e-05,
+      "loss": 1.2189,
+      "step": 11509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1008742857142858e-05,
+      "loss": 1.2845,
+      "step": 11510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1008474285714286e-05,
+      "loss": 1.2944,
+      "step": 11511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1008205714285714e-05,
+      "loss": 0.9196,
+      "step": 11512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1007937142857143e-05,
+      "loss": 1.0975,
+      "step": 11513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1007668571428573e-05,
+      "loss": 1.2495,
+      "step": 11514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1007399999999999e-05,
+      "loss": 1.5102,
+      "step": 11515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1007131428571429e-05,
+      "loss": 1.0975,
+      "step": 11516
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1006862857142857e-05,
+      "loss": 1.3162,
+      "step": 11517
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1006594285714287e-05,
+      "loss": 1.5004,
+      "step": 11518
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1006325714285715e-05,
+      "loss": 1.2641,
+      "step": 11519
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1006057142857143e-05,
+      "loss": 1.4011,
+      "step": 11520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1005788571428573e-05,
+      "loss": 1.2887,
+      "step": 11521
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.100552e-05,
+      "loss": 1.2129,
+      "step": 11522
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.100525142857143e-05,
+      "loss": 0.8957,
+      "step": 11523
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1004982857142858e-05,
+      "loss": 1.1543,
+      "step": 11524
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1004714285714286e-05,
+      "loss": 1.0447,
+      "step": 11525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1004445714285714e-05,
+      "loss": 1.1775,
+      "step": 11526
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1004177142857144e-05,
+      "loss": 0.9045,
+      "step": 11527
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.100390857142857e-05,
+      "loss": 0.8284,
+      "step": 11528
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.100364e-05,
+      "loss": 1.2097,
+      "step": 11529
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1003371428571428e-05,
+      "loss": 1.0646,
+      "step": 11530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1003102857142858e-05,
+      "loss": 1.0882,
+      "step": 11531
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1002834285714286e-05,
+      "loss": 0.7826,
+      "step": 11532
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1002565714285715e-05,
+      "loss": 0.9228,
+      "step": 11533
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1002297142857145e-05,
+      "loss": 1.0758,
+      "step": 11534
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1002028571428571e-05,
+      "loss": 2.0833,
+      "step": 11535
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1001760000000001e-05,
+      "loss": 0.9442,
+      "step": 11536
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1001491428571429e-05,
+      "loss": 1.2734,
+      "step": 11537
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1001222857142857e-05,
+      "loss": 1.3648,
+      "step": 11538
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1000954285714285e-05,
+      "loss": 1.3777,
+      "step": 11539
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1000685714285715e-05,
+      "loss": 1.1158,
+      "step": 11540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1000417142857143e-05,
+      "loss": 1.3274,
+      "step": 11541
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.1000148571428572e-05,
+      "loss": 1.3485,
+      "step": 11542
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.099988e-05,
+      "loss": 1.465,
+      "step": 11543
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.099961142857143e-05,
+      "loss": 1.1911,
+      "step": 11544
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0999342857142858e-05,
+      "loss": 1.28,
+      "step": 11545
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0999074285714286e-05,
+      "loss": 1.3244,
+      "step": 11546
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0998805714285716e-05,
+      "loss": 1.4874,
+      "step": 11547
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0998537142857142e-05,
+      "loss": 1.1024,
+      "step": 11548
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0998268571428572e-05,
+      "loss": 1.5047,
+      "step": 11549
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0998e-05,
+      "loss": 1.764,
+      "step": 11550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0997731428571429e-05,
+      "loss": 1.3214,
+      "step": 11551
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0997462857142857e-05,
+      "loss": 1.0006,
+      "step": 11552
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0997194285714287e-05,
+      "loss": 1.4396,
+      "step": 11553
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0996925714285715e-05,
+      "loss": 0.9649,
+      "step": 11554
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0996657142857143e-05,
+      "loss": 0.9531,
+      "step": 11555
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0996388571428571e-05,
+      "loss": 0.7722,
+      "step": 11556
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0996120000000001e-05,
+      "loss": 0.9181,
+      "step": 11557
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.099585142857143e-05,
+      "loss": 0.9294,
+      "step": 11558
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0995582857142857e-05,
+      "loss": 1.2762,
+      "step": 11559
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0995314285714287e-05,
+      "loss": 0.8093,
+      "step": 11560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0995045714285714e-05,
+      "loss": 0.906,
+      "step": 11561
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0994777142857144e-05,
+      "loss": 1.0794,
+      "step": 11562
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0994508571428572e-05,
+      "loss": 0.9896,
+      "step": 11563
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.099424e-05,
+      "loss": 1.248,
+      "step": 11564
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0993971428571428e-05,
+      "loss": 1.5772,
+      "step": 11565
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0993702857142858e-05,
+      "loss": 0.9275,
+      "step": 11566
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0993434285714286e-05,
+      "loss": 0.83,
+      "step": 11567
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0993165714285714e-05,
+      "loss": 1.3168,
+      "step": 11568
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0992897142857143e-05,
+      "loss": 1.6512,
+      "step": 11569
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0992628571428572e-05,
+      "loss": 1.4329,
+      "step": 11570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.099236e-05,
+      "loss": 1.0693,
+      "step": 11571
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0992091428571429e-05,
+      "loss": 1.5525,
+      "step": 11572
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0991822857142859e-05,
+      "loss": 1.4082,
+      "step": 11573
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0991554285714285e-05,
+      "loss": 1.1283,
+      "step": 11574
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0991285714285715e-05,
+      "loss": 0.9869,
+      "step": 11575
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0991017142857143e-05,
+      "loss": 1.405,
+      "step": 11576
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0990748571428573e-05,
+      "loss": 1.4215,
+      "step": 11577
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.099048e-05,
+      "loss": 1.3808,
+      "step": 11578
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.099021142857143e-05,
+      "loss": 1.2805,
+      "step": 11579
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0989942857142858e-05,
+      "loss": 1.1009,
+      "step": 11580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0989674285714286e-05,
+      "loss": 1.5431,
+      "step": 11581
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0989405714285714e-05,
+      "loss": 1.1988,
+      "step": 11582
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0989137142857144e-05,
+      "loss": 1.5426,
+      "step": 11583
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0988868571428572e-05,
+      "loss": 1.381,
+      "step": 11584
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.09886e-05,
+      "loss": 1.1963,
+      "step": 11585
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.098833142857143e-05,
+      "loss": 1.0153,
+      "step": 11586
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0988062857142856e-05,
+      "loss": 1.4449,
+      "step": 11587
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0987794285714286e-05,
+      "loss": 1.1082,
+      "step": 11588
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0987525714285715e-05,
+      "loss": 1.5454,
+      "step": 11589
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0987257142857144e-05,
+      "loss": 0.7917,
+      "step": 11590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0986988571428571e-05,
+      "loss": 1.1656,
+      "step": 11591
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.098672e-05,
+      "loss": 1.1501,
+      "step": 11592
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0986451428571429e-05,
+      "loss": 1.2402,
+      "step": 11593
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0986182857142857e-05,
+      "loss": 1.4589,
+      "step": 11594
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0985914285714285e-05,
+      "loss": 1.2802,
+      "step": 11595
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0985645714285715e-05,
+      "loss": 1.4431,
+      "step": 11596
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0985377142857143e-05,
+      "loss": 0.9273,
+      "step": 11597
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0985108571428572e-05,
+      "loss": 1.5407,
+      "step": 11598
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0984840000000001e-05,
+      "loss": 1.2388,
+      "step": 11599
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.098457142857143e-05,
+      "loss": 1.3097,
+      "step": 11600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0984302857142858e-05,
+      "loss": 1.2896,
+      "step": 11601
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0984034285714286e-05,
+      "loss": 1.5996,
+      "step": 11602
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0983765714285716e-05,
+      "loss": 1.3889,
+      "step": 11603
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0983497142857142e-05,
+      "loss": 1.4585,
+      "step": 11604
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0983228571428572e-05,
+      "loss": 1.2563,
+      "step": 11605
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.098296e-05,
+      "loss": 0.8324,
+      "step": 11606
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0982691428571428e-05,
+      "loss": 1.151,
+      "step": 11607
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0982422857142857e-05,
+      "loss": 1.2546,
+      "step": 11608
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0982154285714287e-05,
+      "loss": 0.7183,
+      "step": 11609
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0981885714285715e-05,
+      "loss": 1.4706,
+      "step": 11610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0981617142857143e-05,
+      "loss": 1.4422,
+      "step": 11611
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0981348571428573e-05,
+      "loss": 1.2747,
+      "step": 11612
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0981080000000001e-05,
+      "loss": 0.9948,
+      "step": 11613
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0980811428571429e-05,
+      "loss": 1.4229,
+      "step": 11614
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0980542857142857e-05,
+      "loss": 1.3895,
+      "step": 11615
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0980274285714287e-05,
+      "loss": 0.9581,
+      "step": 11616
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0980005714285714e-05,
+      "loss": 1.0404,
+      "step": 11617
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0979737142857143e-05,
+      "loss": 1.1359,
+      "step": 11618
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0979468571428572e-05,
+      "loss": 1.3967,
+      "step": 11619
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.09792e-05,
+      "loss": 1.2633,
+      "step": 11620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0978931428571428e-05,
+      "loss": 0.9196,
+      "step": 11621
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0978662857142858e-05,
+      "loss": 1.4826,
+      "step": 11622
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0978394285714286e-05,
+      "loss": 1.0556,
+      "step": 11623
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0978125714285714e-05,
+      "loss": 1.27,
+      "step": 11624
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0977857142857144e-05,
+      "loss": 1.6049,
+      "step": 11625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0977588571428572e-05,
+      "loss": 1.231,
+      "step": 11626
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.097732e-05,
+      "loss": 1.1675,
+      "step": 11627
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0977051428571429e-05,
+      "loss": 1.2928,
+      "step": 11628
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0976782857142859e-05,
+      "loss": 1.2227,
+      "step": 11629
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0976514285714285e-05,
+      "loss": 1.0372,
+      "step": 11630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0976245714285715e-05,
+      "loss": 1.2252,
+      "step": 11631
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0975977142857143e-05,
+      "loss": 1.5646,
+      "step": 11632
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0975708571428571e-05,
+      "loss": 1.0902,
+      "step": 11633
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.097544e-05,
+      "loss": 1.5762,
+      "step": 11634
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.097517142857143e-05,
+      "loss": 1.3446,
+      "step": 11635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0974902857142857e-05,
+      "loss": 1.3998,
+      "step": 11636
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.0974634285714286e-05,
+      "loss": 1.0949,
+      "step": 11637
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0974365714285715e-05,
+      "loss": 1.3558,
+      "step": 11638
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0974097142857144e-05,
+      "loss": 1.1711,
+      "step": 11639
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0973828571428572e-05,
+      "loss": 1.2452,
+      "step": 11640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.097356e-05,
+      "loss": 1.2488,
+      "step": 11641
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.097329142857143e-05,
+      "loss": 0.7965,
+      "step": 11642
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0973022857142856e-05,
+      "loss": 1.6297,
+      "step": 11643
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0972754285714286e-05,
+      "loss": 1.6457,
+      "step": 11644
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0972485714285714e-05,
+      "loss": 1.0728,
+      "step": 11645
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0972217142857144e-05,
+      "loss": 1.3566,
+      "step": 11646
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.097194857142857e-05,
+      "loss": 1.0351,
+      "step": 11647
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.097168e-05,
+      "loss": 1.2977,
+      "step": 11648
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0971411428571429e-05,
+      "loss": 1.2152,
+      "step": 11649
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0971142857142857e-05,
+      "loss": 1.3303,
+      "step": 11650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0970874285714287e-05,
+      "loss": 1.0502,
+      "step": 11651
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0970605714285715e-05,
+      "loss": 1.4786,
+      "step": 11652
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0970337142857143e-05,
+      "loss": 1.2072,
+      "step": 11653
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0970068571428571e-05,
+      "loss": 1.555,
+      "step": 11654
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0969800000000001e-05,
+      "loss": 1.622,
+      "step": 11655
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0969531428571428e-05,
+      "loss": 1.1972,
+      "step": 11656
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0969262857142858e-05,
+      "loss": 1.1203,
+      "step": 11657
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0968994285714286e-05,
+      "loss": 1.1749,
+      "step": 11658
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0968725714285716e-05,
+      "loss": 1.4855,
+      "step": 11659
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0968457142857142e-05,
+      "loss": 1.1961,
+      "step": 11660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0968188571428572e-05,
+      "loss": 1.4439,
+      "step": 11661
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.096792e-05,
+      "loss": 0.9272,
+      "step": 11662
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0967651428571428e-05,
+      "loss": 1.2213,
+      "step": 11663
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0967382857142858e-05,
+      "loss": 1.0873,
+      "step": 11664
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0967114285714286e-05,
+      "loss": 1.2606,
+      "step": 11665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0966845714285715e-05,
+      "loss": 1.2411,
+      "step": 11666
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0966577142857143e-05,
+      "loss": 1.3269,
+      "step": 11667
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0966308571428573e-05,
+      "loss": 1.1497,
+      "step": 11668
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.096604e-05,
+      "loss": 1.4732,
+      "step": 11669
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0965771428571429e-05,
+      "loss": 1.1226,
+      "step": 11670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0965502857142857e-05,
+      "loss": 1.5289,
+      "step": 11671
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0965234285714287e-05,
+      "loss": 1.1199,
+      "step": 11672
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0964965714285714e-05,
+      "loss": 1.05,
+      "step": 11673
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0964697142857143e-05,
+      "loss": 1.5918,
+      "step": 11674
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0964428571428572e-05,
+      "loss": 1.4858,
+      "step": 11675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.096416e-05,
+      "loss": 1.4666,
+      "step": 11676
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.096389142857143e-05,
+      "loss": 1.0673,
+      "step": 11677
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0963622857142858e-05,
+      "loss": 1.4988,
+      "step": 11678
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0963354285714286e-05,
+      "loss": 1.3541,
+      "step": 11679
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0963085714285714e-05,
+      "loss": 1.526,
+      "step": 11680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0962817142857144e-05,
+      "loss": 1.663,
+      "step": 11681
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0962548571428572e-05,
+      "loss": 1.2646,
+      "step": 11682
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.096228e-05,
+      "loss": 1.4949,
+      "step": 11683
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0962011428571429e-05,
+      "loss": 1.1797,
+      "step": 11684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0961742857142858e-05,
+      "loss": 1.1404,
+      "step": 11685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0961474285714285e-05,
+      "loss": 0.9511,
+      "step": 11686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0961205714285715e-05,
+      "loss": 0.9577,
+      "step": 11687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0960937142857143e-05,
+      "loss": 1.4665,
+      "step": 11688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0960668571428571e-05,
+      "loss": 1.3509,
+      "step": 11689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0960400000000001e-05,
+      "loss": 0.807,
+      "step": 11690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0960131428571429e-05,
+      "loss": 1.264,
+      "step": 11691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0959862857142859e-05,
+      "loss": 1.0865,
+      "step": 11692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0959594285714286e-05,
+      "loss": 1.1434,
+      "step": 11693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0959325714285715e-05,
+      "loss": 1.1848,
+      "step": 11694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0959057142857144e-05,
+      "loss": 1.3461,
+      "step": 11695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0958788571428572e-05,
+      "loss": 1.0119,
+      "step": 11696
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.095852e-05,
+      "loss": 0.9901,
+      "step": 11697
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.095825142857143e-05,
+      "loss": 1.5591,
+      "step": 11698
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0957982857142856e-05,
+      "loss": 1.2491,
+      "step": 11699
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0957714285714286e-05,
+      "loss": 1.0943,
+      "step": 11700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0957445714285714e-05,
+      "loss": 1.278,
+      "step": 11701
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0957177142857142e-05,
+      "loss": 1.1425,
+      "step": 11702
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0956908571428572e-05,
+      "loss": 1.2309,
+      "step": 11703
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.095664e-05,
+      "loss": 1.3064,
+      "step": 11704
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.095637142857143e-05,
+      "loss": 1.2569,
+      "step": 11705
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0956102857142857e-05,
+      "loss": 1.2025,
+      "step": 11706
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0955834285714287e-05,
+      "loss": 1.4091,
+      "step": 11707
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0955565714285715e-05,
+      "loss": 1.103,
+      "step": 11708
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0955297142857143e-05,
+      "loss": 1.1534,
+      "step": 11709
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0955028571428571e-05,
+      "loss": 1.0979,
+      "step": 11710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0954760000000001e-05,
+      "loss": 1.0095,
+      "step": 11711
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0954491428571428e-05,
+      "loss": 1.5799,
+      "step": 11712
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0954222857142857e-05,
+      "loss": 1.2275,
+      "step": 11713
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0953954285714286e-05,
+      "loss": 1.5477,
+      "step": 11714
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0953685714285714e-05,
+      "loss": 1.5367,
+      "step": 11715
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0953417142857144e-05,
+      "loss": 1.0301,
+      "step": 11716
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0953148571428572e-05,
+      "loss": 0.9446,
+      "step": 11717
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0952880000000002e-05,
+      "loss": 1.4209,
+      "step": 11718
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0952611428571428e-05,
+      "loss": 1.2985,
+      "step": 11719
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0952342857142858e-05,
+      "loss": 0.7884,
+      "step": 11720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0952074285714286e-05,
+      "loss": 1.4887,
+      "step": 11721
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0951805714285714e-05,
+      "loss": 1.1024,
+      "step": 11722
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0951537142857143e-05,
+      "loss": 1.4255,
+      "step": 11723
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0951268571428573e-05,
+      "loss": 0.8348,
+      "step": 11724
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0950999999999999e-05,
+      "loss": 1.3862,
+      "step": 11725
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0950731428571429e-05,
+      "loss": 1.0978,
+      "step": 11726
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0950462857142857e-05,
+      "loss": 1.2217,
+      "step": 11727
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0950194285714287e-05,
+      "loss": 1.3367,
+      "step": 11728
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0949925714285715e-05,
+      "loss": 1.1087,
+      "step": 11729
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0949657142857143e-05,
+      "loss": 1.613,
+      "step": 11730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0949388571428573e-05,
+      "loss": 1.0809,
+      "step": 11731
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.094912e-05,
+      "loss": 1.0236,
+      "step": 11732
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.094885142857143e-05,
+      "loss": 1.2627,
+      "step": 11733
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0948582857142858e-05,
+      "loss": 1.2427,
+      "step": 11734
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0948314285714286e-05,
+      "loss": 0.9647,
+      "step": 11735
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0948045714285714e-05,
+      "loss": 0.7486,
+      "step": 11736
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0947777142857144e-05,
+      "loss": 0.9326,
+      "step": 11737
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.094750857142857e-05,
+      "loss": 1.7051,
+      "step": 11738
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.094724e-05,
+      "loss": 1.5495,
+      "step": 11739
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0946971428571428e-05,
+      "loss": 1.5148,
+      "step": 11740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0946702857142858e-05,
+      "loss": 1.3157,
+      "step": 11741
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0946434285714286e-05,
+      "loss": 1.316,
+      "step": 11742
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0946165714285715e-05,
+      "loss": 0.9875,
+      "step": 11743
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0945897142857145e-05,
+      "loss": 1.6644,
+      "step": 11744
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0945628571428571e-05,
+      "loss": 1.4435,
+      "step": 11745
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0945360000000001e-05,
+      "loss": 1.5211,
+      "step": 11746
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0945091428571429e-05,
+      "loss": 1.2122,
+      "step": 11747
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0944822857142857e-05,
+      "loss": 0.9376,
+      "step": 11748
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0944554285714285e-05,
+      "loss": 0.9436,
+      "step": 11749
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0944285714285715e-05,
+      "loss": 1.3046,
+      "step": 11750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0944017142857143e-05,
+      "loss": 1.0078,
+      "step": 11751
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0943748571428572e-05,
+      "loss": 1.6392,
+      "step": 11752
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.094348e-05,
+      "loss": 1.2163,
+      "step": 11753
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.094321142857143e-05,
+      "loss": 1.3033,
+      "step": 11754
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0942942857142858e-05,
+      "loss": 1.2533,
+      "step": 11755
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0942674285714286e-05,
+      "loss": 1.172,
+      "step": 11756
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0942405714285716e-05,
+      "loss": 0.9207,
+      "step": 11757
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0942137142857142e-05,
+      "loss": 1.0307,
+      "step": 11758
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0941868571428572e-05,
+      "loss": 1.3079,
+      "step": 11759
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.09416e-05,
+      "loss": 1.0467,
+      "step": 11760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0941331428571429e-05,
+      "loss": 1.3977,
+      "step": 11761
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0941062857142857e-05,
+      "loss": 1.421,
+      "step": 11762
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0940794285714287e-05,
+      "loss": 1.4664,
+      "step": 11763
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0940525714285715e-05,
+      "loss": 1.2934,
+      "step": 11764
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0940257142857143e-05,
+      "loss": 1.6331,
+      "step": 11765
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0939988571428571e-05,
+      "loss": 1.0169,
+      "step": 11766
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0939720000000001e-05,
+      "loss": 0.9246,
+      "step": 11767
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.093945142857143e-05,
+      "loss": 0.9787,
+      "step": 11768
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0939182857142857e-05,
+      "loss": 0.9696,
+      "step": 11769
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0938914285714287e-05,
+      "loss": 0.8667,
+      "step": 11770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0938645714285714e-05,
+      "loss": 1.3842,
+      "step": 11771
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0938377142857144e-05,
+      "loss": 1.4744,
+      "step": 11772
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0938108571428572e-05,
+      "loss": 1.1401,
+      "step": 11773
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0937840000000002e-05,
+      "loss": 1.3345,
+      "step": 11774
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0937571428571428e-05,
+      "loss": 1.3458,
+      "step": 11775
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0937302857142858e-05,
+      "loss": 1.0111,
+      "step": 11776
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0937034285714286e-05,
+      "loss": 1.4639,
+      "step": 11777
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0936765714285714e-05,
+      "loss": 1.4153,
+      "step": 11778
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0936497142857143e-05,
+      "loss": 1.2992,
+      "step": 11779
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0936228571428572e-05,
+      "loss": 1.1966,
+      "step": 11780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.093596e-05,
+      "loss": 1.2361,
+      "step": 11781
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0935691428571429e-05,
+      "loss": 1.2177,
+      "step": 11782
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0935422857142859e-05,
+      "loss": 1.0917,
+      "step": 11783
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0935154285714285e-05,
+      "loss": 1.5704,
+      "step": 11784
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0934885714285715e-05,
+      "loss": 1.231,
+      "step": 11785
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0934617142857143e-05,
+      "loss": 1.5548,
+      "step": 11786
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0934348571428573e-05,
+      "loss": 1.4008,
+      "step": 11787
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.093408e-05,
+      "loss": 1.09,
+      "step": 11788
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.093381142857143e-05,
+      "loss": 1.4853,
+      "step": 11789
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0933542857142858e-05,
+      "loss": 1.1967,
+      "step": 11790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0933274285714286e-05,
+      "loss": 0.8781,
+      "step": 11791
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0933005714285714e-05,
+      "loss": 1.3389,
+      "step": 11792
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0932737142857144e-05,
+      "loss": 0.9647,
+      "step": 11793
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0932468571428572e-05,
+      "loss": 1.4151,
+      "step": 11794
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.09322e-05,
+      "loss": 1.0504,
+      "step": 11795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.093193142857143e-05,
+      "loss": 0.8923,
+      "step": 11796
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0931662857142858e-05,
+      "loss": 1.0647,
+      "step": 11797
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0931394285714286e-05,
+      "loss": 1.3652,
+      "step": 11798
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0931125714285715e-05,
+      "loss": 1.5039,
+      "step": 11799
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0930857142857144e-05,
+      "loss": 1.0015,
+      "step": 11800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0930588571428571e-05,
+      "loss": 1.1839,
+      "step": 11801
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.093032e-05,
+      "loss": 1.2865,
+      "step": 11802
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0930051428571429e-05,
+      "loss": 0.9557,
+      "step": 11803
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0929782857142857e-05,
+      "loss": 1.1443,
+      "step": 11804
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0929514285714285e-05,
+      "loss": 1.5717,
+      "step": 11805
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0929245714285715e-05,
+      "loss": 1.5194,
+      "step": 11806
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0928977142857143e-05,
+      "loss": 1.3592,
+      "step": 11807
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0928708571428571e-05,
+      "loss": 0.8926,
+      "step": 11808
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0928440000000001e-05,
+      "loss": 1.4255,
+      "step": 11809
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.092817142857143e-05,
+      "loss": 1.268,
+      "step": 11810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0927902857142858e-05,
+      "loss": 0.785,
+      "step": 11811
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.0927634285714286e-05,
+      "loss": 1.4137,
+      "step": 11812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0927365714285716e-05,
+      "loss": 1.4241,
+      "step": 11813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0927097142857142e-05,
+      "loss": 1.0623,
+      "step": 11814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0926828571428572e-05,
+      "loss": 1.4667,
+      "step": 11815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.092656e-05,
+      "loss": 1.341,
+      "step": 11816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0926291428571428e-05,
+      "loss": 1.2167,
+      "step": 11817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0926022857142857e-05,
+      "loss": 1.03,
+      "step": 11818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0925754285714287e-05,
+      "loss": 1.555,
+      "step": 11819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0925485714285715e-05,
+      "loss": 1.1423,
+      "step": 11820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0925217142857143e-05,
+      "loss": 1.4808,
+      "step": 11821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0924948571428573e-05,
+      "loss": 0.9893,
+      "step": 11822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0924680000000001e-05,
+      "loss": 1.1234,
+      "step": 11823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0924411428571429e-05,
+      "loss": 1.5377,
+      "step": 11824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0924142857142857e-05,
+      "loss": 1.3165,
+      "step": 11825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0923874285714287e-05,
+      "loss": 1.539,
+      "step": 11826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0923605714285714e-05,
+      "loss": 1.1022,
+      "step": 11827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0923337142857143e-05,
+      "loss": 1.2093,
+      "step": 11828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0923068571428572e-05,
+      "loss": 1.074,
+      "step": 11829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.09228e-05,
+      "loss": 0.7385,
+      "step": 11830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0922531428571428e-05,
+      "loss": 1.6382,
+      "step": 11831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0922262857142858e-05,
+      "loss": 1.0898,
+      "step": 11832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0921994285714286e-05,
+      "loss": 0.6234,
+      "step": 11833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0921725714285714e-05,
+      "loss": 1.0455,
+      "step": 11834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0921457142857144e-05,
+      "loss": 1.3769,
+      "step": 11835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0921188571428572e-05,
+      "loss": 1.1497,
+      "step": 11836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.092092e-05,
+      "loss": 1.5953,
+      "step": 11837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0920651428571429e-05,
+      "loss": 1.4183,
+      "step": 11838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0920382857142859e-05,
+      "loss": 1.5308,
+      "step": 11839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0920114285714285e-05,
+      "loss": 1.4105,
+      "step": 11840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0919845714285715e-05,
+      "loss": 1.5073,
+      "step": 11841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0919577142857143e-05,
+      "loss": 1.1547,
+      "step": 11842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0919308571428573e-05,
+      "loss": 1.1071,
+      "step": 11843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091904e-05,
+      "loss": 1.0872,
+      "step": 11844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091877142857143e-05,
+      "loss": 1.4375,
+      "step": 11845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0918502857142857e-05,
+      "loss": 1.0007,
+      "step": 11846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0918234285714286e-05,
+      "loss": 1.3372,
+      "step": 11847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0917965714285715e-05,
+      "loss": 1.6852,
+      "step": 11848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0917697142857144e-05,
+      "loss": 1.1575,
+      "step": 11849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0917428571428572e-05,
+      "loss": 1.7481,
+      "step": 11850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091716e-05,
+      "loss": 1.4586,
+      "step": 11851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091689142857143e-05,
+      "loss": 0.8363,
+      "step": 11852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0916622857142856e-05,
+      "loss": 1.2871,
+      "step": 11853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0916354285714286e-05,
+      "loss": 1.1856,
+      "step": 11854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0916085714285714e-05,
+      "loss": 1.3049,
+      "step": 11855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0915817142857144e-05,
+      "loss": 1.0292,
+      "step": 11856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091554857142857e-05,
+      "loss": 1.3117,
+      "step": 11857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091528e-05,
+      "loss": 1.4731,
+      "step": 11858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0915011428571429e-05,
+      "loss": 1.2987,
+      "step": 11859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0914742857142857e-05,
+      "loss": 0.7956,
+      "step": 11860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0914474285714287e-05,
+      "loss": 1.0986,
+      "step": 11861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0914205714285715e-05,
+      "loss": 1.2107,
+      "step": 11862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0913937142857143e-05,
+      "loss": 1.0679,
+      "step": 11863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0913668571428571e-05,
+      "loss": 1.2217,
+      "step": 11864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0913400000000001e-05,
+      "loss": 1.1269,
+      "step": 11865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091313142857143e-05,
+      "loss": 1.4273,
+      "step": 11866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0912862857142858e-05,
+      "loss": 1.246,
+      "step": 11867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0912594285714286e-05,
+      "loss": 1.105,
+      "step": 11868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0912325714285716e-05,
+      "loss": 1.1717,
+      "step": 11869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0912057142857142e-05,
+      "loss": 1.4415,
+      "step": 11870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0911788571428572e-05,
+      "loss": 1.1598,
+      "step": 11871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.091152e-05,
+      "loss": 1.2473,
+      "step": 11872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0911251428571428e-05,
+      "loss": 1.2544,
+      "step": 11873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0910982857142858e-05,
+      "loss": 1.1303,
+      "step": 11874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0910714285714286e-05,
+      "loss": 1.0504,
+      "step": 11875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0910445714285715e-05,
+      "loss": 1.4236,
+      "step": 11876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0910177142857143e-05,
+      "loss": 1.1129,
+      "step": 11877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0909908571428573e-05,
+      "loss": 1.1064,
+      "step": 11878
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.090964e-05,
+      "loss": 1.1339,
+      "step": 11879
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0909371428571429e-05,
+      "loss": 1.0624,
+      "step": 11880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0909102857142857e-05,
+      "loss": 1.7646,
+      "step": 11881
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0908834285714287e-05,
+      "loss": 1.4421,
+      "step": 11882
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0908565714285714e-05,
+      "loss": 1.266,
+      "step": 11883
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0908297142857143e-05,
+      "loss": 1.4178,
+      "step": 11884
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0908028571428572e-05,
+      "loss": 1.2008,
+      "step": 11885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.090776e-05,
+      "loss": 1.366,
+      "step": 11886
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.090749142857143e-05,
+      "loss": 1.4068,
+      "step": 11887
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0907222857142858e-05,
+      "loss": 0.9546,
+      "step": 11888
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0906954285714286e-05,
+      "loss": 0.8105,
+      "step": 11889
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0906685714285714e-05,
+      "loss": 1.1198,
+      "step": 11890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0906417142857144e-05,
+      "loss": 1.0791,
+      "step": 11891
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0906148571428572e-05,
+      "loss": 1.7009,
+      "step": 11892
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.090588e-05,
+      "loss": 1.3627,
+      "step": 11893
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0905611428571429e-05,
+      "loss": 1.2747,
+      "step": 11894
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0905342857142858e-05,
+      "loss": 1.5232,
+      "step": 11895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0905074285714285e-05,
+      "loss": 1.1469,
+      "step": 11896
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0904805714285715e-05,
+      "loss": 1.341,
+      "step": 11897
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0904537142857143e-05,
+      "loss": 1.5496,
+      "step": 11898
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0904268571428571e-05,
+      "loss": 1.4881,
+      "step": 11899
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0904000000000001e-05,
+      "loss": 1.1784,
+      "step": 11900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0903731428571429e-05,
+      "loss": 0.9159,
+      "step": 11901
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0903462857142859e-05,
+      "loss": 1.0071,
+      "step": 11902
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0903194285714286e-05,
+      "loss": 1.2517,
+      "step": 11903
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0902925714285715e-05,
+      "loss": 1.0549,
+      "step": 11904
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0902657142857144e-05,
+      "loss": 1.0787,
+      "step": 11905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0902388571428572e-05,
+      "loss": 1.2014,
+      "step": 11906
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.090212e-05,
+      "loss": 1.18,
+      "step": 11907
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.090185142857143e-05,
+      "loss": 1.2804,
+      "step": 11908
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0901582857142856e-05,
+      "loss": 1.049,
+      "step": 11909
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0901314285714286e-05,
+      "loss": 1.5316,
+      "step": 11910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0901045714285714e-05,
+      "loss": 1.1461,
+      "step": 11911
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0900777142857142e-05,
+      "loss": 1.3134,
+      "step": 11912
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0900508571428572e-05,
+      "loss": 1.236,
+      "step": 11913
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.090024e-05,
+      "loss": 1.0907,
+      "step": 11914
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.089997142857143e-05,
+      "loss": 1.0938,
+      "step": 11915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0899702857142857e-05,
+      "loss": 1.412,
+      "step": 11916
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0899434285714287e-05,
+      "loss": 1.0913,
+      "step": 11917
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0899165714285715e-05,
+      "loss": 1.1641,
+      "step": 11918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0898897142857143e-05,
+      "loss": 1.8693,
+      "step": 11919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0898628571428571e-05,
+      "loss": 1.6631,
+      "step": 11920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0898360000000001e-05,
+      "loss": 1.3844,
+      "step": 11921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0898091428571428e-05,
+      "loss": 1.3357,
+      "step": 11922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0897822857142857e-05,
+      "loss": 1.2014,
+      "step": 11923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0897554285714286e-05,
+      "loss": 1.4922,
+      "step": 11924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0897285714285716e-05,
+      "loss": 1.201,
+      "step": 11925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0897017142857144e-05,
+      "loss": 1.2499,
+      "step": 11926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0896748571428572e-05,
+      "loss": 1.289,
+      "step": 11927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0896480000000002e-05,
+      "loss": 0.9659,
+      "step": 11928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0896211428571428e-05,
+      "loss": 1.1539,
+      "step": 11929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0895942857142858e-05,
+      "loss": 0.8684,
+      "step": 11930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0895674285714286e-05,
+      "loss": 1.1962,
+      "step": 11931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0895405714285714e-05,
+      "loss": 1.4426,
+      "step": 11932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0895137142857143e-05,
+      "loss": 1.0568,
+      "step": 11933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0894868571428573e-05,
+      "loss": 1.2838,
+      "step": 11934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0894599999999999e-05,
+      "loss": 1.2159,
+      "step": 11935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0894331428571429e-05,
+      "loss": 1.3368,
+      "step": 11936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0894062857142857e-05,
+      "loss": 1.1361,
+      "step": 11937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0893794285714287e-05,
+      "loss": 1.1166,
+      "step": 11938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0893525714285715e-05,
+      "loss": 1.2743,
+      "step": 11939
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0893257142857143e-05,
+      "loss": 0.83,
+      "step": 11940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0892988571428573e-05,
+      "loss": 0.8078,
+      "step": 11941
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.089272e-05,
+      "loss": 1.4541,
+      "step": 11942
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.089245142857143e-05,
+      "loss": 1.0897,
+      "step": 11943
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0892182857142858e-05,
+      "loss": 1.1775,
+      "step": 11944
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0891914285714286e-05,
+      "loss": 1.1976,
+      "step": 11945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0891645714285714e-05,
+      "loss": 1.5076,
+      "step": 11946
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0891377142857144e-05,
+      "loss": 0.8928,
+      "step": 11947
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0891108571428572e-05,
+      "loss": 1.3805,
+      "step": 11948
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.089084e-05,
+      "loss": 1.3443,
+      "step": 11949
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0890571428571428e-05,
+      "loss": 0.8877,
+      "step": 11950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0890302857142858e-05,
+      "loss": 1.4352,
+      "step": 11951
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0890034285714286e-05,
+      "loss": 1.0363,
+      "step": 11952
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0889765714285715e-05,
+      "loss": 0.8859,
+      "step": 11953
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0889497142857145e-05,
+      "loss": 0.9125,
+      "step": 11954
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0889228571428571e-05,
+      "loss": 1.493,
+      "step": 11955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0888960000000001e-05,
+      "loss": 1.3763,
+      "step": 11956
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0888691428571429e-05,
+      "loss": 1.1457,
+      "step": 11957
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0888422857142857e-05,
+      "loss": 1.3678,
+      "step": 11958
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0888154285714285e-05,
+      "loss": 1.6519,
+      "step": 11959
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0887885714285715e-05,
+      "loss": 0.8049,
+      "step": 11960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0887617142857143e-05,
+      "loss": 1.4146,
+      "step": 11961
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0887348571428572e-05,
+      "loss": 1.081,
+      "step": 11962
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.088708e-05,
+      "loss": 1.1359,
+      "step": 11963
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.088681142857143e-05,
+      "loss": 1.2616,
+      "step": 11964
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0886542857142858e-05,
+      "loss": 1.323,
+      "step": 11965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0886274285714286e-05,
+      "loss": 0.9546,
+      "step": 11966
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0886005714285716e-05,
+      "loss": 1.325,
+      "step": 11967
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0885737142857142e-05,
+      "loss": 1.2978,
+      "step": 11968
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0885468571428572e-05,
+      "loss": 0.7711,
+      "step": 11969
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.08852e-05,
+      "loss": 1.1575,
+      "step": 11970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.088493142857143e-05,
+      "loss": 0.9068,
+      "step": 11971
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0884662857142857e-05,
+      "loss": 1.699,
+      "step": 11972
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0884394285714287e-05,
+      "loss": 1.642,
+      "step": 11973
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0884125714285715e-05,
+      "loss": 1.0098,
+      "step": 11974
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0883857142857143e-05,
+      "loss": 1.1798,
+      "step": 11975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0883588571428571e-05,
+      "loss": 1.1507,
+      "step": 11976
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0883320000000001e-05,
+      "loss": 0.9917,
+      "step": 11977
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.088305142857143e-05,
+      "loss": 0.9946,
+      "step": 11978
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0882782857142857e-05,
+      "loss": 1.1136,
+      "step": 11979
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0882514285714287e-05,
+      "loss": 1.1011,
+      "step": 11980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0882245714285714e-05,
+      "loss": 1.1506,
+      "step": 11981
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0881977142857144e-05,
+      "loss": 1.1167,
+      "step": 11982
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0881708571428572e-05,
+      "loss": 1.0255,
+      "step": 11983
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0881440000000002e-05,
+      "loss": 1.3331,
+      "step": 11984
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0881171428571428e-05,
+      "loss": 1.3838,
+      "step": 11985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0880902857142858e-05,
+      "loss": 0.9706,
+      "step": 11986
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.0880634285714286e-05,
+      "loss": 1.5595,
+      "step": 11987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0880365714285714e-05,
+      "loss": 1.2013,
+      "step": 11988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0880097142857143e-05,
+      "loss": 1.2712,
+      "step": 11989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0879828571428572e-05,
+      "loss": 1.272,
+      "step": 11990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.087956e-05,
+      "loss": 1.1521,
+      "step": 11991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0879291428571429e-05,
+      "loss": 1.1005,
+      "step": 11992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0879022857142859e-05,
+      "loss": 1.1084,
+      "step": 11993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0878754285714287e-05,
+      "loss": 1.1503,
+      "step": 11994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0878485714285715e-05,
+      "loss": 1.5234,
+      "step": 11995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0878217142857143e-05,
+      "loss": 1.4847,
+      "step": 11996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0877948571428573e-05,
+      "loss": 1.2066,
+      "step": 11997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.087768e-05,
+      "loss": 1.3184,
+      "step": 11998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.087741142857143e-05,
+      "loss": 1.616,
+      "step": 11999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0877142857142858e-05,
+      "loss": 1.3148,
+      "step": 12000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0876874285714286e-05,
+      "loss": 1.5629,
+      "step": 12001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0876605714285714e-05,
+      "loss": 1.4105,
+      "step": 12002
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0876337142857144e-05,
+      "loss": 1.2216,
+      "step": 12003
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0876068571428572e-05,
+      "loss": 1.0556,
+      "step": 12004
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.08758e-05,
+      "loss": 1.5809,
+      "step": 12005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.087553142857143e-05,
+      "loss": 1.2435,
+      "step": 12006
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0875262857142858e-05,
+      "loss": 1.4801,
+      "step": 12007
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0874994285714286e-05,
+      "loss": 0.8953,
+      "step": 12008
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0874725714285715e-05,
+      "loss": 1.6309,
+      "step": 12009
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0874457142857144e-05,
+      "loss": 1.6009,
+      "step": 12010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0874188571428571e-05,
+      "loss": 1.2551,
+      "step": 12011
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.087392e-05,
+      "loss": 1.1571,
+      "step": 12012
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0873651428571429e-05,
+      "loss": 1.0957,
+      "step": 12013
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0873382857142857e-05,
+      "loss": 1.3315,
+      "step": 12014
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0873114285714285e-05,
+      "loss": 1.4394,
+      "step": 12015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0872845714285715e-05,
+      "loss": 1.586,
+      "step": 12016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0872577142857143e-05,
+      "loss": 1.5868,
+      "step": 12017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0872308571428571e-05,
+      "loss": 1.2761,
+      "step": 12018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0872040000000001e-05,
+      "loss": 1.1247,
+      "step": 12019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.087177142857143e-05,
+      "loss": 1.0755,
+      "step": 12020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0871502857142858e-05,
+      "loss": 1.9985,
+      "step": 12021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0871234285714286e-05,
+      "loss": 1.3682,
+      "step": 12022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0870965714285716e-05,
+      "loss": 1.5016,
+      "step": 12023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0870697142857142e-05,
+      "loss": 0.9108,
+      "step": 12024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0870428571428572e-05,
+      "loss": 1.3631,
+      "step": 12025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.087016e-05,
+      "loss": 1.1028,
+      "step": 12026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0869891428571428e-05,
+      "loss": 1.0164,
+      "step": 12027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0869622857142857e-05,
+      "loss": 1.4133,
+      "step": 12028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0869354285714287e-05,
+      "loss": 1.2878,
+      "step": 12029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0869085714285715e-05,
+      "loss": 1.3126,
+      "step": 12030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0868817142857143e-05,
+      "loss": 1.2547,
+      "step": 12031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0868548571428573e-05,
+      "loss": 1.4141,
+      "step": 12032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0868280000000001e-05,
+      "loss": 1.0218,
+      "step": 12033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0868011428571429e-05,
+      "loss": 0.9641,
+      "step": 12034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0867742857142857e-05,
+      "loss": 0.9592,
+      "step": 12035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0867474285714287e-05,
+      "loss": 1.2838,
+      "step": 12036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0867205714285714e-05,
+      "loss": 1.2,
+      "step": 12037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0866937142857143e-05,
+      "loss": 1.2131,
+      "step": 12038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0866668571428572e-05,
+      "loss": 1.0545,
+      "step": 12039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0866400000000002e-05,
+      "loss": 1.129,
+      "step": 12040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0866131428571428e-05,
+      "loss": 1.194,
+      "step": 12041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0865862857142858e-05,
+      "loss": 1.3907,
+      "step": 12042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0865594285714286e-05,
+      "loss": 1.0925,
+      "step": 12043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0865325714285714e-05,
+      "loss": 0.9922,
+      "step": 12044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0865057142857144e-05,
+      "loss": 1.0456,
+      "step": 12045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0864788571428572e-05,
+      "loss": 1.2942,
+      "step": 12046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.086452e-05,
+      "loss": 1.1914,
+      "step": 12047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0864251428571429e-05,
+      "loss": 1.3409,
+      "step": 12048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0863982857142859e-05,
+      "loss": 0.8255,
+      "step": 12049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0863714285714285e-05,
+      "loss": 1.2045,
+      "step": 12050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0863445714285715e-05,
+      "loss": 1.4951,
+      "step": 12051
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0863177142857143e-05,
+      "loss": 1.32,
+      "step": 12052
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0862908571428573e-05,
+      "loss": 1.416,
+      "step": 12053
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.086264e-05,
+      "loss": 1.5465,
+      "step": 12054
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.086237142857143e-05,
+      "loss": 1.4416,
+      "step": 12055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0862102857142857e-05,
+      "loss": 1.3145,
+      "step": 12056
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0861834285714286e-05,
+      "loss": 1.1836,
+      "step": 12057
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0861565714285715e-05,
+      "loss": 1.6195,
+      "step": 12058
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0861297142857144e-05,
+      "loss": 1.3573,
+      "step": 12059
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0861028571428572e-05,
+      "loss": 1.687,
+      "step": 12060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.086076e-05,
+      "loss": 0.8522,
+      "step": 12061
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.086049142857143e-05,
+      "loss": 1.3905,
+      "step": 12062
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0860222857142856e-05,
+      "loss": 1.0268,
+      "step": 12063
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0859954285714286e-05,
+      "loss": 0.6698,
+      "step": 12064
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0859685714285714e-05,
+      "loss": 1.5026,
+      "step": 12065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0859417142857144e-05,
+      "loss": 0.9195,
+      "step": 12066
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.085914857142857e-05,
+      "loss": 0.9632,
+      "step": 12067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.085888e-05,
+      "loss": 1.6847,
+      "step": 12068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0858611428571429e-05,
+      "loss": 1.2715,
+      "step": 12069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0858342857142857e-05,
+      "loss": 1.5469,
+      "step": 12070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0858074285714287e-05,
+      "loss": 1.4398,
+      "step": 12071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0857805714285715e-05,
+      "loss": 1.3079,
+      "step": 12072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0857537142857143e-05,
+      "loss": 0.8919,
+      "step": 12073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0857268571428571e-05,
+      "loss": 1.1729,
+      "step": 12074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0857000000000001e-05,
+      "loss": 1.4058,
+      "step": 12075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.085673142857143e-05,
+      "loss": 1.2788,
+      "step": 12076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0856462857142858e-05,
+      "loss": 1.2259,
+      "step": 12077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0856194285714286e-05,
+      "loss": 1.7022,
+      "step": 12078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0855925714285716e-05,
+      "loss": 1.3546,
+      "step": 12079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0855657142857142e-05,
+      "loss": 0.9872,
+      "step": 12080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0855388571428572e-05,
+      "loss": 1.6576,
+      "step": 12081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.085512e-05,
+      "loss": 1.2105,
+      "step": 12082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0854851428571428e-05,
+      "loss": 1.4111,
+      "step": 12083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0854582857142858e-05,
+      "loss": 1.2789,
+      "step": 12084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0854314285714286e-05,
+      "loss": 1.023,
+      "step": 12085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0854045714285715e-05,
+      "loss": 0.9582,
+      "step": 12086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0853777142857143e-05,
+      "loss": 1.3637,
+      "step": 12087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0853508571428573e-05,
+      "loss": 1.3187,
+      "step": 12088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.085324e-05,
+      "loss": 1.0212,
+      "step": 12089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0852971428571429e-05,
+      "loss": 1.4889,
+      "step": 12090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0852702857142857e-05,
+      "loss": 1.4888,
+      "step": 12091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0852434285714287e-05,
+      "loss": 1.6613,
+      "step": 12092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0852165714285714e-05,
+      "loss": 0.9139,
+      "step": 12093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0851897142857143e-05,
+      "loss": 1.3978,
+      "step": 12094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0851628571428572e-05,
+      "loss": 1.4528,
+      "step": 12095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.085136e-05,
+      "loss": 0.874,
+      "step": 12096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.085109142857143e-05,
+      "loss": 1.4762,
+      "step": 12097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0850822857142858e-05,
+      "loss": 1.5085,
+      "step": 12098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0850554285714288e-05,
+      "loss": 1.2608,
+      "step": 12099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0850285714285714e-05,
+      "loss": 1.2807,
+      "step": 12100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0850017142857144e-05,
+      "loss": 1.4219,
+      "step": 12101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0849748571428572e-05,
+      "loss": 1.1991,
+      "step": 12102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.084948e-05,
+      "loss": 1.0215,
+      "step": 12103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0849211428571429e-05,
+      "loss": 1.786,
+      "step": 12104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0848942857142858e-05,
+      "loss": 1.4157,
+      "step": 12105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0848674285714285e-05,
+      "loss": 1.4414,
+      "step": 12106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0848405714285715e-05,
+      "loss": 1.0324,
+      "step": 12107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0848137142857143e-05,
+      "loss": 1.4952,
+      "step": 12108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0847868571428571e-05,
+      "loss": 1.1564,
+      "step": 12109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0847600000000001e-05,
+      "loss": 1.3097,
+      "step": 12110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0847331428571429e-05,
+      "loss": 0.8303,
+      "step": 12111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0847062857142859e-05,
+      "loss": 1.4706,
+      "step": 12112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0846794285714285e-05,
+      "loss": 1.4338,
+      "step": 12113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0846525714285715e-05,
+      "loss": 1.124,
+      "step": 12114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0846257142857144e-05,
+      "loss": 1.378,
+      "step": 12115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0845988571428572e-05,
+      "loss": 1.1496,
+      "step": 12116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.084572e-05,
+      "loss": 1.2422,
+      "step": 12117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.084545142857143e-05,
+      "loss": 1.2103,
+      "step": 12118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0845182857142856e-05,
+      "loss": 1.4664,
+      "step": 12119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0844914285714286e-05,
+      "loss": 0.7475,
+      "step": 12120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0844645714285714e-05,
+      "loss": 1.129,
+      "step": 12121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0844377142857144e-05,
+      "loss": 1.3032,
+      "step": 12122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.084410857142857e-05,
+      "loss": 1.2715,
+      "step": 12123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.084384e-05,
+      "loss": 0.9455,
+      "step": 12124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.084357142857143e-05,
+      "loss": 1.2821,
+      "step": 12125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0843302857142857e-05,
+      "loss": 0.9937,
+      "step": 12126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0843034285714287e-05,
+      "loss": 0.7338,
+      "step": 12127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0842765714285715e-05,
+      "loss": 1.0681,
+      "step": 12128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0842497142857143e-05,
+      "loss": 1.0582,
+      "step": 12129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0842228571428571e-05,
+      "loss": 1.0435,
+      "step": 12130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0841960000000001e-05,
+      "loss": 1.1354,
+      "step": 12131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0841691428571428e-05,
+      "loss": 0.9481,
+      "step": 12132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0841422857142857e-05,
+      "loss": 0.9391,
+      "step": 12133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0841154285714286e-05,
+      "loss": 1.3199,
+      "step": 12134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0840885714285716e-05,
+      "loss": 1.2079,
+      "step": 12135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0840617142857142e-05,
+      "loss": 1.1842,
+      "step": 12136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0840348571428572e-05,
+      "loss": 1.0914,
+      "step": 12137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0840080000000002e-05,
+      "loss": 1.3717,
+      "step": 12138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0839811428571428e-05,
+      "loss": 1.1733,
+      "step": 12139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0839542857142858e-05,
+      "loss": 1.6478,
+      "step": 12140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0839274285714286e-05,
+      "loss": 1.3651,
+      "step": 12141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0839005714285714e-05,
+      "loss": 1.2261,
+      "step": 12142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0838737142857143e-05,
+      "loss": 1.1176,
+      "step": 12143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0838468571428573e-05,
+      "loss": 1.0151,
+      "step": 12144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.08382e-05,
+      "loss": 1.8009,
+      "step": 12145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0837931428571429e-05,
+      "loss": 1.4162,
+      "step": 12146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0837662857142857e-05,
+      "loss": 0.869,
+      "step": 12147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0837394285714287e-05,
+      "loss": 1.4844,
+      "step": 12148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0837125714285713e-05,
+      "loss": 1.2762,
+      "step": 12149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0836857142857143e-05,
+      "loss": 1.3956,
+      "step": 12150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0836588571428573e-05,
+      "loss": 0.9271,
+      "step": 12151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.083632e-05,
+      "loss": 1.188,
+      "step": 12152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.083605142857143e-05,
+      "loss": 1.0744,
+      "step": 12153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0835782857142858e-05,
+      "loss": 1.183,
+      "step": 12154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0835514285714286e-05,
+      "loss": 1.3992,
+      "step": 12155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0835245714285714e-05,
+      "loss": 1.2931,
+      "step": 12156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0834977142857144e-05,
+      "loss": 1.0768,
+      "step": 12157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0834708571428572e-05,
+      "loss": 1.0473,
+      "step": 12158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.083444e-05,
+      "loss": 1.6229,
+      "step": 12159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0834171428571428e-05,
+      "loss": 1.1872,
+      "step": 12160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0833902857142858e-05,
+      "loss": 1.4868,
+      "step": 12161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.0833634285714285e-05,
+      "loss": 1.4433,
+      "step": 12162
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0833365714285715e-05,
+      "loss": 1.5557,
+      "step": 12163
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0833097142857145e-05,
+      "loss": 1.4014,
+      "step": 12164
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0832828571428571e-05,
+      "loss": 1.2527,
+      "step": 12165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0832560000000001e-05,
+      "loss": 1.0933,
+      "step": 12166
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0832291428571429e-05,
+      "loss": 1.1375,
+      "step": 12167
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0832022857142859e-05,
+      "loss": 0.8436,
+      "step": 12168
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0831754285714285e-05,
+      "loss": 0.9368,
+      "step": 12169
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0831485714285715e-05,
+      "loss": 1.2795,
+      "step": 12170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0831217142857143e-05,
+      "loss": 1.1183,
+      "step": 12171
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0830948571428572e-05,
+      "loss": 1.0326,
+      "step": 12172
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.083068e-05,
+      "loss": 1.4458,
+      "step": 12173
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.083041142857143e-05,
+      "loss": 1.3537,
+      "step": 12174
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0830142857142856e-05,
+      "loss": 1.7149,
+      "step": 12175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0829874285714286e-05,
+      "loss": 1.6058,
+      "step": 12176
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0829605714285716e-05,
+      "loss": 1.2114,
+      "step": 12177
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0829337142857142e-05,
+      "loss": 1.3555,
+      "step": 12178
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0829068571428572e-05,
+      "loss": 1.3717,
+      "step": 12179
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.08288e-05,
+      "loss": 0.9382,
+      "step": 12180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.082853142857143e-05,
+      "loss": 1.1944,
+      "step": 12181
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0828262857142857e-05,
+      "loss": 1.1212,
+      "step": 12182
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0827994285714287e-05,
+      "loss": 1.4971,
+      "step": 12183
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0827725714285715e-05,
+      "loss": 1.106,
+      "step": 12184
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0827457142857143e-05,
+      "loss": 1.0692,
+      "step": 12185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0827188571428571e-05,
+      "loss": 1.3683,
+      "step": 12186
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0826920000000001e-05,
+      "loss": 1.0583,
+      "step": 12187
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0826651428571428e-05,
+      "loss": 0.9279,
+      "step": 12188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0826382857142857e-05,
+      "loss": 1.506,
+      "step": 12189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0826114285714287e-05,
+      "loss": 1.0693,
+      "step": 12190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0825845714285715e-05,
+      "loss": 1.3792,
+      "step": 12191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0825577142857144e-05,
+      "loss": 0.9437,
+      "step": 12192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0825308571428572e-05,
+      "loss": 1.2155,
+      "step": 12193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0825040000000002e-05,
+      "loss": 1.5951,
+      "step": 12194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0824771428571428e-05,
+      "loss": 0.8746,
+      "step": 12195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0824502857142858e-05,
+      "loss": 1.368,
+      "step": 12196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0824234285714286e-05,
+      "loss": 1.5457,
+      "step": 12197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0823965714285714e-05,
+      "loss": 1.1713,
+      "step": 12198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0823697142857143e-05,
+      "loss": 1.2759,
+      "step": 12199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0823428571428572e-05,
+      "loss": 1.1932,
+      "step": 12200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0823159999999999e-05,
+      "loss": 1.1989,
+      "step": 12201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0822891428571429e-05,
+      "loss": 1.3061,
+      "step": 12202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0822622857142859e-05,
+      "loss": 1.2288,
+      "step": 12203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0822354285714287e-05,
+      "loss": 1.7438,
+      "step": 12204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0822085714285715e-05,
+      "loss": 1.2385,
+      "step": 12205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0821817142857143e-05,
+      "loss": 1.3256,
+      "step": 12206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0821548571428573e-05,
+      "loss": 0.7684,
+      "step": 12207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.082128e-05,
+      "loss": 1.2639,
+      "step": 12208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.082101142857143e-05,
+      "loss": 1.2872,
+      "step": 12209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0820742857142858e-05,
+      "loss": 0.9209,
+      "step": 12210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0820474285714286e-05,
+      "loss": 0.8888,
+      "step": 12211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0820205714285714e-05,
+      "loss": 1.1708,
+      "step": 12212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0819937142857144e-05,
+      "loss": 1.0129,
+      "step": 12213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.081966857142857e-05,
+      "loss": 1.4566,
+      "step": 12214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.08194e-05,
+      "loss": 1.2787,
+      "step": 12215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.081913142857143e-05,
+      "loss": 1.2675,
+      "step": 12216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0818862857142858e-05,
+      "loss": 1.1905,
+      "step": 12217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0818594285714286e-05,
+      "loss": 1.5539,
+      "step": 12218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0818325714285715e-05,
+      "loss": 1.1942,
+      "step": 12219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0818057142857144e-05,
+      "loss": 0.9254,
+      "step": 12220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0817788571428571e-05,
+      "loss": 1.3144,
+      "step": 12221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.081752e-05,
+      "loss": 0.9138,
+      "step": 12222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0817251428571429e-05,
+      "loss": 1.1762,
+      "step": 12223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0816982857142857e-05,
+      "loss": 1.1828,
+      "step": 12224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0816714285714285e-05,
+      "loss": 1.3333,
+      "step": 12225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0816445714285715e-05,
+      "loss": 1.0787,
+      "step": 12226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0816177142857143e-05,
+      "loss": 0.6459,
+      "step": 12227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0815908571428571e-05,
+      "loss": 1.5064,
+      "step": 12228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.081564e-05,
+      "loss": 1.3176,
+      "step": 12229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.081537142857143e-05,
+      "loss": 0.8054,
+      "step": 12230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0815102857142858e-05,
+      "loss": 1.0227,
+      "step": 12231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0814834285714286e-05,
+      "loss": 1.1805,
+      "step": 12232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0814565714285716e-05,
+      "loss": 1.5672,
+      "step": 12233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0814297142857142e-05,
+      "loss": 1.2743,
+      "step": 12234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0814028571428572e-05,
+      "loss": 0.931,
+      "step": 12235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.081376e-05,
+      "loss": 1.0699,
+      "step": 12236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0813491428571428e-05,
+      "loss": 1.3725,
+      "step": 12237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0813222857142857e-05,
+      "loss": 1.6315,
+      "step": 12238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0812954285714287e-05,
+      "loss": 1.1614,
+      "step": 12239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0812685714285715e-05,
+      "loss": 1.001,
+      "step": 12240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0812417142857143e-05,
+      "loss": 1.0233,
+      "step": 12241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0812148571428571e-05,
+      "loss": 1.1369,
+      "step": 12242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0811880000000001e-05,
+      "loss": 1.1039,
+      "step": 12243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0811611428571429e-05,
+      "loss": 1.1624,
+      "step": 12244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0811342857142857e-05,
+      "loss": 0.9342,
+      "step": 12245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0811074285714287e-05,
+      "loss": 1.1751,
+      "step": 12246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0810805714285714e-05,
+      "loss": 1.4045,
+      "step": 12247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0810537142857143e-05,
+      "loss": 1.453,
+      "step": 12248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0810268571428572e-05,
+      "loss": 1.152,
+      "step": 12249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0810000000000002e-05,
+      "loss": 1.2015,
+      "step": 12250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0809731428571428e-05,
+      "loss": 0.884,
+      "step": 12251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0809462857142858e-05,
+      "loss": 1.5777,
+      "step": 12252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0809194285714286e-05,
+      "loss": 1.1672,
+      "step": 12253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0808925714285714e-05,
+      "loss": 1.4656,
+      "step": 12254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0808657142857142e-05,
+      "loss": 0.9822,
+      "step": 12255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0808388571428572e-05,
+      "loss": 1.307,
+      "step": 12256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080812e-05,
+      "loss": 0.9341,
+      "step": 12257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0807851428571429e-05,
+      "loss": 1.2444,
+      "step": 12258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0807582857142859e-05,
+      "loss": 1.5253,
+      "step": 12259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0807314285714285e-05,
+      "loss": 1.2758,
+      "step": 12260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0807045714285715e-05,
+      "loss": 1.2798,
+      "step": 12261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0806777142857143e-05,
+      "loss": 1.3977,
+      "step": 12262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0806508571428573e-05,
+      "loss": 1.6173,
+      "step": 12263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080624e-05,
+      "loss": 1.2191,
+      "step": 12264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080597142857143e-05,
+      "loss": 1.323,
+      "step": 12265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0805702857142857e-05,
+      "loss": 1.0352,
+      "step": 12266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0805434285714286e-05,
+      "loss": 1.2487,
+      "step": 12267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0805165714285714e-05,
+      "loss": 1.0445,
+      "step": 12268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0804897142857144e-05,
+      "loss": 1.2163,
+      "step": 12269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0804628571428572e-05,
+      "loss": 1.1284,
+      "step": 12270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080436e-05,
+      "loss": 1.1667,
+      "step": 12271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080409142857143e-05,
+      "loss": 0.9065,
+      "step": 12272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0803822857142858e-05,
+      "loss": 1.3918,
+      "step": 12273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0803554285714286e-05,
+      "loss": 1.3493,
+      "step": 12274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0803285714285714e-05,
+      "loss": 1.3658,
+      "step": 12275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0803017142857144e-05,
+      "loss": 1.1724,
+      "step": 12276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080274857142857e-05,
+      "loss": 1.4212,
+      "step": 12277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080248e-05,
+      "loss": 1.2774,
+      "step": 12278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0802211428571429e-05,
+      "loss": 1.4014,
+      "step": 12279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0801942857142857e-05,
+      "loss": 0.7483,
+      "step": 12280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0801674285714285e-05,
+      "loss": 1.1186,
+      "step": 12281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0801405714285715e-05,
+      "loss": 1.325,
+      "step": 12282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0801137142857143e-05,
+      "loss": 0.9152,
+      "step": 12283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0800868571428571e-05,
+      "loss": 1.154,
+      "step": 12284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0800600000000001e-05,
+      "loss": 1.3573,
+      "step": 12285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080033142857143e-05,
+      "loss": 1.3087,
+      "step": 12286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0800062857142858e-05,
+      "loss": 1.3136,
+      "step": 12287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0799794285714286e-05,
+      "loss": 0.93,
+      "step": 12288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0799525714285716e-05,
+      "loss": 1.1042,
+      "step": 12289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0799257142857142e-05,
+      "loss": 1.2685,
+      "step": 12290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0798988571428572e-05,
+      "loss": 1.5328,
+      "step": 12291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.079872e-05,
+      "loss": 1.2833,
+      "step": 12292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0798451428571428e-05,
+      "loss": 1.3019,
+      "step": 12293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0798182857142857e-05,
+      "loss": 1.1244,
+      "step": 12294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0797914285714286e-05,
+      "loss": 1.348,
+      "step": 12295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0797645714285716e-05,
+      "loss": 0.7086,
+      "step": 12296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0797377142857143e-05,
+      "loss": 1.4032,
+      "step": 12297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0797108571428573e-05,
+      "loss": 1.0446,
+      "step": 12298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.079684e-05,
+      "loss": 1.4,
+      "step": 12299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0796571428571429e-05,
+      "loss": 1.4285,
+      "step": 12300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0796302857142857e-05,
+      "loss": 1.682,
+      "step": 12301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0796034285714287e-05,
+      "loss": 1.4597,
+      "step": 12302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0795765714285714e-05,
+      "loss": 1.7315,
+      "step": 12303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0795497142857143e-05,
+      "loss": 1.4467,
+      "step": 12304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0795228571428572e-05,
+      "loss": 1.5879,
+      "step": 12305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.079496e-05,
+      "loss": 1.3122,
+      "step": 12306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0794691428571428e-05,
+      "loss": 1.3863,
+      "step": 12307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0794422857142858e-05,
+      "loss": 1.449,
+      "step": 12308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0794154285714288e-05,
+      "loss": 1.0272,
+      "step": 12309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0793885714285714e-05,
+      "loss": 1.3786,
+      "step": 12310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0793617142857144e-05,
+      "loss": 0.8437,
+      "step": 12311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0793348571428572e-05,
+      "loss": 0.8557,
+      "step": 12312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.079308e-05,
+      "loss": 0.7911,
+      "step": 12313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0792811428571429e-05,
+      "loss": 1.2587,
+      "step": 12314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0792542857142858e-05,
+      "loss": 1.4166,
+      "step": 12315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0792274285714285e-05,
+      "loss": 1.2024,
+      "step": 12316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0792005714285715e-05,
+      "loss": 0.9535,
+      "step": 12317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0791737142857143e-05,
+      "loss": 1.1559,
+      "step": 12318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0791468571428573e-05,
+      "loss": 1.1935,
+      "step": 12319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.07912e-05,
+      "loss": 0.9186,
+      "step": 12320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0790931428571429e-05,
+      "loss": 1.3439,
+      "step": 12321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0790662857142859e-05,
+      "loss": 0.8713,
+      "step": 12322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0790394285714285e-05,
+      "loss": 1.13,
+      "step": 12323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0790125714285715e-05,
+      "loss": 1.06,
+      "step": 12324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0789857142857144e-05,
+      "loss": 1.1202,
+      "step": 12325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0789588571428572e-05,
+      "loss": 1.4581,
+      "step": 12326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.078932e-05,
+      "loss": 1.5513,
+      "step": 12327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.078905142857143e-05,
+      "loss": 1.5186,
+      "step": 12328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0788782857142856e-05,
+      "loss": 1.3487,
+      "step": 12329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0788514285714286e-05,
+      "loss": 0.9929,
+      "step": 12330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0788245714285714e-05,
+      "loss": 1.2418,
+      "step": 12331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0787977142857144e-05,
+      "loss": 1.1543,
+      "step": 12332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.078770857142857e-05,
+      "loss": 1.6131,
+      "step": 12333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.078744e-05,
+      "loss": 0.9344,
+      "step": 12334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0787171428571429e-05,
+      "loss": 1.504,
+      "step": 12335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0786902857142857e-05,
+      "loss": 1.1136,
+      "step": 12336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0786634285714287e-05,
+      "loss": 0.8557,
+      "step": 12337
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0786365714285715e-05,
+      "loss": 1.0458,
+      "step": 12338
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0786097142857143e-05,
+      "loss": 0.9315,
+      "step": 12339
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0785828571428571e-05,
+      "loss": 1.386,
+      "step": 12340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0785560000000001e-05,
+      "loss": 1.258,
+      "step": 12341
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.078529142857143e-05,
+      "loss": 1.2401,
+      "step": 12342
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0785022857142857e-05,
+      "loss": 1.4772,
+      "step": 12343
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0784754285714286e-05,
+      "loss": 1.1529,
+      "step": 12344
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0784485714285716e-05,
+      "loss": 1.5584,
+      "step": 12345
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0784217142857142e-05,
+      "loss": 1.0096,
+      "step": 12346
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0783948571428572e-05,
+      "loss": 0.918,
+      "step": 12347
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.078368e-05,
+      "loss": 0.9857,
+      "step": 12348
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0783411428571428e-05,
+      "loss": 1.7958,
+      "step": 12349
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0783142857142858e-05,
+      "loss": 1.2731,
+      "step": 12350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0782874285714286e-05,
+      "loss": 1.18,
+      "step": 12351
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0782605714285714e-05,
+      "loss": 1.292,
+      "step": 12352
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0782337142857143e-05,
+      "loss": 1.0652,
+      "step": 12353
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0782068571428573e-05,
+      "loss": 1.0185,
+      "step": 12354
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.07818e-05,
+      "loss": 1.0256,
+      "step": 12355
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0781531428571429e-05,
+      "loss": 1.2786,
+      "step": 12356
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0781262857142857e-05,
+      "loss": 1.1221,
+      "step": 12357
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0780994285714287e-05,
+      "loss": 1.5562,
+      "step": 12358
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0780725714285713e-05,
+      "loss": 1.1748,
+      "step": 12359
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0780457142857143e-05,
+      "loss": 1.5643,
+      "step": 12360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0780188571428571e-05,
+      "loss": 1.4318,
+      "step": 12361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.077992e-05,
+      "loss": 1.0653,
+      "step": 12362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.077965142857143e-05,
+      "loss": 0.7251,
+      "step": 12363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0779382857142858e-05,
+      "loss": 1.0619,
+      "step": 12364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0779114285714288e-05,
+      "loss": 1.3074,
+      "step": 12365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0778845714285714e-05,
+      "loss": 1.3187,
+      "step": 12366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0778577142857144e-05,
+      "loss": 1.1449,
+      "step": 12367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0778308571428572e-05,
+      "loss": 1.1661,
+      "step": 12368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.077804e-05,
+      "loss": 1.2862,
+      "step": 12369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0777771428571428e-05,
+      "loss": 1.1177,
+      "step": 12370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0777502857142858e-05,
+      "loss": 1.2608,
+      "step": 12371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0777234285714285e-05,
+      "loss": 0.6729,
+      "step": 12372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0776965714285715e-05,
+      "loss": 1.1694,
+      "step": 12373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0776697142857143e-05,
+      "loss": 1.4726,
+      "step": 12374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0776428571428571e-05,
+      "loss": 0.829,
+      "step": 12375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0776160000000001e-05,
+      "loss": 0.9158,
+      "step": 12376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0775891428571429e-05,
+      "loss": 0.9169,
+      "step": 12377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0775622857142859e-05,
+      "loss": 0.9368,
+      "step": 12378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0775354285714285e-05,
+      "loss": 0.779,
+      "step": 12379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0775085714285715e-05,
+      "loss": 1.4286,
+      "step": 12380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0774817142857143e-05,
+      "loss": 1.4832,
+      "step": 12381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0774548571428572e-05,
+      "loss": 1.2988,
+      "step": 12382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.077428e-05,
+      "loss": 1.4242,
+      "step": 12383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.077401142857143e-05,
+      "loss": 0.9249,
+      "step": 12384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0773742857142856e-05,
+      "loss": 1.1127,
+      "step": 12385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0773474285714286e-05,
+      "loss": 1.1193,
+      "step": 12386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0773205714285714e-05,
+      "loss": 1.4188,
+      "step": 12387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0772937142857142e-05,
+      "loss": 1.5545,
+      "step": 12388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0772668571428572e-05,
+      "loss": 1.7927,
+      "step": 12389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.07724e-05,
+      "loss": 1.4808,
+      "step": 12390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.077213142857143e-05,
+      "loss": 1.084,
+      "step": 12391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0771862857142857e-05,
+      "loss": 1.2229,
+      "step": 12392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0771594285714287e-05,
+      "loss": 0.9838,
+      "step": 12393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0771325714285715e-05,
+      "loss": 0.9888,
+      "step": 12394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0771057142857143e-05,
+      "loss": 1.3949,
+      "step": 12395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0770788571428571e-05,
+      "loss": 1.1712,
+      "step": 12396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0770520000000001e-05,
+      "loss": 1.3971,
+      "step": 12397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0770251428571428e-05,
+      "loss": 1.4655,
+      "step": 12398
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0769982857142857e-05,
+      "loss": 1.3013,
+      "step": 12399
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0769714285714286e-05,
+      "loss": 1.0188,
+      "step": 12400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0769445714285715e-05,
+      "loss": 1.5871,
+      "step": 12401
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0769177142857144e-05,
+      "loss": 1.2227,
+      "step": 12402
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0768908571428572e-05,
+      "loss": 1.4212,
+      "step": 12403
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0768640000000002e-05,
+      "loss": 1.3386,
+      "step": 12404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0768371428571428e-05,
+      "loss": 1.1665,
+      "step": 12405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0768102857142858e-05,
+      "loss": 0.8384,
+      "step": 12406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0767834285714286e-05,
+      "loss": 1.3592,
+      "step": 12407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0767565714285714e-05,
+      "loss": 1.081,
+      "step": 12408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0767297142857143e-05,
+      "loss": 1.5428,
+      "step": 12409
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0767028571428572e-05,
+      "loss": 0.7344,
+      "step": 12410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0766759999999999e-05,
+      "loss": 1.4729,
+      "step": 12411
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0766491428571429e-05,
+      "loss": 1.6533,
+      "step": 12412
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0766222857142857e-05,
+      "loss": 1.0244,
+      "step": 12413
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0765954285714287e-05,
+      "loss": 0.7723,
+      "step": 12414
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0765685714285715e-05,
+      "loss": 1.157,
+      "step": 12415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0765417142857143e-05,
+      "loss": 1.3384,
+      "step": 12416
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0765148571428573e-05,
+      "loss": 1.2853,
+      "step": 12417
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.076488e-05,
+      "loss": 1.4287,
+      "step": 12418
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.076461142857143e-05,
+      "loss": 1.0675,
+      "step": 12419
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0764342857142858e-05,
+      "loss": 1.1023,
+      "step": 12420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0764074285714286e-05,
+      "loss": 0.8784,
+      "step": 12421
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0763805714285714e-05,
+      "loss": 1.4846,
+      "step": 12422
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0763537142857144e-05,
+      "loss": 0.8629,
+      "step": 12423
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0763268571428572e-05,
+      "loss": 1.0928,
+      "step": 12424
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0763e-05,
+      "loss": 1.1218,
+      "step": 12425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0762731428571428e-05,
+      "loss": 1.5913,
+      "step": 12426
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0762462857142858e-05,
+      "loss": 1.1677,
+      "step": 12427
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0762194285714286e-05,
+      "loss": 0.9752,
+      "step": 12428
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0761925714285715e-05,
+      "loss": 1.5667,
+      "step": 12429
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0761657142857144e-05,
+      "loss": 1.5679,
+      "step": 12430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0761388571428571e-05,
+      "loss": 0.8316,
+      "step": 12431
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.076112e-05,
+      "loss": 1.6059,
+      "step": 12432
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0760851428571429e-05,
+      "loss": 1.2524,
+      "step": 12433
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0760582857142857e-05,
+      "loss": 1.2888,
+      "step": 12434
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0760314285714285e-05,
+      "loss": 0.7313,
+      "step": 12435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0760045714285715e-05,
+      "loss": 1.4674,
+      "step": 12436
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0759777142857143e-05,
+      "loss": 1.1546,
+      "step": 12437
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0759508571428571e-05,
+      "loss": 1.5023,
+      "step": 12438
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.075924e-05,
+      "loss": 1.4642,
+      "step": 12439
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.075897142857143e-05,
+      "loss": 1.0784,
+      "step": 12440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0758702857142858e-05,
+      "loss": 1.2014,
+      "step": 12441
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0758434285714286e-05,
+      "loss": 1.5205,
+      "step": 12442
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0758165714285716e-05,
+      "loss": 0.6949,
+      "step": 12443
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0757897142857142e-05,
+      "loss": 0.7181,
+      "step": 12444
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0757628571428572e-05,
+      "loss": 1.2786,
+      "step": 12445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.075736e-05,
+      "loss": 1.5256,
+      "step": 12446
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.075709142857143e-05,
+      "loss": 1.1331,
+      "step": 12447
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0756822857142857e-05,
+      "loss": 0.9027,
+      "step": 12448
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0756554285714287e-05,
+      "loss": 1.1164,
+      "step": 12449
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0756285714285715e-05,
+      "loss": 1.3109,
+      "step": 12450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0756017142857143e-05,
+      "loss": 1.393,
+      "step": 12451
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0755748571428571e-05,
+      "loss": 1.0913,
+      "step": 12452
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0755480000000001e-05,
+      "loss": 1.3639,
+      "step": 12453
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0755211428571429e-05,
+      "loss": 1.4538,
+      "step": 12454
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0754942857142857e-05,
+      "loss": 1.6347,
+      "step": 12455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0754674285714287e-05,
+      "loss": 1.1721,
+      "step": 12456
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0754405714285714e-05,
+      "loss": 1.1669,
+      "step": 12457
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0754137142857143e-05,
+      "loss": 1.1285,
+      "step": 12458
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0753868571428572e-05,
+      "loss": 1.0792,
+      "step": 12459
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0753600000000002e-05,
+      "loss": 1.3517,
+      "step": 12460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0753331428571428e-05,
+      "loss": 0.9394,
+      "step": 12461
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0753062857142858e-05,
+      "loss": 1.205,
+      "step": 12462
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0752794285714286e-05,
+      "loss": 1.5757,
+      "step": 12463
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0752525714285714e-05,
+      "loss": 1.4842,
+      "step": 12464
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0752257142857142e-05,
+      "loss": 1.0894,
+      "step": 12465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0751988571428572e-05,
+      "loss": 1.0349,
+      "step": 12466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.075172e-05,
+      "loss": 1.4162,
+      "step": 12467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0751451428571429e-05,
+      "loss": 1.111,
+      "step": 12468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0751182857142859e-05,
+      "loss": 1.0453,
+      "step": 12469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0750914285714287e-05,
+      "loss": 1.1635,
+      "step": 12470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0750645714285715e-05,
+      "loss": 1.4098,
+      "step": 12471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0750377142857143e-05,
+      "loss": 0.8227,
+      "step": 12472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0750108571428573e-05,
+      "loss": 1.4315,
+      "step": 12473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074984e-05,
+      "loss": 1.0625,
+      "step": 12474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074957142857143e-05,
+      "loss": 1.3024,
+      "step": 12475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0749302857142857e-05,
+      "loss": 1.088,
+      "step": 12476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0749034285714286e-05,
+      "loss": 1.4839,
+      "step": 12477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0748765714285714e-05,
+      "loss": 1.0915,
+      "step": 12478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0748497142857144e-05,
+      "loss": 1.454,
+      "step": 12479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0748228571428572e-05,
+      "loss": 0.6885,
+      "step": 12480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074796e-05,
+      "loss": 0.9546,
+      "step": 12481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074769142857143e-05,
+      "loss": 1.6234,
+      "step": 12482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0747422857142858e-05,
+      "loss": 0.9912,
+      "step": 12483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0747154285714286e-05,
+      "loss": 1.2432,
+      "step": 12484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0746885714285714e-05,
+      "loss": 1.2683,
+      "step": 12485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0746617142857144e-05,
+      "loss": 1.1738,
+      "step": 12486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074634857142857e-05,
+      "loss": 1.1085,
+      "step": 12487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074608e-05,
+      "loss": 1.1319,
+      "step": 12488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0745811428571429e-05,
+      "loss": 1.4736,
+      "step": 12489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0745542857142857e-05,
+      "loss": 0.9854,
+      "step": 12490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0745274285714285e-05,
+      "loss": 1.2657,
+      "step": 12491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0745005714285715e-05,
+      "loss": 1.1363,
+      "step": 12492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0744737142857143e-05,
+      "loss": 1.1228,
+      "step": 12493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0744468571428571e-05,
+      "loss": 1.2457,
+      "step": 12494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0744200000000001e-05,
+      "loss": 1.6499,
+      "step": 12495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074393142857143e-05,
+      "loss": 0.6999,
+      "step": 12496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0743662857142858e-05,
+      "loss": 1.348,
+      "step": 12497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0743394285714286e-05,
+      "loss": 1.2194,
+      "step": 12498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0743125714285716e-05,
+      "loss": 1.1235,
+      "step": 12499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0742857142857142e-05,
+      "loss": 1.3099,
+      "step": 12500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0742588571428572e-05,
+      "loss": 1.0774,
+      "step": 12501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074232e-05,
+      "loss": 1.2775,
+      "step": 12502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0742051428571428e-05,
+      "loss": 1.5784,
+      "step": 12503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0741782857142857e-05,
+      "loss": 1.16,
+      "step": 12504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0741514285714286e-05,
+      "loss": 0.9521,
+      "step": 12505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0741245714285715e-05,
+      "loss": 1.3213,
+      "step": 12506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0740977142857143e-05,
+      "loss": 1.2214,
+      "step": 12507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0740708571428573e-05,
+      "loss": 1.2855,
+      "step": 12508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.074044e-05,
+      "loss": 1.4951,
+      "step": 12509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0740171428571429e-05,
+      "loss": 1.1282,
+      "step": 12510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0739902857142857e-05,
+      "loss": 1.9065,
+      "step": 12511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0739634285714287e-05,
+      "loss": 1.1133,
+      "step": 12512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0739365714285714e-05,
+      "loss": 1.3499,
+      "step": 12513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0739097142857143e-05,
+      "loss": 1.3925,
+      "step": 12514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0738828571428572e-05,
+      "loss": 1.2206,
+      "step": 12515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0738560000000001e-05,
+      "loss": 1.4736,
+      "step": 12516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0738291428571428e-05,
+      "loss": 1.5917,
+      "step": 12517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0738022857142858e-05,
+      "loss": 1.4757,
+      "step": 12518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0737754285714286e-05,
+      "loss": 1.1203,
+      "step": 12519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0737485714285714e-05,
+      "loss": 1.5993,
+      "step": 12520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0737217142857144e-05,
+      "loss": 1.8602,
+      "step": 12521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0736948571428572e-05,
+      "loss": 1.2217,
+      "step": 12522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.073668e-05,
+      "loss": 1.2019,
+      "step": 12523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0736411428571429e-05,
+      "loss": 1.3391,
+      "step": 12524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0736142857142858e-05,
+      "loss": 1.2219,
+      "step": 12525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0735874285714285e-05,
+      "loss": 1.2693,
+      "step": 12526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0735605714285715e-05,
+      "loss": 1.3054,
+      "step": 12527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0735337142857143e-05,
+      "loss": 1.1496,
+      "step": 12528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0735068571428573e-05,
+      "loss": 1.5204,
+      "step": 12529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.07348e-05,
+      "loss": 1.3539,
+      "step": 12530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0734531428571429e-05,
+      "loss": 1.092,
+      "step": 12531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0734262857142857e-05,
+      "loss": 1.1913,
+      "step": 12532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0733994285714285e-05,
+      "loss": 1.2355,
+      "step": 12533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0733725714285715e-05,
+      "loss": 1.3206,
+      "step": 12534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0733457142857144e-05,
+      "loss": 1.1777,
+      "step": 12535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0733188571428572e-05,
+      "loss": 1.0673,
+      "step": 12536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.073292e-05,
+      "loss": 1.2776,
+      "step": 12537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.073265142857143e-05,
+      "loss": 1.1141,
+      "step": 12538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0732382857142858e-05,
+      "loss": 1.3863,
+      "step": 12539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0732114285714286e-05,
+      "loss": 1.3676,
+      "step": 12540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0731845714285714e-05,
+      "loss": 1.2913,
+      "step": 12541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0731577142857144e-05,
+      "loss": 1.4354,
+      "step": 12542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.073130857142857e-05,
+      "loss": 1.2255,
+      "step": 12543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.073104e-05,
+      "loss": 1.1593,
+      "step": 12544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0730771428571429e-05,
+      "loss": 0.7968,
+      "step": 12545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0730502857142857e-05,
+      "loss": 1.1729,
+      "step": 12546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0730234285714287e-05,
+      "loss": 1.49,
+      "step": 12547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0729965714285715e-05,
+      "loss": 0.9739,
+      "step": 12548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0729697142857143e-05,
+      "loss": 1.33,
+      "step": 12549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0729428571428571e-05,
+      "loss": 1.0305,
+      "step": 12550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0729160000000001e-05,
+      "loss": 0.7971,
+      "step": 12551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.072889142857143e-05,
+      "loss": 1.0584,
+      "step": 12552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0728622857142857e-05,
+      "loss": 0.7216,
+      "step": 12553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0728354285714286e-05,
+      "loss": 1.0437,
+      "step": 12554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0728085714285716e-05,
+      "loss": 1.2522,
+      "step": 12555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0727817142857142e-05,
+      "loss": 1.4486,
+      "step": 12556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0727548571428572e-05,
+      "loss": 1.2512,
+      "step": 12557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.072728e-05,
+      "loss": 1.0683,
+      "step": 12558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0727011428571428e-05,
+      "loss": 1.0528,
+      "step": 12559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0726742857142858e-05,
+      "loss": 1.1443,
+      "step": 12560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0726474285714286e-05,
+      "loss": 1.7649,
+      "step": 12561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0726205714285714e-05,
+      "loss": 1.1884,
+      "step": 12562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0725937142857143e-05,
+      "loss": 1.3826,
+      "step": 12563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0725668571428573e-05,
+      "loss": 1.5463,
+      "step": 12564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.07254e-05,
+      "loss": 1.4478,
+      "step": 12565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0725131428571429e-05,
+      "loss": 1.1813,
+      "step": 12566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0724862857142857e-05,
+      "loss": 0.9307,
+      "step": 12567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0724594285714287e-05,
+      "loss": 1.0695,
+      "step": 12568
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0724325714285713e-05,
+      "loss": 1.3797,
+      "step": 12569
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0724057142857143e-05,
+      "loss": 1.0673,
+      "step": 12570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0723788571428571e-05,
+      "loss": 1.2231,
+      "step": 12571
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.072352e-05,
+      "loss": 1.2391,
+      "step": 12572
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.072325142857143e-05,
+      "loss": 1.0511,
+      "step": 12573
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0722982857142858e-05,
+      "loss": 0.9952,
+      "step": 12574
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0722714285714288e-05,
+      "loss": 1.2816,
+      "step": 12575
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0722445714285714e-05,
+      "loss": 1.2124,
+      "step": 12576
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0722177142857144e-05,
+      "loss": 1.4273,
+      "step": 12577
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0721908571428572e-05,
+      "loss": 1.3454,
+      "step": 12578
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.072164e-05,
+      "loss": 1.5465,
+      "step": 12579
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0721371428571428e-05,
+      "loss": 0.9947,
+      "step": 12580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0721102857142858e-05,
+      "loss": 1.4235,
+      "step": 12581
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0720834285714285e-05,
+      "loss": 1.3218,
+      "step": 12582
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0720565714285715e-05,
+      "loss": 0.8517,
+      "step": 12583
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0720297142857143e-05,
+      "loss": 1.1637,
+      "step": 12584
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0720028571428571e-05,
+      "loss": 1.3511,
+      "step": 12585
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0719760000000001e-05,
+      "loss": 1.373,
+      "step": 12586
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0719491428571429e-05,
+      "loss": 1.4525,
+      "step": 12587
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0719222857142859e-05,
+      "loss": 1.5682,
+      "step": 12588
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0718954285714285e-05,
+      "loss": 1.3636,
+      "step": 12589
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0718685714285715e-05,
+      "loss": 1.2565,
+      "step": 12590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0718417142857143e-05,
+      "loss": 1.1848,
+      "step": 12591
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0718148571428572e-05,
+      "loss": 0.8414,
+      "step": 12592
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.071788e-05,
+      "loss": 1.3135,
+      "step": 12593
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.071761142857143e-05,
+      "loss": 1.1219,
+      "step": 12594
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0717342857142856e-05,
+      "loss": 1.5202,
+      "step": 12595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0717074285714286e-05,
+      "loss": 1.3263,
+      "step": 12596
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0716805714285714e-05,
+      "loss": 0.9585,
+      "step": 12597
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0716537142857144e-05,
+      "loss": 1.0885,
+      "step": 12598
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0716268571428572e-05,
+      "loss": 1.1141,
+      "step": 12599
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0716e-05,
+      "loss": 1.3062,
+      "step": 12600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.071573142857143e-05,
+      "loss": 1.168,
+      "step": 12601
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0715462857142857e-05,
+      "loss": 1.0739,
+      "step": 12602
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0715194285714287e-05,
+      "loss": 1.1227,
+      "step": 12603
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0714925714285715e-05,
+      "loss": 1.4293,
+      "step": 12604
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0714657142857143e-05,
+      "loss": 0.9964,
+      "step": 12605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0714388571428571e-05,
+      "loss": 1.0239,
+      "step": 12606
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0714120000000001e-05,
+      "loss": 1.0694,
+      "step": 12607
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0713851428571428e-05,
+      "loss": 1.1644,
+      "step": 12608
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0713582857142857e-05,
+      "loss": 1.4202,
+      "step": 12609
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0713314285714286e-05,
+      "loss": 1.0053,
+      "step": 12610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0713045714285715e-05,
+      "loss": 1.137,
+      "step": 12611
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0712777142857144e-05,
+      "loss": 1.4067,
+      "step": 12612
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0712508571428572e-05,
+      "loss": 1.1126,
+      "step": 12613
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0712240000000002e-05,
+      "loss": 1.5798,
+      "step": 12614
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0711971428571428e-05,
+      "loss": 1.3414,
+      "step": 12615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0711702857142858e-05,
+      "loss": 1.1039,
+      "step": 12616
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0711434285714286e-05,
+      "loss": 1.1305,
+      "step": 12617
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0711165714285714e-05,
+      "loss": 1.4508,
+      "step": 12618
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0710897142857143e-05,
+      "loss": 1.0149,
+      "step": 12619
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0710628571428572e-05,
+      "loss": 0.9929,
+      "step": 12620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.071036e-05,
+      "loss": 0.9832,
+      "step": 12621
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0710091428571429e-05,
+      "loss": 1.1797,
+      "step": 12622
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0709822857142857e-05,
+      "loss": 1.3341,
+      "step": 12623
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0709554285714287e-05,
+      "loss": 1.1696,
+      "step": 12624
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0709285714285715e-05,
+      "loss": 1.3821,
+      "step": 12625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0709017142857143e-05,
+      "loss": 1.3462,
+      "step": 12626
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0708748571428573e-05,
+      "loss": 1.3857,
+      "step": 12627
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.070848e-05,
+      "loss": 1.1125,
+      "step": 12628
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.070821142857143e-05,
+      "loss": 0.7287,
+      "step": 12629
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0707942857142858e-05,
+      "loss": 1.5406,
+      "step": 12630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0707674285714286e-05,
+      "loss": 1.1807,
+      "step": 12631
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0707405714285714e-05,
+      "loss": 1.6775,
+      "step": 12632
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0707137142857144e-05,
+      "loss": 1.4665,
+      "step": 12633
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0706868571428572e-05,
+      "loss": 0.9888,
+      "step": 12634
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.07066e-05,
+      "loss": 1.4841,
+      "step": 12635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0706331428571428e-05,
+      "loss": 1.1308,
+      "step": 12636
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0706062857142858e-05,
+      "loss": 1.5256,
+      "step": 12637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0705794285714286e-05,
+      "loss": 1.2056,
+      "step": 12638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0705525714285715e-05,
+      "loss": 1.2892,
+      "step": 12639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0705257142857144e-05,
+      "loss": 1.2962,
+      "step": 12640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0704988571428571e-05,
+      "loss": 1.1165,
+      "step": 12641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.070472e-05,
+      "loss": 1.1086,
+      "step": 12642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0704451428571429e-05,
+      "loss": 1.3153,
+      "step": 12643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0704182857142859e-05,
+      "loss": 1.332,
+      "step": 12644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0703914285714285e-05,
+      "loss": 1.1386,
+      "step": 12645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0703645714285715e-05,
+      "loss": 0.9666,
+      "step": 12646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0703377142857143e-05,
+      "loss": 1.641,
+      "step": 12647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0703108571428571e-05,
+      "loss": 0.8996,
+      "step": 12648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.070284e-05,
+      "loss": 1.138,
+      "step": 12649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.070257142857143e-05,
+      "loss": 0.819,
+      "step": 12650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0702302857142858e-05,
+      "loss": 1.4219,
+      "step": 12651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0702034285714286e-05,
+      "loss": 1.3967,
+      "step": 12652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0701765714285716e-05,
+      "loss": 0.9312,
+      "step": 12653
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0701497142857142e-05,
+      "loss": 1.5047,
+      "step": 12654
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0701228571428572e-05,
+      "loss": 1.288,
+      "step": 12655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.070096e-05,
+      "loss": 1.9211,
+      "step": 12656
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.070069142857143e-05,
+      "loss": 1.3215,
+      "step": 12657
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0700422857142857e-05,
+      "loss": 1.1251,
+      "step": 12658
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0700154285714287e-05,
+      "loss": 1.2742,
+      "step": 12659
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0699885714285715e-05,
+      "loss": 0.749,
+      "step": 12660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0699617142857143e-05,
+      "loss": 1.0346,
+      "step": 12661
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0699348571428571e-05,
+      "loss": 1.1259,
+      "step": 12662
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0699080000000001e-05,
+      "loss": 1.2608,
+      "step": 12663
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0698811428571429e-05,
+      "loss": 1.3032,
+      "step": 12664
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0698542857142857e-05,
+      "loss": 1.1917,
+      "step": 12665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0698274285714287e-05,
+      "loss": 1.0806,
+      "step": 12666
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0698005714285715e-05,
+      "loss": 1.3175,
+      "step": 12667
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0697737142857143e-05,
+      "loss": 1.0058,
+      "step": 12668
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0697468571428572e-05,
+      "loss": 1.5602,
+      "step": 12669
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0697200000000002e-05,
+      "loss": 1.2363,
+      "step": 12670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0696931428571428e-05,
+      "loss": 1.2103,
+      "step": 12671
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0696662857142858e-05,
+      "loss": 0.8985,
+      "step": 12672
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0696394285714286e-05,
+      "loss": 1.0365,
+      "step": 12673
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0696125714285714e-05,
+      "loss": 1.1687,
+      "step": 12674
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0695857142857142e-05,
+      "loss": 1.2005,
+      "step": 12675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0695588571428572e-05,
+      "loss": 1.487,
+      "step": 12676
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.069532e-05,
+      "loss": 1.4747,
+      "step": 12677
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0695051428571429e-05,
+      "loss": 0.7919,
+      "step": 12678
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0694782857142859e-05,
+      "loss": 1.1296,
+      "step": 12679
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0694514285714287e-05,
+      "loss": 1.1932,
+      "step": 12680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0694245714285715e-05,
+      "loss": 1.099,
+      "step": 12681
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0693977142857143e-05,
+      "loss": 1.2571,
+      "step": 12682
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0693708571428573e-05,
+      "loss": 1.1218,
+      "step": 12683
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.069344e-05,
+      "loss": 1.0176,
+      "step": 12684
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.069317142857143e-05,
+      "loss": 1.1786,
+      "step": 12685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0692902857142857e-05,
+      "loss": 0.9832,
+      "step": 12686
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.0692634285714286e-05,
+      "loss": 1.0255,
+      "step": 12687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0692365714285714e-05,
+      "loss": 1.6454,
+      "step": 12688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0692097142857144e-05,
+      "loss": 1.3613,
+      "step": 12689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0691828571428572e-05,
+      "loss": 1.3301,
+      "step": 12690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.069156e-05,
+      "loss": 1.5991,
+      "step": 12691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.069129142857143e-05,
+      "loss": 0.8648,
+      "step": 12692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0691022857142858e-05,
+      "loss": 1.7101,
+      "step": 12693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0690754285714286e-05,
+      "loss": 1.2865,
+      "step": 12694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0690485714285714e-05,
+      "loss": 1.445,
+      "step": 12695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0690217142857144e-05,
+      "loss": 1.3389,
+      "step": 12696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.068994857142857e-05,
+      "loss": 1.4318,
+      "step": 12697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.068968e-05,
+      "loss": 1.1759,
+      "step": 12698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0689411428571429e-05,
+      "loss": 0.8716,
+      "step": 12699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0689142857142857e-05,
+      "loss": 1.1786,
+      "step": 12700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0688874285714285e-05,
+      "loss": 1.3724,
+      "step": 12701
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0688605714285715e-05,
+      "loss": 0.9507,
+      "step": 12702
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0688337142857143e-05,
+      "loss": 1.0731,
+      "step": 12703
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0688068571428571e-05,
+      "loss": 1.4083,
+      "step": 12704
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0687800000000001e-05,
+      "loss": 1.1087,
+      "step": 12705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.068753142857143e-05,
+      "loss": 1.3476,
+      "step": 12706
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0687262857142858e-05,
+      "loss": 1.1594,
+      "step": 12707
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0686994285714286e-05,
+      "loss": 1.3374,
+      "step": 12708
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0686725714285716e-05,
+      "loss": 1.4665,
+      "step": 12709
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0686457142857142e-05,
+      "loss": 1.0213,
+      "step": 12710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0686188571428572e-05,
+      "loss": 1.2334,
+      "step": 12711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.068592e-05,
+      "loss": 1.2165,
+      "step": 12712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0685651428571428e-05,
+      "loss": 0.9589,
+      "step": 12713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0685382857142857e-05,
+      "loss": 1.2604,
+      "step": 12714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0685114285714286e-05,
+      "loss": 1.5257,
+      "step": 12715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0684845714285715e-05,
+      "loss": 1.2987,
+      "step": 12716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0684577142857143e-05,
+      "loss": 1.2511,
+      "step": 12717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0684308571428573e-05,
+      "loss": 1.2585,
+      "step": 12718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.068404e-05,
+      "loss": 1.5246,
+      "step": 12719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0683771428571429e-05,
+      "loss": 0.8044,
+      "step": 12720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0683502857142857e-05,
+      "loss": 0.7713,
+      "step": 12721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0683234285714287e-05,
+      "loss": 1.1038,
+      "step": 12722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0682965714285713e-05,
+      "loss": 1.2254,
+      "step": 12723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0682697142857143e-05,
+      "loss": 1.1404,
+      "step": 12724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0682428571428572e-05,
+      "loss": 1.2488,
+      "step": 12725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0682160000000001e-05,
+      "loss": 1.2599,
+      "step": 12726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0681891428571428e-05,
+      "loss": 1.2334,
+      "step": 12727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0681622857142858e-05,
+      "loss": 1.3819,
+      "step": 12728
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0681354285714286e-05,
+      "loss": 1.337,
+      "step": 12729
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0681085714285714e-05,
+      "loss": 1.2214,
+      "step": 12730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0680817142857144e-05,
+      "loss": 1.4038,
+      "step": 12731
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0680548571428572e-05,
+      "loss": 1.1856,
+      "step": 12732
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.068028e-05,
+      "loss": 1.1505,
+      "step": 12733
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0680011428571429e-05,
+      "loss": 1.3282,
+      "step": 12734
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0679742857142858e-05,
+      "loss": 1.117,
+      "step": 12735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0679474285714285e-05,
+      "loss": 1.602,
+      "step": 12736
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0679205714285715e-05,
+      "loss": 0.9736,
+      "step": 12737
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0678937142857143e-05,
+      "loss": 1.5047,
+      "step": 12738
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0678668571428573e-05,
+      "loss": 1.0622,
+      "step": 12739
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.06784e-05,
+      "loss": 1.2884,
+      "step": 12740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0678131428571429e-05,
+      "loss": 1.3441,
+      "step": 12741
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0677862857142857e-05,
+      "loss": 1.0804,
+      "step": 12742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0677594285714285e-05,
+      "loss": 1.218,
+      "step": 12743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0677325714285715e-05,
+      "loss": 1.1152,
+      "step": 12744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0677057142857144e-05,
+      "loss": 1.5778,
+      "step": 12745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0676788571428572e-05,
+      "loss": 1.1213,
+      "step": 12746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.067652e-05,
+      "loss": 1.1469,
+      "step": 12747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.067625142857143e-05,
+      "loss": 1.2802,
+      "step": 12748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0675982857142858e-05,
+      "loss": 0.9836,
+      "step": 12749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0675714285714286e-05,
+      "loss": 0.8114,
+      "step": 12750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0675445714285714e-05,
+      "loss": 1.3753,
+      "step": 12751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0675177142857144e-05,
+      "loss": 1.1139,
+      "step": 12752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.067490857142857e-05,
+      "loss": 1.3641,
+      "step": 12753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.067464e-05,
+      "loss": 0.9907,
+      "step": 12754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0674371428571429e-05,
+      "loss": 1.6453,
+      "step": 12755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0674102857142857e-05,
+      "loss": 1.4166,
+      "step": 12756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0673834285714287e-05,
+      "loss": 1.1142,
+      "step": 12757
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0673565714285715e-05,
+      "loss": 1.4617,
+      "step": 12758
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0673297142857143e-05,
+      "loss": 1.1677,
+      "step": 12759
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0673028571428571e-05,
+      "loss": 1.0811,
+      "step": 12760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0672760000000001e-05,
+      "loss": 1.1806,
+      "step": 12761
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.067249142857143e-05,
+      "loss": 1.1534,
+      "step": 12762
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0672222857142857e-05,
+      "loss": 1.213,
+      "step": 12763
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0671954285714286e-05,
+      "loss": 1.3063,
+      "step": 12764
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0671685714285716e-05,
+      "loss": 1.5879,
+      "step": 12765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0671417142857142e-05,
+      "loss": 1.2746,
+      "step": 12766
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0671148571428572e-05,
+      "loss": 1.4834,
+      "step": 12767
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.067088e-05,
+      "loss": 1.2267,
+      "step": 12768
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0670611428571428e-05,
+      "loss": 1.0909,
+      "step": 12769
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0670342857142858e-05,
+      "loss": 1.8649,
+      "step": 12770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0670074285714286e-05,
+      "loss": 1.2611,
+      "step": 12771
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0669805714285716e-05,
+      "loss": 1.1502,
+      "step": 12772
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0669537142857143e-05,
+      "loss": 1.1955,
+      "step": 12773
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0669268571428573e-05,
+      "loss": 1.5144,
+      "step": 12774
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0669e-05,
+      "loss": 1.3303,
+      "step": 12775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0668731428571429e-05,
+      "loss": 1.0949,
+      "step": 12776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0668462857142857e-05,
+      "loss": 0.6324,
+      "step": 12777
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0668194285714287e-05,
+      "loss": 1.5045,
+      "step": 12778
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0667925714285713e-05,
+      "loss": 1.0845,
+      "step": 12779
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0667657142857143e-05,
+      "loss": 1.4675,
+      "step": 12780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0667388571428571e-05,
+      "loss": 1.5766,
+      "step": 12781
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.066712e-05,
+      "loss": 1.2987,
+      "step": 12782
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.066685142857143e-05,
+      "loss": 1.3359,
+      "step": 12783
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0666582857142858e-05,
+      "loss": 1.3624,
+      "step": 12784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0666314285714288e-05,
+      "loss": 1.6157,
+      "step": 12785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0666045714285714e-05,
+      "loss": 1.1325,
+      "step": 12786
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0665777142857144e-05,
+      "loss": 1.2537,
+      "step": 12787
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0665508571428572e-05,
+      "loss": 1.0108,
+      "step": 12788
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.066524e-05,
+      "loss": 0.8779,
+      "step": 12789
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0664971428571428e-05,
+      "loss": 1.6096,
+      "step": 12790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0664702857142858e-05,
+      "loss": 1.4864,
+      "step": 12791
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0664434285714285e-05,
+      "loss": 0.984,
+      "step": 12792
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0664165714285715e-05,
+      "loss": 1.3418,
+      "step": 12793
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0663897142857143e-05,
+      "loss": 1.2629,
+      "step": 12794
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0663628571428573e-05,
+      "loss": 1.3469,
+      "step": 12795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0663360000000001e-05,
+      "loss": 1.2057,
+      "step": 12796
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0663091428571429e-05,
+      "loss": 0.9091,
+      "step": 12797
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0662822857142859e-05,
+      "loss": 1.3633,
+      "step": 12798
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0662554285714285e-05,
+      "loss": 0.8681,
+      "step": 12799
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0662285714285715e-05,
+      "loss": 1.2372,
+      "step": 12800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0662017142857143e-05,
+      "loss": 1.2137,
+      "step": 12801
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0661748571428572e-05,
+      "loss": 0.8955,
+      "step": 12802
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.066148e-05,
+      "loss": 1.4777,
+      "step": 12803
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.066121142857143e-05,
+      "loss": 1.6692,
+      "step": 12804
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0660942857142856e-05,
+      "loss": 1.0413,
+      "step": 12805
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0660674285714286e-05,
+      "loss": 1.4772,
+      "step": 12806
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0660405714285714e-05,
+      "loss": 0.7767,
+      "step": 12807
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0660137142857144e-05,
+      "loss": 1.1869,
+      "step": 12808
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0659868571428572e-05,
+      "loss": 0.8289,
+      "step": 12809
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.06596e-05,
+      "loss": 1.1236,
+      "step": 12810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.065933142857143e-05,
+      "loss": 1.0094,
+      "step": 12811
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0659062857142857e-05,
+      "loss": 1.3032,
+      "step": 12812
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0658794285714287e-05,
+      "loss": 0.837,
+      "step": 12813
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0658525714285715e-05,
+      "loss": 1.0869,
+      "step": 12814
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0658257142857143e-05,
+      "loss": 1.2503,
+      "step": 12815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0657988571428571e-05,
+      "loss": 1.3394,
+      "step": 12816
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0657720000000001e-05,
+      "loss": 1.3142,
+      "step": 12817
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.065745142857143e-05,
+      "loss": 1.6018,
+      "step": 12818
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0657182857142857e-05,
+      "loss": 1.0528,
+      "step": 12819
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0656914285714286e-05,
+      "loss": 1.1799,
+      "step": 12820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0656645714285715e-05,
+      "loss": 0.8895,
+      "step": 12821
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0656377142857144e-05,
+      "loss": 1.4841,
+      "step": 12822
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0656108571428572e-05,
+      "loss": 1.454,
+      "step": 12823
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0655840000000002e-05,
+      "loss": 1.3849,
+      "step": 12824
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0655571428571428e-05,
+      "loss": 1.1236,
+      "step": 12825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0655302857142858e-05,
+      "loss": 1.045,
+      "step": 12826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0655034285714286e-05,
+      "loss": 1.1615,
+      "step": 12827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0654765714285714e-05,
+      "loss": 1.4611,
+      "step": 12828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0654497142857143e-05,
+      "loss": 1.2384,
+      "step": 12829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0654228571428572e-05,
+      "loss": 1.4973,
+      "step": 12830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.065396e-05,
+      "loss": 1.2093,
+      "step": 12831
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0653691428571429e-05,
+      "loss": 1.0195,
+      "step": 12832
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0653422857142857e-05,
+      "loss": 1.574,
+      "step": 12833
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0653154285714287e-05,
+      "loss": 1.2612,
+      "step": 12834
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0652885714285715e-05,
+      "loss": 1.2696,
+      "step": 12835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0652617142857143e-05,
+      "loss": 1.0681,
+      "step": 12836
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0652348571428573e-05,
+      "loss": 1.0343,
+      "step": 12837
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.065208e-05,
+      "loss": 1.2403,
+      "step": 12838
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.065181142857143e-05,
+      "loss": 0.8762,
+      "step": 12839
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0651542857142858e-05,
+      "loss": 1.059,
+      "step": 12840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0651274285714287e-05,
+      "loss": 1.4662,
+      "step": 12841
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0651005714285714e-05,
+      "loss": 0.8816,
+      "step": 12842
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0650737142857144e-05,
+      "loss": 1.2668,
+      "step": 12843
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0650468571428572e-05,
+      "loss": 1.3196,
+      "step": 12844
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.06502e-05,
+      "loss": 1.3658,
+      "step": 12845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0649931428571428e-05,
+      "loss": 1.6113,
+      "step": 12846
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0649662857142858e-05,
+      "loss": 1.6645,
+      "step": 12847
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0649394285714286e-05,
+      "loss": 1.3502,
+      "step": 12848
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0649125714285715e-05,
+      "loss": 1.186,
+      "step": 12849
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0648857142857144e-05,
+      "loss": 1.6381,
+      "step": 12850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0648588571428571e-05,
+      "loss": 1.044,
+      "step": 12851
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.064832e-05,
+      "loss": 1.2913,
+      "step": 12852
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0648051428571429e-05,
+      "loss": 1.0849,
+      "step": 12853
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0647782857142859e-05,
+      "loss": 0.9584,
+      "step": 12854
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0647514285714285e-05,
+      "loss": 0.8049,
+      "step": 12855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0647245714285715e-05,
+      "loss": 1.5341,
+      "step": 12856
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0646977142857143e-05,
+      "loss": 1.4275,
+      "step": 12857
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0646708571428571e-05,
+      "loss": 1.0626,
+      "step": 12858
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.064644e-05,
+      "loss": 1.3581,
+      "step": 12859
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.064617142857143e-05,
+      "loss": 1.2018,
+      "step": 12860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0645902857142858e-05,
+      "loss": 0.9106,
+      "step": 12861
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.0645634285714286e-05,
+      "loss": 0.9405,
+      "step": 12862
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0645365714285716e-05,
+      "loss": 1.5268,
+      "step": 12863
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0645097142857144e-05,
+      "loss": 1.2421,
+      "step": 12864
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0644828571428572e-05,
+      "loss": 1.1523,
+      "step": 12865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.064456e-05,
+      "loss": 1.0856,
+      "step": 12866
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.064429142857143e-05,
+      "loss": 1.1226,
+      "step": 12867
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0644022857142857e-05,
+      "loss": 0.9035,
+      "step": 12868
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0643754285714287e-05,
+      "loss": 1.1915,
+      "step": 12869
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0643485714285715e-05,
+      "loss": 1.1108,
+      "step": 12870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0643217142857143e-05,
+      "loss": 1.2163,
+      "step": 12871
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0642948571428571e-05,
+      "loss": 1.0977,
+      "step": 12872
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0642680000000001e-05,
+      "loss": 0.9168,
+      "step": 12873
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0642411428571429e-05,
+      "loss": 1.229,
+      "step": 12874
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0642142857142857e-05,
+      "loss": 1.1484,
+      "step": 12875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0641874285714287e-05,
+      "loss": 1.1922,
+      "step": 12876
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0641605714285715e-05,
+      "loss": 1.3332,
+      "step": 12877
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0641337142857143e-05,
+      "loss": 0.9645,
+      "step": 12878
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0641068571428572e-05,
+      "loss": 1.0407,
+      "step": 12879
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0640800000000002e-05,
+      "loss": 1.4671,
+      "step": 12880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0640531428571428e-05,
+      "loss": 1.4479,
+      "step": 12881
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0640262857142858e-05,
+      "loss": 1.4451,
+      "step": 12882
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0639994285714286e-05,
+      "loss": 1.6999,
+      "step": 12883
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0639725714285714e-05,
+      "loss": 0.7505,
+      "step": 12884
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0639457142857142e-05,
+      "loss": 1.1396,
+      "step": 12885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0639188571428572e-05,
+      "loss": 1.0559,
+      "step": 12886
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063892e-05,
+      "loss": 1.0331,
+      "step": 12887
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0638651428571429e-05,
+      "loss": 1.4322,
+      "step": 12888
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0638382857142859e-05,
+      "loss": 1.1901,
+      "step": 12889
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0638114285714287e-05,
+      "loss": 1.4886,
+      "step": 12890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0637845714285715e-05,
+      "loss": 1.1704,
+      "step": 12891
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0637577142857143e-05,
+      "loss": 1.492,
+      "step": 12892
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0637308571428573e-05,
+      "loss": 1.4451,
+      "step": 12893
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063704e-05,
+      "loss": 0.952,
+      "step": 12894
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063677142857143e-05,
+      "loss": 1.1877,
+      "step": 12895
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0636502857142857e-05,
+      "loss": 1.0477,
+      "step": 12896
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0636234285714286e-05,
+      "loss": 1.3557,
+      "step": 12897
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0635965714285714e-05,
+      "loss": 1.064,
+      "step": 12898
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0635697142857144e-05,
+      "loss": 1.2421,
+      "step": 12899
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0635428571428572e-05,
+      "loss": 1.2336,
+      "step": 12900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063516e-05,
+      "loss": 1.2418,
+      "step": 12901
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063489142857143e-05,
+      "loss": 1.1108,
+      "step": 12902
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0634622857142858e-05,
+      "loss": 1.4953,
+      "step": 12903
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0634354285714286e-05,
+      "loss": 1.3939,
+      "step": 12904
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0634085714285714e-05,
+      "loss": 1.5428,
+      "step": 12905
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0633817142857144e-05,
+      "loss": 1.4404,
+      "step": 12906
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063354857142857e-05,
+      "loss": 1.1101,
+      "step": 12907
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063328e-05,
+      "loss": 1.2045,
+      "step": 12908
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0633011428571429e-05,
+      "loss": 1.2469,
+      "step": 12909
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0632742857142857e-05,
+      "loss": 1.2278,
+      "step": 12910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0632474285714285e-05,
+      "loss": 1.0933,
+      "step": 12911
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0632205714285715e-05,
+      "loss": 1.0269,
+      "step": 12912
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0631937142857143e-05,
+      "loss": 1.3736,
+      "step": 12913
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0631668571428571e-05,
+      "loss": 1.3045,
+      "step": 12914
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0631400000000001e-05,
+      "loss": 1.2055,
+      "step": 12915
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.063113142857143e-05,
+      "loss": 1.4621,
+      "step": 12916
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0630862857142858e-05,
+      "loss": 1.3276,
+      "step": 12917
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0630594285714286e-05,
+      "loss": 1.5007,
+      "step": 12918
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0630325714285716e-05,
+      "loss": 1.6499,
+      "step": 12919
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0630057142857142e-05,
+      "loss": 1.4638,
+      "step": 12920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0629788571428572e-05,
+      "loss": 1.2407,
+      "step": 12921
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.062952e-05,
+      "loss": 1.0406,
+      "step": 12922
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.062925142857143e-05,
+      "loss": 1.3822,
+      "step": 12923
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0628982857142857e-05,
+      "loss": 1.2713,
+      "step": 12924
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0628714285714286e-05,
+      "loss": 1.8326,
+      "step": 12925
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0628445714285715e-05,
+      "loss": 1.2895,
+      "step": 12926
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0628177142857143e-05,
+      "loss": 0.9713,
+      "step": 12927
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0627908571428573e-05,
+      "loss": 0.9841,
+      "step": 12928
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.062764e-05,
+      "loss": 1.176,
+      "step": 12929
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0627371428571429e-05,
+      "loss": 0.8657,
+      "step": 12930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0627102857142857e-05,
+      "loss": 1.0925,
+      "step": 12931
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0626834285714287e-05,
+      "loss": 0.9899,
+      "step": 12932
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0626565714285713e-05,
+      "loss": 1.4547,
+      "step": 12933
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0626297142857143e-05,
+      "loss": 1.1637,
+      "step": 12934
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0626028571428572e-05,
+      "loss": 1.0033,
+      "step": 12935
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0625760000000001e-05,
+      "loss": 1.2207,
+      "step": 12936
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0625491428571428e-05,
+      "loss": 1.4331,
+      "step": 12937
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0625222857142858e-05,
+      "loss": 1.5548,
+      "step": 12938
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0624954285714286e-05,
+      "loss": 1.467,
+      "step": 12939
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0624685714285714e-05,
+      "loss": 1.2877,
+      "step": 12940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0624417142857144e-05,
+      "loss": 0.8924,
+      "step": 12941
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0624148571428572e-05,
+      "loss": 1.4035,
+      "step": 12942
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.062388e-05,
+      "loss": 1.1872,
+      "step": 12943
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0623611428571429e-05,
+      "loss": 0.8977,
+      "step": 12944
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0623342857142858e-05,
+      "loss": 1.1615,
+      "step": 12945
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0623074285714287e-05,
+      "loss": 0.9435,
+      "step": 12946
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0622805714285715e-05,
+      "loss": 1.434,
+      "step": 12947
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0622537142857143e-05,
+      "loss": 0.886,
+      "step": 12948
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0622268571428573e-05,
+      "loss": 1.6295,
+      "step": 12949
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0622e-05,
+      "loss": 1.4751,
+      "step": 12950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0621731428571429e-05,
+      "loss": 1.2206,
+      "step": 12951
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0621462857142857e-05,
+      "loss": 1.1287,
+      "step": 12952
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0621194285714285e-05,
+      "loss": 0.9651,
+      "step": 12953
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0620925714285715e-05,
+      "loss": 1.2195,
+      "step": 12954
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0620657142857144e-05,
+      "loss": 1.1747,
+      "step": 12955
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0620388571428572e-05,
+      "loss": 1.4638,
+      "step": 12956
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.062012e-05,
+      "loss": 1.2803,
+      "step": 12957
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.061985142857143e-05,
+      "loss": 1.6817,
+      "step": 12958
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0619582857142858e-05,
+      "loss": 0.7789,
+      "step": 12959
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0619314285714286e-05,
+      "loss": 1.5056,
+      "step": 12960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0619045714285714e-05,
+      "loss": 1.0399,
+      "step": 12961
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0618777142857144e-05,
+      "loss": 1.1867,
+      "step": 12962
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.061850857142857e-05,
+      "loss": 0.8534,
+      "step": 12963
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.061824e-05,
+      "loss": 1.3454,
+      "step": 12964
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0617971428571429e-05,
+      "loss": 1.2576,
+      "step": 12965
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0617702857142857e-05,
+      "loss": 1.2518,
+      "step": 12966
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0617434285714287e-05,
+      "loss": 1.2852,
+      "step": 12967
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0617165714285715e-05,
+      "loss": 1.1036,
+      "step": 12968
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0616897142857145e-05,
+      "loss": 1.3687,
+      "step": 12969
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0616628571428571e-05,
+      "loss": 1.4147,
+      "step": 12970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0616360000000001e-05,
+      "loss": 1.6096,
+      "step": 12971
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.061609142857143e-05,
+      "loss": 1.3941,
+      "step": 12972
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0615822857142857e-05,
+      "loss": 1.4889,
+      "step": 12973
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0615554285714286e-05,
+      "loss": 0.9959,
+      "step": 12974
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0615285714285716e-05,
+      "loss": 1.3228,
+      "step": 12975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0615017142857142e-05,
+      "loss": 1.0831,
+      "step": 12976
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0614748571428572e-05,
+      "loss": 1.0733,
+      "step": 12977
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.061448e-05,
+      "loss": 1.6835,
+      "step": 12978
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0614211428571428e-05,
+      "loss": 1.346,
+      "step": 12979
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0613942857142858e-05,
+      "loss": 0.939,
+      "step": 12980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0613674285714286e-05,
+      "loss": 1.3277,
+      "step": 12981
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0613405714285716e-05,
+      "loss": 0.8797,
+      "step": 12982
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0613137142857143e-05,
+      "loss": 0.7759,
+      "step": 12983
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0612868571428573e-05,
+      "loss": 1.2098,
+      "step": 12984
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.06126e-05,
+      "loss": 1.0229,
+      "step": 12985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0612331428571429e-05,
+      "loss": 1.3015,
+      "step": 12986
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0612062857142857e-05,
+      "loss": 1.2088,
+      "step": 12987
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0611794285714287e-05,
+      "loss": 1.3052,
+      "step": 12988
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0611525714285713e-05,
+      "loss": 0.9758,
+      "step": 12989
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0611257142857143e-05,
+      "loss": 1.2149,
+      "step": 12990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0610988571428571e-05,
+      "loss": 1.1608,
+      "step": 12991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0610720000000001e-05,
+      "loss": 1.12,
+      "step": 12992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.061045142857143e-05,
+      "loss": 0.8911,
+      "step": 12993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0610182857142858e-05,
+      "loss": 0.9865,
+      "step": 12994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0609914285714288e-05,
+      "loss": 1.2495,
+      "step": 12995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0609645714285714e-05,
+      "loss": 1.0217,
+      "step": 12996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0609377142857144e-05,
+      "loss": 1.2919,
+      "step": 12997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0609108571428572e-05,
+      "loss": 1.198,
+      "step": 12998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.060884e-05,
+      "loss": 1.2926,
+      "step": 12999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0608571428571428e-05,
+      "loss": 1.3688,
+      "step": 13000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0608302857142858e-05,
+      "loss": 0.8919,
+      "step": 13001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0608034285714285e-05,
+      "loss": 1.0606,
+      "step": 13002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0607765714285715e-05,
+      "loss": 0.8306,
+      "step": 13003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0607497142857143e-05,
+      "loss": 0.9609,
+      "step": 13004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0607228571428573e-05,
+      "loss": 1.5645,
+      "step": 13005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0606960000000001e-05,
+      "loss": 0.9298,
+      "step": 13006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0606691428571429e-05,
+      "loss": 1.2035,
+      "step": 13007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0606422857142859e-05,
+      "loss": 1.0262,
+      "step": 13008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0606154285714285e-05,
+      "loss": 1.4052,
+      "step": 13009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0605885714285715e-05,
+      "loss": 1.2121,
+      "step": 13010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0605617142857143e-05,
+      "loss": 1.2582,
+      "step": 13011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0605348571428572e-05,
+      "loss": 1.0488,
+      "step": 13012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.060508e-05,
+      "loss": 1.1801,
+      "step": 13013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.060481142857143e-05,
+      "loss": 1.2022,
+      "step": 13014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0604542857142858e-05,
+      "loss": 1.7324,
+      "step": 13015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0604274285714286e-05,
+      "loss": 1.5237,
+      "step": 13016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0604005714285714e-05,
+      "loss": 0.9778,
+      "step": 13017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0603737142857144e-05,
+      "loss": 1.1268,
+      "step": 13018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.060346857142857e-05,
+      "loss": 1.4664,
+      "step": 13019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.06032e-05,
+      "loss": 1.1047,
+      "step": 13020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.060293142857143e-05,
+      "loss": 1.3665,
+      "step": 13021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0602662857142857e-05,
+      "loss": 1.6196,
+      "step": 13022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0602394285714287e-05,
+      "loss": 1.3195,
+      "step": 13023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0602125714285715e-05,
+      "loss": 1.3109,
+      "step": 13024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0601857142857143e-05,
+      "loss": 1.1779,
+      "step": 13025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0601588571428571e-05,
+      "loss": 1.2377,
+      "step": 13026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0601320000000001e-05,
+      "loss": 0.8035,
+      "step": 13027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.060105142857143e-05,
+      "loss": 0.8895,
+      "step": 13028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0600782857142857e-05,
+      "loss": 1.3944,
+      "step": 13029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0600514285714286e-05,
+      "loss": 1.0556,
+      "step": 13030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0600245714285715e-05,
+      "loss": 0.9611,
+      "step": 13031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0599977142857142e-05,
+      "loss": 1.3143,
+      "step": 13032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0599708571428572e-05,
+      "loss": 1.2276,
+      "step": 13033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0599440000000002e-05,
+      "loss": 0.9851,
+      "step": 13034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0599171428571428e-05,
+      "loss": 1.1512,
+      "step": 13035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0598902857142858e-05,
+      "loss": 1.2875,
+      "step": 13036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.0598634285714286e-05,
+      "loss": 1.5955,
+      "step": 13037
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0598365714285716e-05,
+      "loss": 1.2927,
+      "step": 13038
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0598097142857143e-05,
+      "loss": 1.116,
+      "step": 13039
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0597828571428572e-05,
+      "loss": 1.3398,
+      "step": 13040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.059756e-05,
+      "loss": 0.8863,
+      "step": 13041
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0597291428571429e-05,
+      "loss": 1.1805,
+      "step": 13042
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0597022857142857e-05,
+      "loss": 1.2995,
+      "step": 13043
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0596754285714287e-05,
+      "loss": 1.252,
+      "step": 13044
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0596485714285713e-05,
+      "loss": 0.8909,
+      "step": 13045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0596217142857143e-05,
+      "loss": 1.2561,
+      "step": 13046
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0595948571428573e-05,
+      "loss": 1.2663,
+      "step": 13047
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.059568e-05,
+      "loss": 0.9926,
+      "step": 13048
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.059541142857143e-05,
+      "loss": 0.9529,
+      "step": 13049
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0595142857142858e-05,
+      "loss": 1.0287,
+      "step": 13050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0594874285714287e-05,
+      "loss": 0.8958,
+      "step": 13051
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0594605714285714e-05,
+      "loss": 1.4648,
+      "step": 13052
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0594337142857144e-05,
+      "loss": 0.9444,
+      "step": 13053
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0594068571428572e-05,
+      "loss": 0.9102,
+      "step": 13054
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.05938e-05,
+      "loss": 0.8693,
+      "step": 13055
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0593531428571428e-05,
+      "loss": 0.8931,
+      "step": 13056
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0593262857142858e-05,
+      "loss": 1.1063,
+      "step": 13057
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0592994285714285e-05,
+      "loss": 1.622,
+      "step": 13058
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0592725714285715e-05,
+      "loss": 1.4309,
+      "step": 13059
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0592457142857144e-05,
+      "loss": 1.209,
+      "step": 13060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0592188571428571e-05,
+      "loss": 1.6621,
+      "step": 13061
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.059192e-05,
+      "loss": 1.0384,
+      "step": 13062
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0591651428571429e-05,
+      "loss": 1.1209,
+      "step": 13063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0591382857142859e-05,
+      "loss": 1.1618,
+      "step": 13064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0591114285714285e-05,
+      "loss": 1.8411,
+      "step": 13065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0590845714285715e-05,
+      "loss": 0.8616,
+      "step": 13066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0590577142857143e-05,
+      "loss": 1.1981,
+      "step": 13067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0590308571428571e-05,
+      "loss": 1.3813,
+      "step": 13068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.059004e-05,
+      "loss": 1.2945,
+      "step": 13069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.058977142857143e-05,
+      "loss": 1.5139,
+      "step": 13070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0589502857142856e-05,
+      "loss": 0.7379,
+      "step": 13071
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0589234285714286e-05,
+      "loss": 1.3749,
+      "step": 13072
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0588965714285716e-05,
+      "loss": 1.3346,
+      "step": 13073
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0588697142857144e-05,
+      "loss": 1.4172,
+      "step": 13074
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0588428571428572e-05,
+      "loss": 1.515,
+      "step": 13075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.058816e-05,
+      "loss": 1.2846,
+      "step": 13076
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.058789142857143e-05,
+      "loss": 1.009,
+      "step": 13077
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0587622857142857e-05,
+      "loss": 1.3948,
+      "step": 13078
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0587354285714287e-05,
+      "loss": 1.197,
+      "step": 13079
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0587085714285715e-05,
+      "loss": 1.3098,
+      "step": 13080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0586817142857143e-05,
+      "loss": 1.1582,
+      "step": 13081
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0586548571428571e-05,
+      "loss": 1.5808,
+      "step": 13082
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0586280000000001e-05,
+      "loss": 1.1062,
+      "step": 13083
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0586011428571427e-05,
+      "loss": 1.3658,
+      "step": 13084
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0585742857142857e-05,
+      "loss": 1.4873,
+      "step": 13085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0585474285714287e-05,
+      "loss": 1.1627,
+      "step": 13086
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0585205714285715e-05,
+      "loss": 1.6511,
+      "step": 13087
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0584937142857143e-05,
+      "loss": 1.5248,
+      "step": 13088
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0584668571428572e-05,
+      "loss": 0.9934,
+      "step": 13089
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0584400000000002e-05,
+      "loss": 1.1511,
+      "step": 13090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0584131428571428e-05,
+      "loss": 1.0838,
+      "step": 13091
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0583862857142858e-05,
+      "loss": 0.8138,
+      "step": 13092
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0583594285714286e-05,
+      "loss": 1.6236,
+      "step": 13093
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0583325714285714e-05,
+      "loss": 1.2262,
+      "step": 13094
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0583057142857142e-05,
+      "loss": 1.3973,
+      "step": 13095
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0582788571428572e-05,
+      "loss": 1.0261,
+      "step": 13096
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.058252e-05,
+      "loss": 1.4959,
+      "step": 13097
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0582251428571429e-05,
+      "loss": 1.8362,
+      "step": 13098
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0581982857142858e-05,
+      "loss": 1.1338,
+      "step": 13099
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0581714285714287e-05,
+      "loss": 1.6861,
+      "step": 13100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0581445714285715e-05,
+      "loss": 1.2562,
+      "step": 13101
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0581177142857143e-05,
+      "loss": 0.9668,
+      "step": 13102
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0580908571428573e-05,
+      "loss": 1.3107,
+      "step": 13103
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.058064e-05,
+      "loss": 0.8748,
+      "step": 13104
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.058037142857143e-05,
+      "loss": 1.244,
+      "step": 13105
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0580102857142857e-05,
+      "loss": 1.2441,
+      "step": 13106
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0579834285714286e-05,
+      "loss": 1.0272,
+      "step": 13107
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0579565714285714e-05,
+      "loss": 1.2077,
+      "step": 13108
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0579297142857144e-05,
+      "loss": 1.3118,
+      "step": 13109
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0579028571428572e-05,
+      "loss": 1.5894,
+      "step": 13110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057876e-05,
+      "loss": 1.5079,
+      "step": 13111
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057849142857143e-05,
+      "loss": 1.373,
+      "step": 13112
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0578222857142858e-05,
+      "loss": 1.4321,
+      "step": 13113
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0577954285714286e-05,
+      "loss": 1.1709,
+      "step": 13114
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0577685714285714e-05,
+      "loss": 1.1352,
+      "step": 13115
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0577417142857144e-05,
+      "loss": 1.4794,
+      "step": 13116
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057714857142857e-05,
+      "loss": 1.0828,
+      "step": 13117
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057688e-05,
+      "loss": 1.392,
+      "step": 13118
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0576611428571429e-05,
+      "loss": 1.362,
+      "step": 13119
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0576342857142859e-05,
+      "loss": 1.2326,
+      "step": 13120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0576074285714285e-05,
+      "loss": 1.1584,
+      "step": 13121
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0575805714285715e-05,
+      "loss": 1.3515,
+      "step": 13122
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0575537142857143e-05,
+      "loss": 1.3905,
+      "step": 13123
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0575268571428571e-05,
+      "loss": 1.362,
+      "step": 13124
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0575e-05,
+      "loss": 1.2867,
+      "step": 13125
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057473142857143e-05,
+      "loss": 0.9779,
+      "step": 13126
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0574462857142858e-05,
+      "loss": 1.482,
+      "step": 13127
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0574194285714286e-05,
+      "loss": 1.0056,
+      "step": 13128
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0573925714285716e-05,
+      "loss": 1.0985,
+      "step": 13129
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0573657142857142e-05,
+      "loss": 1.2662,
+      "step": 13130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0573388571428572e-05,
+      "loss": 1.2446,
+      "step": 13131
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057312e-05,
+      "loss": 1.2855,
+      "step": 13132
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057285142857143e-05,
+      "loss": 0.9042,
+      "step": 13133
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0572582857142857e-05,
+      "loss": 1.477,
+      "step": 13134
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0572314285714286e-05,
+      "loss": 1.2005,
+      "step": 13135
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0572045714285715e-05,
+      "loss": 1.3194,
+      "step": 13136
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0571777142857143e-05,
+      "loss": 1.0523,
+      "step": 13137
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0571508571428571e-05,
+      "loss": 1.1388,
+      "step": 13138
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.057124e-05,
+      "loss": 1.2526,
+      "step": 13139
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0570971428571429e-05,
+      "loss": 1.3737,
+      "step": 13140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0570702857142857e-05,
+      "loss": 1.3966,
+      "step": 13141
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0570434285714287e-05,
+      "loss": 1.1045,
+      "step": 13142
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0570165714285715e-05,
+      "loss": 1.2451,
+      "step": 13143
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0569897142857143e-05,
+      "loss": 1.3436,
+      "step": 13144
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0569628571428572e-05,
+      "loss": 1.1076,
+      "step": 13145
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0569360000000001e-05,
+      "loss": 1.5284,
+      "step": 13146
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0569091428571428e-05,
+      "loss": 1.371,
+      "step": 13147
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0568822857142858e-05,
+      "loss": 1.493,
+      "step": 13148
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0568554285714286e-05,
+      "loss": 1.218,
+      "step": 13149
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0568285714285714e-05,
+      "loss": 1.4919,
+      "step": 13150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0568017142857142e-05,
+      "loss": 0.73,
+      "step": 13151
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0567748571428572e-05,
+      "loss": 1.1197,
+      "step": 13152
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.056748e-05,
+      "loss": 1.2736,
+      "step": 13153
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0567211428571429e-05,
+      "loss": 1.4173,
+      "step": 13154
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0566942857142858e-05,
+      "loss": 1.3121,
+      "step": 13155
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0566674285714287e-05,
+      "loss": 0.8998,
+      "step": 13156
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0566405714285715e-05,
+      "loss": 1.6355,
+      "step": 13157
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0566137142857143e-05,
+      "loss": 1.6995,
+      "step": 13158
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0565868571428573e-05,
+      "loss": 0.9537,
+      "step": 13159
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.05656e-05,
+      "loss": 1.1629,
+      "step": 13160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0565331428571429e-05,
+      "loss": 1.2395,
+      "step": 13161
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0565062857142857e-05,
+      "loss": 1.2368,
+      "step": 13162
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0564794285714285e-05,
+      "loss": 1.4642,
+      "step": 13163
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0564525714285714e-05,
+      "loss": 1.3362,
+      "step": 13164
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0564257142857144e-05,
+      "loss": 1.1758,
+      "step": 13165
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0563988571428573e-05,
+      "loss": 1.2788,
+      "step": 13166
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.056372e-05,
+      "loss": 1.3645,
+      "step": 13167
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.056345142857143e-05,
+      "loss": 1.4435,
+      "step": 13168
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0563182857142858e-05,
+      "loss": 0.8695,
+      "step": 13169
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0562914285714286e-05,
+      "loss": 1.5103,
+      "step": 13170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0562645714285714e-05,
+      "loss": 1.2834,
+      "step": 13171
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0562377142857144e-05,
+      "loss": 1.3964,
+      "step": 13172
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.056210857142857e-05,
+      "loss": 1.4578,
+      "step": 13173
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.056184e-05,
+      "loss": 1.2355,
+      "step": 13174
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0561571428571429e-05,
+      "loss": 1.7433,
+      "step": 13175
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0561302857142857e-05,
+      "loss": 1.2549,
+      "step": 13176
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0561034285714285e-05,
+      "loss": 1.3637,
+      "step": 13177
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0560765714285715e-05,
+      "loss": 1.2854,
+      "step": 13178
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0560497142857145e-05,
+      "loss": 1.1439,
+      "step": 13179
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0560228571428571e-05,
+      "loss": 1.1389,
+      "step": 13180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0559960000000001e-05,
+      "loss": 1.6915,
+      "step": 13181
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.055969142857143e-05,
+      "loss": 1.362,
+      "step": 13182
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0559422857142857e-05,
+      "loss": 1.1471,
+      "step": 13183
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0559154285714286e-05,
+      "loss": 1.3072,
+      "step": 13184
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0558885714285716e-05,
+      "loss": 1.4128,
+      "step": 13185
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0558617142857142e-05,
+      "loss": 1.2454,
+      "step": 13186
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0558348571428572e-05,
+      "loss": 1.4545,
+      "step": 13187
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.055808e-05,
+      "loss": 1.3243,
+      "step": 13188
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.055781142857143e-05,
+      "loss": 1.5758,
+      "step": 13189
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0557542857142856e-05,
+      "loss": 1.0565,
+      "step": 13190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0557274285714286e-05,
+      "loss": 0.9134,
+      "step": 13191
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0557005714285716e-05,
+      "loss": 1.35,
+      "step": 13192
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0556737142857143e-05,
+      "loss": 1.0287,
+      "step": 13193
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0556468571428573e-05,
+      "loss": 1.4354,
+      "step": 13194
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.05562e-05,
+      "loss": 1.5656,
+      "step": 13195
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0555931428571429e-05,
+      "loss": 0.9777,
+      "step": 13196
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0555662857142857e-05,
+      "loss": 1.371,
+      "step": 13197
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0555394285714287e-05,
+      "loss": 1.1267,
+      "step": 13198
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0555125714285713e-05,
+      "loss": 1.3671,
+      "step": 13199
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0554857142857143e-05,
+      "loss": 1.5868,
+      "step": 13200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0554588571428571e-05,
+      "loss": 1.2235,
+      "step": 13201
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0554320000000001e-05,
+      "loss": 0.9265,
+      "step": 13202
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0554051428571428e-05,
+      "loss": 1.4708,
+      "step": 13203
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0553782857142858e-05,
+      "loss": 1.7296,
+      "step": 13204
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0553514285714288e-05,
+      "loss": 1.3717,
+      "step": 13205
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0553245714285714e-05,
+      "loss": 1.577,
+      "step": 13206
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0552977142857144e-05,
+      "loss": 1.2874,
+      "step": 13207
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0552708571428572e-05,
+      "loss": 1.2488,
+      "step": 13208
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.055244e-05,
+      "loss": 0.8913,
+      "step": 13209
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0552171428571428e-05,
+      "loss": 1.09,
+      "step": 13210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0551902857142858e-05,
+      "loss": 1.2325,
+      "step": 13211
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.0551634285714285e-05,
+      "loss": 1.258,
+      "step": 13212
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0551365714285715e-05,
+      "loss": 1.2197,
+      "step": 13213
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0551097142857143e-05,
+      "loss": 1.2463,
+      "step": 13214
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0550828571428573e-05,
+      "loss": 1.4068,
+      "step": 13215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0550559999999999e-05,
+      "loss": 0.891,
+      "step": 13216
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0550291428571429e-05,
+      "loss": 1.5267,
+      "step": 13217
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0550022857142859e-05,
+      "loss": 1.1476,
+      "step": 13218
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0549754285714285e-05,
+      "loss": 1.2615,
+      "step": 13219
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0549485714285715e-05,
+      "loss": 1.193,
+      "step": 13220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0549217142857143e-05,
+      "loss": 1.259,
+      "step": 13221
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0548948571428572e-05,
+      "loss": 1.5026,
+      "step": 13222
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.054868e-05,
+      "loss": 1.7634,
+      "step": 13223
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.054841142857143e-05,
+      "loss": 1.4059,
+      "step": 13224
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0548142857142858e-05,
+      "loss": 1.34,
+      "step": 13225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0547874285714286e-05,
+      "loss": 1.2437,
+      "step": 13226
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0547605714285714e-05,
+      "loss": 1.2836,
+      "step": 13227
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0547337142857144e-05,
+      "loss": 0.8607,
+      "step": 13228
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.054706857142857e-05,
+      "loss": 1.0444,
+      "step": 13229
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.05468e-05,
+      "loss": 1.1273,
+      "step": 13230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.054653142857143e-05,
+      "loss": 1.2854,
+      "step": 13231
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0546262857142857e-05,
+      "loss": 0.9153,
+      "step": 13232
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0545994285714287e-05,
+      "loss": 1.1973,
+      "step": 13233
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0545725714285715e-05,
+      "loss": 0.868,
+      "step": 13234
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0545457142857143e-05,
+      "loss": 1.255,
+      "step": 13235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0545188571428571e-05,
+      "loss": 1.2142,
+      "step": 13236
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0544920000000001e-05,
+      "loss": 1.1714,
+      "step": 13237
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.054465142857143e-05,
+      "loss": 1.2308,
+      "step": 13238
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0544382857142857e-05,
+      "loss": 1.1953,
+      "step": 13239
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0544114285714286e-05,
+      "loss": 1.4873,
+      "step": 13240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0543845714285715e-05,
+      "loss": 0.8208,
+      "step": 13241
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0543577142857142e-05,
+      "loss": 1.3623,
+      "step": 13242
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0543308571428572e-05,
+      "loss": 1.5147,
+      "step": 13243
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.054304e-05,
+      "loss": 1.3369,
+      "step": 13244
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0542771428571428e-05,
+      "loss": 1.0291,
+      "step": 13245
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0542502857142858e-05,
+      "loss": 1.2347,
+      "step": 13246
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0542234285714286e-05,
+      "loss": 1.2608,
+      "step": 13247
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0541965714285716e-05,
+      "loss": 0.9868,
+      "step": 13248
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0541697142857143e-05,
+      "loss": 1.2227,
+      "step": 13249
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0541428571428572e-05,
+      "loss": 0.9618,
+      "step": 13250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.054116e-05,
+      "loss": 1.0105,
+      "step": 13251
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0540891428571429e-05,
+      "loss": 1.0985,
+      "step": 13252
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0540622857142857e-05,
+      "loss": 1.1624,
+      "step": 13253
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0540354285714287e-05,
+      "loss": 1.207,
+      "step": 13254
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0540085714285713e-05,
+      "loss": 1.1166,
+      "step": 13255
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0539817142857143e-05,
+      "loss": 1.3545,
+      "step": 13256
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0539548571428571e-05,
+      "loss": 1.3124,
+      "step": 13257
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.053928e-05,
+      "loss": 1.01,
+      "step": 13258
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.053901142857143e-05,
+      "loss": 1.3148,
+      "step": 13259
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0538742857142858e-05,
+      "loss": 1.3083,
+      "step": 13260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0538474285714287e-05,
+      "loss": 1.0161,
+      "step": 13261
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0538205714285714e-05,
+      "loss": 1.3666,
+      "step": 13262
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0537937142857144e-05,
+      "loss": 0.9754,
+      "step": 13263
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0537668571428572e-05,
+      "loss": 1.2419,
+      "step": 13264
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.05374e-05,
+      "loss": 1.223,
+      "step": 13265
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0537131428571428e-05,
+      "loss": 0.8558,
+      "step": 13266
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0536862857142858e-05,
+      "loss": 1.0431,
+      "step": 13267
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0536594285714285e-05,
+      "loss": 1.6587,
+      "step": 13268
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0536325714285715e-05,
+      "loss": 1.423,
+      "step": 13269
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0536057142857143e-05,
+      "loss": 1.1189,
+      "step": 13270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0535788571428573e-05,
+      "loss": 1.1089,
+      "step": 13271
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.053552e-05,
+      "loss": 1.2385,
+      "step": 13272
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0535251428571429e-05,
+      "loss": 1.0616,
+      "step": 13273
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0534982857142859e-05,
+      "loss": 1.2601,
+      "step": 13274
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0534714285714285e-05,
+      "loss": 1.3645,
+      "step": 13275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0534445714285715e-05,
+      "loss": 0.9658,
+      "step": 13276
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0534177142857143e-05,
+      "loss": 0.8311,
+      "step": 13277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0533908571428571e-05,
+      "loss": 1.4787,
+      "step": 13278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.053364e-05,
+      "loss": 1.5571,
+      "step": 13279
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.053337142857143e-05,
+      "loss": 1.2256,
+      "step": 13280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0533102857142856e-05,
+      "loss": 1.3836,
+      "step": 13281
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0532834285714286e-05,
+      "loss": 1.5569,
+      "step": 13282
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0532565714285714e-05,
+      "loss": 1.2039,
+      "step": 13283
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0532297142857144e-05,
+      "loss": 1.5787,
+      "step": 13284
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0532028571428572e-05,
+      "loss": 1.1766,
+      "step": 13285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.053176e-05,
+      "loss": 1.3387,
+      "step": 13286
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.053149142857143e-05,
+      "loss": 1.4258,
+      "step": 13287
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0531222857142857e-05,
+      "loss": 1.5588,
+      "step": 13288
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0530954285714287e-05,
+      "loss": 1.2404,
+      "step": 13289
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0530685714285715e-05,
+      "loss": 1.1251,
+      "step": 13290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0530417142857143e-05,
+      "loss": 0.9708,
+      "step": 13291
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0530148571428571e-05,
+      "loss": 1.0814,
+      "step": 13292
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0529880000000001e-05,
+      "loss": 1.0549,
+      "step": 13293
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0529611428571429e-05,
+      "loss": 1.5394,
+      "step": 13294
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0529342857142857e-05,
+      "loss": 1.4529,
+      "step": 13295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0529074285714285e-05,
+      "loss": 0.7589,
+      "step": 13296
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0528805714285715e-05,
+      "loss": 1.4302,
+      "step": 13297
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0528537142857143e-05,
+      "loss": 1.2578,
+      "step": 13298
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0528268571428572e-05,
+      "loss": 1.0762,
+      "step": 13299
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0528000000000002e-05,
+      "loss": 0.9235,
+      "step": 13300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0527731428571428e-05,
+      "loss": 1.2957,
+      "step": 13301
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0527462857142858e-05,
+      "loss": 1.3485,
+      "step": 13302
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0527194285714286e-05,
+      "loss": 1.0824,
+      "step": 13303
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0526925714285714e-05,
+      "loss": 1.3297,
+      "step": 13304
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0526657142857142e-05,
+      "loss": 1.5471,
+      "step": 13305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0526388571428572e-05,
+      "loss": 1.1316,
+      "step": 13306
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.052612e-05,
+      "loss": 1.4716,
+      "step": 13307
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0525851428571429e-05,
+      "loss": 1.481,
+      "step": 13308
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0525582857142857e-05,
+      "loss": 1.3401,
+      "step": 13309
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0525314285714287e-05,
+      "loss": 1.0741,
+      "step": 13310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0525045714285715e-05,
+      "loss": 1.4667,
+      "step": 13311
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0524777142857143e-05,
+      "loss": 1.3881,
+      "step": 13312
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0524508571428573e-05,
+      "loss": 1.4174,
+      "step": 13313
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.052424e-05,
+      "loss": 1.1109,
+      "step": 13314
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.052397142857143e-05,
+      "loss": 1.0613,
+      "step": 13315
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0523702857142857e-05,
+      "loss": 1.1815,
+      "step": 13316
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0523434285714287e-05,
+      "loss": 1.1825,
+      "step": 13317
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0523165714285714e-05,
+      "loss": 1.5613,
+      "step": 13318
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0522897142857144e-05,
+      "loss": 1.0295,
+      "step": 13319
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0522628571428572e-05,
+      "loss": 2.0253,
+      "step": 13320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.052236e-05,
+      "loss": 1.1668,
+      "step": 13321
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0522091428571428e-05,
+      "loss": 1.2967,
+      "step": 13322
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0521822857142858e-05,
+      "loss": 1.037,
+      "step": 13323
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0521554285714286e-05,
+      "loss": 1.1065,
+      "step": 13324
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0521285714285714e-05,
+      "loss": 1.7889,
+      "step": 13325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0521017142857144e-05,
+      "loss": 1.4439,
+      "step": 13326
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.052074857142857e-05,
+      "loss": 1.1732,
+      "step": 13327
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.052048e-05,
+      "loss": 1.4115,
+      "step": 13328
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0520211428571429e-05,
+      "loss": 1.3574,
+      "step": 13329
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0519942857142859e-05,
+      "loss": 1.0362,
+      "step": 13330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0519674285714285e-05,
+      "loss": 1.4197,
+      "step": 13331
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0519405714285715e-05,
+      "loss": 1.2333,
+      "step": 13332
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0519137142857143e-05,
+      "loss": 1.1516,
+      "step": 13333
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0518868571428571e-05,
+      "loss": 1.186,
+      "step": 13334
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.05186e-05,
+      "loss": 1.2834,
+      "step": 13335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.051833142857143e-05,
+      "loss": 1.4136,
+      "step": 13336
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0518062857142858e-05,
+      "loss": 1.1111,
+      "step": 13337
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0517794285714286e-05,
+      "loss": 1.0325,
+      "step": 13338
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0517525714285716e-05,
+      "loss": 0.9502,
+      "step": 13339
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0517257142857144e-05,
+      "loss": 1.4264,
+      "step": 13340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0516988571428572e-05,
+      "loss": 1.0285,
+      "step": 13341
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.051672e-05,
+      "loss": 1.6021,
+      "step": 13342
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.051645142857143e-05,
+      "loss": 1.0139,
+      "step": 13343
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0516182857142857e-05,
+      "loss": 1.3958,
+      "step": 13344
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0515914285714286e-05,
+      "loss": 0.9989,
+      "step": 13345
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0515645714285715e-05,
+      "loss": 1.2473,
+      "step": 13346
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0515377142857143e-05,
+      "loss": 1.0618,
+      "step": 13347
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0515108571428571e-05,
+      "loss": 1.0107,
+      "step": 13348
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.051484e-05,
+      "loss": 1.0198,
+      "step": 13349
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0514571428571429e-05,
+      "loss": 1.3532,
+      "step": 13350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0514302857142857e-05,
+      "loss": 1.624,
+      "step": 13351
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0514034285714287e-05,
+      "loss": 1.2402,
+      "step": 13352
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0513765714285715e-05,
+      "loss": 1.0829,
+      "step": 13353
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0513497142857143e-05,
+      "loss": 1.0056,
+      "step": 13354
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0513228571428572e-05,
+      "loss": 1.3771,
+      "step": 13355
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0512960000000001e-05,
+      "loss": 1.2341,
+      "step": 13356
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0512691428571428e-05,
+      "loss": 1.347,
+      "step": 13357
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0512422857142858e-05,
+      "loss": 0.8832,
+      "step": 13358
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0512154285714286e-05,
+      "loss": 1.3775,
+      "step": 13359
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0511885714285714e-05,
+      "loss": 1.1462,
+      "step": 13360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0511617142857142e-05,
+      "loss": 0.808,
+      "step": 13361
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0511348571428572e-05,
+      "loss": 1.4712,
+      "step": 13362
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.051108e-05,
+      "loss": 1.3175,
+      "step": 13363
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0510811428571429e-05,
+      "loss": 0.8202,
+      "step": 13364
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0510542857142858e-05,
+      "loss": 1.5935,
+      "step": 13365
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0510274285714287e-05,
+      "loss": 1.4011,
+      "step": 13366
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0510005714285715e-05,
+      "loss": 0.9214,
+      "step": 13367
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0509737142857143e-05,
+      "loss": 1.2642,
+      "step": 13368
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0509468571428573e-05,
+      "loss": 0.7728,
+      "step": 13369
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.05092e-05,
+      "loss": 1.0524,
+      "step": 13370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0508931428571429e-05,
+      "loss": 1.2965,
+      "step": 13371
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0508662857142857e-05,
+      "loss": 1.7093,
+      "step": 13372
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0508394285714285e-05,
+      "loss": 0.6782,
+      "step": 13373
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0508125714285714e-05,
+      "loss": 1.0114,
+      "step": 13374
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0507857142857144e-05,
+      "loss": 1.417,
+      "step": 13375
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0507588571428572e-05,
+      "loss": 1.0794,
+      "step": 13376
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.050732e-05,
+      "loss": 1.6045,
+      "step": 13377
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.050705142857143e-05,
+      "loss": 1.1313,
+      "step": 13378
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0506782857142858e-05,
+      "loss": 1.1514,
+      "step": 13379
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0506514285714286e-05,
+      "loss": 1.446,
+      "step": 13380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0506245714285714e-05,
+      "loss": 1.4599,
+      "step": 13381
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0505977142857144e-05,
+      "loss": 1.1485,
+      "step": 13382
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.050570857142857e-05,
+      "loss": 1.2216,
+      "step": 13383
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.050544e-05,
+      "loss": 1.2492,
+      "step": 13384
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0505171428571429e-05,
+      "loss": 0.9948,
+      "step": 13385
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0504902857142857e-05,
+      "loss": 0.9555,
+      "step": 13386
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.0504634285714285e-05,
+      "loss": 0.9674,
+      "step": 13387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0504365714285715e-05,
+      "loss": 1.2608,
+      "step": 13388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0504097142857143e-05,
+      "loss": 1.3057,
+      "step": 13389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0503828571428571e-05,
+      "loss": 0.9932,
+      "step": 13390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0503560000000001e-05,
+      "loss": 1.2876,
+      "step": 13391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.050329142857143e-05,
+      "loss": 1.0665,
+      "step": 13392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0503022857142857e-05,
+      "loss": 1.1854,
+      "step": 13393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0502754285714286e-05,
+      "loss": 1.4949,
+      "step": 13394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0502485714285716e-05,
+      "loss": 1.5251,
+      "step": 13395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0502217142857142e-05,
+      "loss": 0.9975,
+      "step": 13396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0501948571428572e-05,
+      "loss": 1.3466,
+      "step": 13397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.050168e-05,
+      "loss": 0.9323,
+      "step": 13398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.050141142857143e-05,
+      "loss": 1.0893,
+      "step": 13399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0501142857142856e-05,
+      "loss": 1.4386,
+      "step": 13400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0500874285714286e-05,
+      "loss": 1.0417,
+      "step": 13401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0500605714285714e-05,
+      "loss": 1.502,
+      "step": 13402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0500337142857143e-05,
+      "loss": 1.2762,
+      "step": 13403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0500068571428572e-05,
+      "loss": 0.9468,
+      "step": 13404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.04998e-05,
+      "loss": 1.0882,
+      "step": 13405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0499531428571429e-05,
+      "loss": 1.1547,
+      "step": 13406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0499262857142857e-05,
+      "loss": 1.5091,
+      "step": 13407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0498994285714287e-05,
+      "loss": 1.1652,
+      "step": 13408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0498725714285713e-05,
+      "loss": 1.1787,
+      "step": 13409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0498457142857143e-05,
+      "loss": 1.3116,
+      "step": 13410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0498188571428571e-05,
+      "loss": 1.6084,
+      "step": 13411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0497920000000001e-05,
+      "loss": 0.9382,
+      "step": 13412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0497651428571428e-05,
+      "loss": 1.2312,
+      "step": 13413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0497382857142858e-05,
+      "loss": 1.477,
+      "step": 13414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0497114285714286e-05,
+      "loss": 1.0118,
+      "step": 13415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0496845714285714e-05,
+      "loss": 1.4232,
+      "step": 13416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0496577142857144e-05,
+      "loss": 0.7752,
+      "step": 13417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0496308571428572e-05,
+      "loss": 1.3063,
+      "step": 13418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.049604e-05,
+      "loss": 1.141,
+      "step": 13419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0495771428571428e-05,
+      "loss": 1.6134,
+      "step": 13420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0495502857142858e-05,
+      "loss": 1.3441,
+      "step": 13421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0495234285714286e-05,
+      "loss": 1.0725,
+      "step": 13422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0494965714285715e-05,
+      "loss": 1.4226,
+      "step": 13423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0494697142857143e-05,
+      "loss": 1.203,
+      "step": 13424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0494428571428573e-05,
+      "loss": 1.2484,
+      "step": 13425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0494159999999999e-05,
+      "loss": 1.3989,
+      "step": 13426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0493891428571429e-05,
+      "loss": 0.8565,
+      "step": 13427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0493622857142857e-05,
+      "loss": 1.1967,
+      "step": 13428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0493354285714285e-05,
+      "loss": 1.1001,
+      "step": 13429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0493085714285715e-05,
+      "loss": 1.4172,
+      "step": 13430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0492817142857143e-05,
+      "loss": 1.2072,
+      "step": 13431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0492548571428572e-05,
+      "loss": 1.4512,
+      "step": 13432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.049228e-05,
+      "loss": 0.9812,
+      "step": 13433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.049201142857143e-05,
+      "loss": 0.8762,
+      "step": 13434
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0491742857142858e-05,
+      "loss": 1.2127,
+      "step": 13435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0491474285714286e-05,
+      "loss": 0.9458,
+      "step": 13436
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0491205714285714e-05,
+      "loss": 1.5816,
+      "step": 13437
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0490937142857144e-05,
+      "loss": 1.2246,
+      "step": 13438
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.049066857142857e-05,
+      "loss": 1.0788,
+      "step": 13439
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.04904e-05,
+      "loss": 1.0247,
+      "step": 13440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0490131428571429e-05,
+      "loss": 1.2709,
+      "step": 13441
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0489862857142857e-05,
+      "loss": 1.3893,
+      "step": 13442
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0489594285714287e-05,
+      "loss": 1.3552,
+      "step": 13443
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0489325714285715e-05,
+      "loss": 1.3177,
+      "step": 13444
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0489057142857145e-05,
+      "loss": 1.35,
+      "step": 13445
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0488788571428571e-05,
+      "loss": 1.3015,
+      "step": 13446
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0488520000000001e-05,
+      "loss": 1.2357,
+      "step": 13447
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.048825142857143e-05,
+      "loss": 0.9072,
+      "step": 13448
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0487982857142857e-05,
+      "loss": 1.0059,
+      "step": 13449
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0487714285714286e-05,
+      "loss": 1.3037,
+      "step": 13450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0487445714285715e-05,
+      "loss": 1.4655,
+      "step": 13451
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0487177142857142e-05,
+      "loss": 1.2466,
+      "step": 13452
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0486908571428572e-05,
+      "loss": 1.3217,
+      "step": 13453
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.048664e-05,
+      "loss": 1.2344,
+      "step": 13454
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0486371428571428e-05,
+      "loss": 1.6071,
+      "step": 13455
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0486102857142858e-05,
+      "loss": 1.0883,
+      "step": 13456
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0485834285714286e-05,
+      "loss": 0.9019,
+      "step": 13457
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0485565714285716e-05,
+      "loss": 0.6711,
+      "step": 13458
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0485297142857143e-05,
+      "loss": 1.2032,
+      "step": 13459
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0485028571428572e-05,
+      "loss": 1.4617,
+      "step": 13460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.048476e-05,
+      "loss": 1.3606,
+      "step": 13461
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0484491428571429e-05,
+      "loss": 1.4221,
+      "step": 13462
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0484222857142857e-05,
+      "loss": 1.2309,
+      "step": 13463
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0483954285714287e-05,
+      "loss": 1.1974,
+      "step": 13464
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0483685714285713e-05,
+      "loss": 1.2169,
+      "step": 13465
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0483417142857143e-05,
+      "loss": 1.5148,
+      "step": 13466
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0483148571428571e-05,
+      "loss": 1.0397,
+      "step": 13467
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0482880000000001e-05,
+      "loss": 1.1079,
+      "step": 13468
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.048261142857143e-05,
+      "loss": 1.2454,
+      "step": 13469
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0482342857142858e-05,
+      "loss": 0.9499,
+      "step": 13470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0482074285714287e-05,
+      "loss": 1.3498,
+      "step": 13471
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0481805714285714e-05,
+      "loss": 0.9529,
+      "step": 13472
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0481537142857144e-05,
+      "loss": 1.0697,
+      "step": 13473
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0481268571428572e-05,
+      "loss": 1.1732,
+      "step": 13474
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0481e-05,
+      "loss": 1.4644,
+      "step": 13475
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0480731428571428e-05,
+      "loss": 1.0788,
+      "step": 13476
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0480462857142858e-05,
+      "loss": 1.1309,
+      "step": 13477
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0480194285714285e-05,
+      "loss": 1.4414,
+      "step": 13478
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0479925714285715e-05,
+      "loss": 1.4168,
+      "step": 13479
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0479657142857143e-05,
+      "loss": 0.9842,
+      "step": 13480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0479388571428573e-05,
+      "loss": 0.8303,
+      "step": 13481
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.047912e-05,
+      "loss": 1.5875,
+      "step": 13482
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0478851428571429e-05,
+      "loss": 1.2416,
+      "step": 13483
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0478582857142859e-05,
+      "loss": 1.15,
+      "step": 13484
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0478314285714285e-05,
+      "loss": 1.2609,
+      "step": 13485
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0478045714285715e-05,
+      "loss": 1.602,
+      "step": 13486
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0477777142857143e-05,
+      "loss": 1.3621,
+      "step": 13487
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0477508571428571e-05,
+      "loss": 1.0302,
+      "step": 13488
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.047724e-05,
+      "loss": 1.6009,
+      "step": 13489
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.047697142857143e-05,
+      "loss": 1.1454,
+      "step": 13490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0476702857142858e-05,
+      "loss": 1.2738,
+      "step": 13491
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0476434285714286e-05,
+      "loss": 1.6238,
+      "step": 13492
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0476165714285714e-05,
+      "loss": 1.8275,
+      "step": 13493
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0475897142857144e-05,
+      "loss": 1.4768,
+      "step": 13494
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0475628571428572e-05,
+      "loss": 1.6625,
+      "step": 13495
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.047536e-05,
+      "loss": 0.9185,
+      "step": 13496
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.047509142857143e-05,
+      "loss": 1.0101,
+      "step": 13497
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0474822857142857e-05,
+      "loss": 1.0857,
+      "step": 13498
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0474554285714287e-05,
+      "loss": 1.3303,
+      "step": 13499
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0474285714285715e-05,
+      "loss": 1.2344,
+      "step": 13500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0474017142857143e-05,
+      "loss": 1.158,
+      "step": 13501
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0473748571428571e-05,
+      "loss": 1.3479,
+      "step": 13502
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0473480000000001e-05,
+      "loss": 1.188,
+      "step": 13503
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0473211428571429e-05,
+      "loss": 1.4685,
+      "step": 13504
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0472942857142857e-05,
+      "loss": 1.1998,
+      "step": 13505
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0472674285714285e-05,
+      "loss": 1.2592,
+      "step": 13506
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0472405714285715e-05,
+      "loss": 1.1219,
+      "step": 13507
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0472137142857143e-05,
+      "loss": 1.2942,
+      "step": 13508
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0471868571428572e-05,
+      "loss": 1.4141,
+      "step": 13509
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0471600000000002e-05,
+      "loss": 1.4817,
+      "step": 13510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0471331428571428e-05,
+      "loss": 1.3508,
+      "step": 13511
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0471062857142858e-05,
+      "loss": 1.0399,
+      "step": 13512
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0470794285714286e-05,
+      "loss": 1.3811,
+      "step": 13513
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0470525714285716e-05,
+      "loss": 1.5865,
+      "step": 13514
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0470257142857142e-05,
+      "loss": 1.1345,
+      "step": 13515
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0469988571428572e-05,
+      "loss": 1.4545,
+      "step": 13516
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046972e-05,
+      "loss": 1.3142,
+      "step": 13517
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0469451428571429e-05,
+      "loss": 1.0607,
+      "step": 13518
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0469182857142857e-05,
+      "loss": 1.4591,
+      "step": 13519
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0468914285714287e-05,
+      "loss": 1.574,
+      "step": 13520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0468645714285715e-05,
+      "loss": 1.4411,
+      "step": 13521
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0468377142857143e-05,
+      "loss": 1.3895,
+      "step": 13522
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0468108571428573e-05,
+      "loss": 1.1033,
+      "step": 13523
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046784e-05,
+      "loss": 1.307,
+      "step": 13524
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046757142857143e-05,
+      "loss": 0.8485,
+      "step": 13525
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0467302857142857e-05,
+      "loss": 1.2362,
+      "step": 13526
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0467034285714287e-05,
+      "loss": 1.2218,
+      "step": 13527
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0466765714285714e-05,
+      "loss": 1.6261,
+      "step": 13528
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0466497142857144e-05,
+      "loss": 0.999,
+      "step": 13529
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0466228571428572e-05,
+      "loss": 1.5074,
+      "step": 13530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046596e-05,
+      "loss": 1.1808,
+      "step": 13531
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0465691428571428e-05,
+      "loss": 1.4291,
+      "step": 13532
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0465422857142858e-05,
+      "loss": 0.8775,
+      "step": 13533
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0465154285714286e-05,
+      "loss": 0.7585,
+      "step": 13534
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0464885714285714e-05,
+      "loss": 1.0138,
+      "step": 13535
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0464617142857144e-05,
+      "loss": 0.7643,
+      "step": 13536
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0464348571428572e-05,
+      "loss": 1.3543,
+      "step": 13537
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046408e-05,
+      "loss": 1.075,
+      "step": 13538
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0463811428571429e-05,
+      "loss": 1.2504,
+      "step": 13539
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0463542857142859e-05,
+      "loss": 1.1079,
+      "step": 13540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0463274285714285e-05,
+      "loss": 1.3073,
+      "step": 13541
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0463005714285715e-05,
+      "loss": 1.3554,
+      "step": 13542
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0462737142857143e-05,
+      "loss": 1.197,
+      "step": 13543
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0462468571428571e-05,
+      "loss": 1.2984,
+      "step": 13544
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.04622e-05,
+      "loss": 1.4943,
+      "step": 13545
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046193142857143e-05,
+      "loss": 1.6705,
+      "step": 13546
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0461662857142858e-05,
+      "loss": 0.8896,
+      "step": 13547
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0461394285714286e-05,
+      "loss": 0.9569,
+      "step": 13548
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0461125714285716e-05,
+      "loss": 1.5671,
+      "step": 13549
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0460857142857144e-05,
+      "loss": 1.2471,
+      "step": 13550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0460588571428572e-05,
+      "loss": 1.0552,
+      "step": 13551
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046032e-05,
+      "loss": 1.5759,
+      "step": 13552
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.046005142857143e-05,
+      "loss": 1.0833,
+      "step": 13553
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0459782857142857e-05,
+      "loss": 1.6849,
+      "step": 13554
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0459514285714286e-05,
+      "loss": 1.4169,
+      "step": 13555
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0459245714285715e-05,
+      "loss": 0.9198,
+      "step": 13556
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0458977142857143e-05,
+      "loss": 0.9335,
+      "step": 13557
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0458708571428571e-05,
+      "loss": 0.8699,
+      "step": 13558
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.045844e-05,
+      "loss": 1.1605,
+      "step": 13559
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0458171428571429e-05,
+      "loss": 1.0432,
+      "step": 13560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0457902857142857e-05,
+      "loss": 0.9943,
+      "step": 13561
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.0457634285714287e-05,
+      "loss": 1.3828,
+      "step": 13562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0457365714285715e-05,
+      "loss": 1.3316,
+      "step": 13563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0457097142857143e-05,
+      "loss": 0.982,
+      "step": 13564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0456828571428572e-05,
+      "loss": 1.5864,
+      "step": 13565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0456560000000001e-05,
+      "loss": 1.0023,
+      "step": 13566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0456291428571428e-05,
+      "loss": 0.9498,
+      "step": 13567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0456022857142858e-05,
+      "loss": 0.8367,
+      "step": 13568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0455754285714286e-05,
+      "loss": 1.1534,
+      "step": 13569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0455485714285714e-05,
+      "loss": 1.3421,
+      "step": 13570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0455217142857142e-05,
+      "loss": 1.2715,
+      "step": 13571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0454948571428572e-05,
+      "loss": 1.3412,
+      "step": 13572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.045468e-05,
+      "loss": 1.3241,
+      "step": 13573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0454411428571429e-05,
+      "loss": 1.1342,
+      "step": 13574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0454142857142858e-05,
+      "loss": 1.1562,
+      "step": 13575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0453874285714287e-05,
+      "loss": 1.4809,
+      "step": 13576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0453605714285715e-05,
+      "loss": 1.2872,
+      "step": 13577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0453337142857143e-05,
+      "loss": 1.2079,
+      "step": 13578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0453068571428573e-05,
+      "loss": 1.5391,
+      "step": 13579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.04528e-05,
+      "loss": 1.5527,
+      "step": 13580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0452531428571429e-05,
+      "loss": 1.2002,
+      "step": 13581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0452262857142857e-05,
+      "loss": 1.2485,
+      "step": 13582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0451994285714285e-05,
+      "loss": 0.8785,
+      "step": 13583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0451725714285714e-05,
+      "loss": 1.0939,
+      "step": 13584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0451457142857144e-05,
+      "loss": 1.5017,
+      "step": 13585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0451188571428572e-05,
+      "loss": 1.1408,
+      "step": 13586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.045092e-05,
+      "loss": 0.9251,
+      "step": 13587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.045065142857143e-05,
+      "loss": 0.7392,
+      "step": 13588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0450382857142858e-05,
+      "loss": 1.3215,
+      "step": 13589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0450114285714286e-05,
+      "loss": 0.9908,
+      "step": 13590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0449845714285714e-05,
+      "loss": 1.1908,
+      "step": 13591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0449577142857144e-05,
+      "loss": 0.8321,
+      "step": 13592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.044930857142857e-05,
+      "loss": 1.257,
+      "step": 13593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.044904e-05,
+      "loss": 1.4217,
+      "step": 13594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0448771428571429e-05,
+      "loss": 0.8778,
+      "step": 13595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0448502857142859e-05,
+      "loss": 1.2648,
+      "step": 13596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0448234285714285e-05,
+      "loss": 1.26,
+      "step": 13597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0447965714285715e-05,
+      "loss": 0.8564,
+      "step": 13598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0447697142857143e-05,
+      "loss": 0.8139,
+      "step": 13599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0447428571428571e-05,
+      "loss": 1.3466,
+      "step": 13600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0447160000000001e-05,
+      "loss": 1.1421,
+      "step": 13601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.044689142857143e-05,
+      "loss": 1.1234,
+      "step": 13602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0446622857142857e-05,
+      "loss": 1.4075,
+      "step": 13603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0446354285714286e-05,
+      "loss": 1.0885,
+      "step": 13604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0446085714285716e-05,
+      "loss": 0.6114,
+      "step": 13605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0445817142857142e-05,
+      "loss": 0.7256,
+      "step": 13606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0445548571428572e-05,
+      "loss": 1.4718,
+      "step": 13607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.044528e-05,
+      "loss": 1.1689,
+      "step": 13608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.044501142857143e-05,
+      "loss": 1.0944,
+      "step": 13609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0444742857142856e-05,
+      "loss": 1.4121,
+      "step": 13610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0444474285714286e-05,
+      "loss": 1.112,
+      "step": 13611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0444205714285714e-05,
+      "loss": 1.229,
+      "step": 13612
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0443937142857143e-05,
+      "loss": 0.8505,
+      "step": 13613
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0443668571428572e-05,
+      "loss": 1.3786,
+      "step": 13614
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.04434e-05,
+      "loss": 1.6294,
+      "step": 13615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0443131428571429e-05,
+      "loss": 1.5158,
+      "step": 13616
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0442862857142857e-05,
+      "loss": 1.0579,
+      "step": 13617
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0442594285714287e-05,
+      "loss": 1.4251,
+      "step": 13618
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0442325714285715e-05,
+      "loss": 1.3979,
+      "step": 13619
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0442057142857143e-05,
+      "loss": 0.7127,
+      "step": 13620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0441788571428571e-05,
+      "loss": 0.9704,
+      "step": 13621
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0441520000000001e-05,
+      "loss": 1.3806,
+      "step": 13622
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0441251428571428e-05,
+      "loss": 1.231,
+      "step": 13623
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0440982857142858e-05,
+      "loss": 0.9144,
+      "step": 13624
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0440714285714286e-05,
+      "loss": 1.0259,
+      "step": 13625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0440445714285714e-05,
+      "loss": 1.1361,
+      "step": 13626
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0440177142857144e-05,
+      "loss": 1.0661,
+      "step": 13627
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0439908571428572e-05,
+      "loss": 1.3735,
+      "step": 13628
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.043964e-05,
+      "loss": 1.1756,
+      "step": 13629
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0439371428571428e-05,
+      "loss": 1.9708,
+      "step": 13630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0439102857142858e-05,
+      "loss": 0.9394,
+      "step": 13631
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0438834285714286e-05,
+      "loss": 1.2165,
+      "step": 13632
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0438565714285715e-05,
+      "loss": 1.3225,
+      "step": 13633
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0438297142857143e-05,
+      "loss": 1.1903,
+      "step": 13634
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0438028571428573e-05,
+      "loss": 1.0693,
+      "step": 13635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0437759999999999e-05,
+      "loss": 1.5382,
+      "step": 13636
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0437491428571429e-05,
+      "loss": 1.1798,
+      "step": 13637
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0437222857142857e-05,
+      "loss": 1.1083,
+      "step": 13638
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0436954285714285e-05,
+      "loss": 1.3937,
+      "step": 13639
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0436685714285715e-05,
+      "loss": 1.0425,
+      "step": 13640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0436417142857143e-05,
+      "loss": 1.1043,
+      "step": 13641
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0436148571428573e-05,
+      "loss": 1.5262,
+      "step": 13642
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.043588e-05,
+      "loss": 1.3735,
+      "step": 13643
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.043561142857143e-05,
+      "loss": 1.1516,
+      "step": 13644
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0435342857142858e-05,
+      "loss": 0.7985,
+      "step": 13645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0435074285714286e-05,
+      "loss": 1.3143,
+      "step": 13646
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0434805714285714e-05,
+      "loss": 1.3533,
+      "step": 13647
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0434537142857144e-05,
+      "loss": 1.4988,
+      "step": 13648
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.043426857142857e-05,
+      "loss": 0.7833,
+      "step": 13649
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0434e-05,
+      "loss": 1.3586,
+      "step": 13650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0433731428571429e-05,
+      "loss": 1.4918,
+      "step": 13651
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0433462857142857e-05,
+      "loss": 1.2919,
+      "step": 13652
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0433194285714287e-05,
+      "loss": 1.4746,
+      "step": 13653
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0432925714285715e-05,
+      "loss": 1.5924,
+      "step": 13654
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0432657142857145e-05,
+      "loss": 1.2007,
+      "step": 13655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0432388571428571e-05,
+      "loss": 1.2952,
+      "step": 13656
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0432120000000001e-05,
+      "loss": 1.2089,
+      "step": 13657
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.043185142857143e-05,
+      "loss": 1.2123,
+      "step": 13658
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0431582857142857e-05,
+      "loss": 1.1052,
+      "step": 13659
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0431314285714286e-05,
+      "loss": 1.3253,
+      "step": 13660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0431045714285715e-05,
+      "loss": 1.3731,
+      "step": 13661
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0430777142857142e-05,
+      "loss": 1.376,
+      "step": 13662
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0430508571428572e-05,
+      "loss": 1.3661,
+      "step": 13663
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.043024e-05,
+      "loss": 1.2749,
+      "step": 13664
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.042997142857143e-05,
+      "loss": 1.0746,
+      "step": 13665
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0429702857142858e-05,
+      "loss": 1.4687,
+      "step": 13666
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0429434285714286e-05,
+      "loss": 1.6663,
+      "step": 13667
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0429165714285716e-05,
+      "loss": 1.5883,
+      "step": 13668
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0428897142857143e-05,
+      "loss": 1.1409,
+      "step": 13669
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0428628571428572e-05,
+      "loss": 1.3305,
+      "step": 13670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.042836e-05,
+      "loss": 1.1179,
+      "step": 13671
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0428091428571429e-05,
+      "loss": 1.2476,
+      "step": 13672
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0427822857142857e-05,
+      "loss": 1.329,
+      "step": 13673
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0427554285714287e-05,
+      "loss": 1.7009,
+      "step": 13674
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0427285714285713e-05,
+      "loss": 1.3055,
+      "step": 13675
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0427017142857143e-05,
+      "loss": 1.3691,
+      "step": 13676
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0426748571428571e-05,
+      "loss": 1.43,
+      "step": 13677
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0426480000000001e-05,
+      "loss": 1.3056,
+      "step": 13678
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.042621142857143e-05,
+      "loss": 1.3121,
+      "step": 13679
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0425942857142858e-05,
+      "loss": 1.8753,
+      "step": 13680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0425674285714287e-05,
+      "loss": 1.0197,
+      "step": 13681
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0425405714285714e-05,
+      "loss": 1.1656,
+      "step": 13682
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0425137142857144e-05,
+      "loss": 1.1464,
+      "step": 13683
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0424868571428572e-05,
+      "loss": 1.4386,
+      "step": 13684
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.04246e-05,
+      "loss": 1.5791,
+      "step": 13685
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0424331428571428e-05,
+      "loss": 1.1743,
+      "step": 13686
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0424062857142858e-05,
+      "loss": 0.8218,
+      "step": 13687
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0423794285714286e-05,
+      "loss": 1.1837,
+      "step": 13688
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0423525714285715e-05,
+      "loss": 1.2834,
+      "step": 13689
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0423257142857143e-05,
+      "loss": 1.2404,
+      "step": 13690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0422988571428573e-05,
+      "loss": 1.2612,
+      "step": 13691
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.042272e-05,
+      "loss": 1.4268,
+      "step": 13692
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0422451428571429e-05,
+      "loss": 1.3192,
+      "step": 13693
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0422182857142859e-05,
+      "loss": 1.0489,
+      "step": 13694
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0421914285714285e-05,
+      "loss": 1.074,
+      "step": 13695
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0421645714285715e-05,
+      "loss": 1.1757,
+      "step": 13696
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0421377142857143e-05,
+      "loss": 1.3293,
+      "step": 13697
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0421108571428571e-05,
+      "loss": 1.4488,
+      "step": 13698
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.042084e-05,
+      "loss": 1.122,
+      "step": 13699
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.042057142857143e-05,
+      "loss": 1.4204,
+      "step": 13700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0420302857142858e-05,
+      "loss": 1.2555,
+      "step": 13701
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0420034285714286e-05,
+      "loss": 1.1064,
+      "step": 13702
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0419765714285714e-05,
+      "loss": 1.0256,
+      "step": 13703
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0419497142857144e-05,
+      "loss": 1.4481,
+      "step": 13704
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0419228571428572e-05,
+      "loss": 1.3511,
+      "step": 13705
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.041896e-05,
+      "loss": 1.0359,
+      "step": 13706
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.041869142857143e-05,
+      "loss": 1.4057,
+      "step": 13707
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0418422857142857e-05,
+      "loss": 1.2671,
+      "step": 13708
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0418154285714286e-05,
+      "loss": 1.37,
+      "step": 13709
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0417885714285715e-05,
+      "loss": 1.3851,
+      "step": 13710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0417617142857143e-05,
+      "loss": 1.3329,
+      "step": 13711
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0417348571428571e-05,
+      "loss": 1.0309,
+      "step": 13712
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0417080000000001e-05,
+      "loss": 0.9781,
+      "step": 13713
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0416811428571429e-05,
+      "loss": 1.1366,
+      "step": 13714
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0416542857142857e-05,
+      "loss": 0.9477,
+      "step": 13715
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0416274285714285e-05,
+      "loss": 1.0737,
+      "step": 13716
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0416005714285715e-05,
+      "loss": 1.4088,
+      "step": 13717
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0415737142857143e-05,
+      "loss": 1.7296,
+      "step": 13718
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0415468571428572e-05,
+      "loss": 1.3329,
+      "step": 13719
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0415200000000002e-05,
+      "loss": 1.1173,
+      "step": 13720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0414931428571428e-05,
+      "loss": 1.0814,
+      "step": 13721
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0414662857142858e-05,
+      "loss": 1.4092,
+      "step": 13722
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0414394285714286e-05,
+      "loss": 0.7625,
+      "step": 13723
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0414125714285716e-05,
+      "loss": 1.6577,
+      "step": 13724
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0413857142857142e-05,
+      "loss": 1.4386,
+      "step": 13725
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0413588571428572e-05,
+      "loss": 1.0597,
+      "step": 13726
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.041332e-05,
+      "loss": 1.4041,
+      "step": 13727
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0413051428571429e-05,
+      "loss": 1.0337,
+      "step": 13728
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0412782857142857e-05,
+      "loss": 1.4216,
+      "step": 13729
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0412514285714287e-05,
+      "loss": 1.3124,
+      "step": 13730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0412245714285715e-05,
+      "loss": 1.1934,
+      "step": 13731
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0411977142857143e-05,
+      "loss": 1.3008,
+      "step": 13732
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0411708571428573e-05,
+      "loss": 1.089,
+      "step": 13733
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.041144e-05,
+      "loss": 1.1039,
+      "step": 13734
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.041117142857143e-05,
+      "loss": 1.5502,
+      "step": 13735
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0410902857142857e-05,
+      "loss": 1.3877,
+      "step": 13736
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0410634285714287e-05,
+      "loss": 1.4664,
+      "step": 13737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0410365714285714e-05,
+      "loss": 1.1665,
+      "step": 13738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0410097142857144e-05,
+      "loss": 1.0563,
+      "step": 13739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0409828571428572e-05,
+      "loss": 1.4138,
+      "step": 13740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.040956e-05,
+      "loss": 1.3905,
+      "step": 13741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0409291428571428e-05,
+      "loss": 1.5566,
+      "step": 13742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0409022857142858e-05,
+      "loss": 1.1483,
+      "step": 13743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0408754285714286e-05,
+      "loss": 1.0689,
+      "step": 13744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0408485714285714e-05,
+      "loss": 1.3,
+      "step": 13745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0408217142857144e-05,
+      "loss": 0.9596,
+      "step": 13746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0407948571428572e-05,
+      "loss": 0.8295,
+      "step": 13747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.040768e-05,
+      "loss": 1.6196,
+      "step": 13748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0407411428571429e-05,
+      "loss": 1.6542,
+      "step": 13749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0407142857142859e-05,
+      "loss": 1.2517,
+      "step": 13750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0406874285714285e-05,
+      "loss": 1.3361,
+      "step": 13751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0406605714285715e-05,
+      "loss": 1.2291,
+      "step": 13752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0406337142857143e-05,
+      "loss": 1.044,
+      "step": 13753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0406068571428571e-05,
+      "loss": 1.397,
+      "step": 13754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.04058e-05,
+      "loss": 0.957,
+      "step": 13755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.040553142857143e-05,
+      "loss": 0.894,
+      "step": 13756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0405262857142858e-05,
+      "loss": 0.9617,
+      "step": 13757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0404994285714286e-05,
+      "loss": 1.2401,
+      "step": 13758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0404725714285716e-05,
+      "loss": 1.2255,
+      "step": 13759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0404457142857144e-05,
+      "loss": 1.1104,
+      "step": 13760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0404188571428572e-05,
+      "loss": 1.6825,
+      "step": 13761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.040392e-05,
+      "loss": 1.3182,
+      "step": 13762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.040365142857143e-05,
+      "loss": 1.1594,
+      "step": 13763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0403382857142857e-05,
+      "loss": 1.7524,
+      "step": 13764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0403114285714286e-05,
+      "loss": 0.8181,
+      "step": 13765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0402845714285715e-05,
+      "loss": 1.0258,
+      "step": 13766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0402577142857143e-05,
+      "loss": 1.3665,
+      "step": 13767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0402308571428571e-05,
+      "loss": 0.9192,
+      "step": 13768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.040204e-05,
+      "loss": 1.3576,
+      "step": 13769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0401771428571429e-05,
+      "loss": 1.029,
+      "step": 13770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0401502857142857e-05,
+      "loss": 0.945,
+      "step": 13771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0401234285714287e-05,
+      "loss": 0.9946,
+      "step": 13772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0400965714285715e-05,
+      "loss": 1.38,
+      "step": 13773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0400697142857143e-05,
+      "loss": 1.1328,
+      "step": 13774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0400428571428572e-05,
+      "loss": 1.2964,
+      "step": 13775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0400160000000001e-05,
+      "loss": 1.3329,
+      "step": 13776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0399891428571428e-05,
+      "loss": 0.8975,
+      "step": 13777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0399622857142858e-05,
+      "loss": 1.4258,
+      "step": 13778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0399354285714286e-05,
+      "loss": 1.1535,
+      "step": 13779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0399085714285714e-05,
+      "loss": 1.1397,
+      "step": 13780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0398817142857142e-05,
+      "loss": 1.4874,
+      "step": 13781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0398548571428572e-05,
+      "loss": 1.4655,
+      "step": 13782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.039828e-05,
+      "loss": 1.3155,
+      "step": 13783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0398011428571429e-05,
+      "loss": 1.2657,
+      "step": 13784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0397742857142858e-05,
+      "loss": 0.8935,
+      "step": 13785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0397474285714287e-05,
+      "loss": 1.4585,
+      "step": 13786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0397205714285715e-05,
+      "loss": 0.8062,
+      "step": 13787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0396937142857143e-05,
+      "loss": 0.7369,
+      "step": 13788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0396668571428573e-05,
+      "loss": 1.1928,
+      "step": 13789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.03964e-05,
+      "loss": 0.9768,
+      "step": 13790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0396131428571429e-05,
+      "loss": 1.4839,
+      "step": 13791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0395862857142857e-05,
+      "loss": 0.8325,
+      "step": 13792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0395594285714287e-05,
+      "loss": 1.4858,
+      "step": 13793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0395325714285714e-05,
+      "loss": 1.4775,
+      "step": 13794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0395057142857144e-05,
+      "loss": 1.149,
+      "step": 13795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0394788571428572e-05,
+      "loss": 1.2243,
+      "step": 13796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.039452e-05,
+      "loss": 1.1046,
+      "step": 13797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.039425142857143e-05,
+      "loss": 1.2594,
+      "step": 13798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0393982857142858e-05,
+      "loss": 1.4551,
+      "step": 13799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0393714285714286e-05,
+      "loss": 0.9772,
+      "step": 13800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0393445714285714e-05,
+      "loss": 1.3368,
+      "step": 13801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0393177142857144e-05,
+      "loss": 1.0121,
+      "step": 13802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.039290857142857e-05,
+      "loss": 1.172,
+      "step": 13803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.039264e-05,
+      "loss": 1.1444,
+      "step": 13804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0392371428571429e-05,
+      "loss": 0.8567,
+      "step": 13805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0392102857142859e-05,
+      "loss": 1.1879,
+      "step": 13806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0391834285714285e-05,
+      "loss": 1.4311,
+      "step": 13807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0391565714285715e-05,
+      "loss": 1.5175,
+      "step": 13808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0391297142857143e-05,
+      "loss": 1.2548,
+      "step": 13809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0391028571428571e-05,
+      "loss": 1.6966,
+      "step": 13810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0390760000000001e-05,
+      "loss": 1.4744,
+      "step": 13811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.039049142857143e-05,
+      "loss": 1.438,
+      "step": 13812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0390222857142857e-05,
+      "loss": 1.3913,
+      "step": 13813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0389954285714286e-05,
+      "loss": 0.8996,
+      "step": 13814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0389685714285716e-05,
+      "loss": 1.5044,
+      "step": 13815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0389417142857144e-05,
+      "loss": 1.0305,
+      "step": 13816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0389148571428572e-05,
+      "loss": 1.2944,
+      "step": 13817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.038888e-05,
+      "loss": 0.9454,
+      "step": 13818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.038861142857143e-05,
+      "loss": 1.4056,
+      "step": 13819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0388342857142856e-05,
+      "loss": 1.1531,
+      "step": 13820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0388074285714286e-05,
+      "loss": 1.0909,
+      "step": 13821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0387805714285714e-05,
+      "loss": 1.8304,
+      "step": 13822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0387537142857143e-05,
+      "loss": 1.4535,
+      "step": 13823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0387268571428572e-05,
+      "loss": 1.0403,
+      "step": 13824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0387e-05,
+      "loss": 1.0954,
+      "step": 13825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0386731428571429e-05,
+      "loss": 1.6338,
+      "step": 13826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0386462857142857e-05,
+      "loss": 1.0812,
+      "step": 13827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0386194285714287e-05,
+      "loss": 1.6107,
+      "step": 13828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0385925714285715e-05,
+      "loss": 1.0266,
+      "step": 13829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0385657142857143e-05,
+      "loss": 1.2173,
+      "step": 13830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0385388571428571e-05,
+      "loss": 1.6354,
+      "step": 13831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0385120000000001e-05,
+      "loss": 0.9864,
+      "step": 13832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0384851428571428e-05,
+      "loss": 1.3989,
+      "step": 13833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0384582857142858e-05,
+      "loss": 1.18,
+      "step": 13834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0384314285714286e-05,
+      "loss": 1.5471,
+      "step": 13835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0384045714285714e-05,
+      "loss": 1.3951,
+      "step": 13836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0383777142857144e-05,
+      "loss": 1.2663,
+      "step": 13837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0383508571428572e-05,
+      "loss": 0.7212,
+      "step": 13838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0383240000000002e-05,
+      "loss": 1.1898,
+      "step": 13839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0382971428571428e-05,
+      "loss": 1.08,
+      "step": 13840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0382702857142858e-05,
+      "loss": 1.469,
+      "step": 13841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0382434285714286e-05,
+      "loss": 1.5743,
+      "step": 13842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0382165714285715e-05,
+      "loss": 1.0135,
+      "step": 13843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0381897142857143e-05,
+      "loss": 1.246,
+      "step": 13844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0381628571428573e-05,
+      "loss": 1.4338,
+      "step": 13845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0381359999999999e-05,
+      "loss": 1.2873,
+      "step": 13846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0381091428571429e-05,
+      "loss": 1.1138,
+      "step": 13847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0380822857142857e-05,
+      "loss": 1.5613,
+      "step": 13848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0380554285714285e-05,
+      "loss": 1.1868,
+      "step": 13849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0380285714285715e-05,
+      "loss": 1.3095,
+      "step": 13850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0380017142857143e-05,
+      "loss": 0.9562,
+      "step": 13851
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0379748571428573e-05,
+      "loss": 1.2859,
+      "step": 13852
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037948e-05,
+      "loss": 1.3649,
+      "step": 13853
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037921142857143e-05,
+      "loss": 1.1317,
+      "step": 13854
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0378942857142858e-05,
+      "loss": 0.8507,
+      "step": 13855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0378674285714286e-05,
+      "loss": 1.0129,
+      "step": 13856
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0378405714285714e-05,
+      "loss": 1.288,
+      "step": 13857
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0378137142857144e-05,
+      "loss": 0.8871,
+      "step": 13858
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037786857142857e-05,
+      "loss": 1.0598,
+      "step": 13859
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.03776e-05,
+      "loss": 1.2518,
+      "step": 13860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0377331428571429e-05,
+      "loss": 1.4874,
+      "step": 13861
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0377062857142858e-05,
+      "loss": 0.9475,
+      "step": 13862
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0376794285714287e-05,
+      "loss": 1.334,
+      "step": 13863
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0376525714285715e-05,
+      "loss": 1.5405,
+      "step": 13864
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0376257142857145e-05,
+      "loss": 1.3619,
+      "step": 13865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0375988571428571e-05,
+      "loss": 1.304,
+      "step": 13866
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0375720000000001e-05,
+      "loss": 0.7472,
+      "step": 13867
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037545142857143e-05,
+      "loss": 1.3375,
+      "step": 13868
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0375182857142857e-05,
+      "loss": 1.0782,
+      "step": 13869
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0374914285714286e-05,
+      "loss": 1.134,
+      "step": 13870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0374645714285715e-05,
+      "loss": 1.8466,
+      "step": 13871
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0374377142857142e-05,
+      "loss": 1.293,
+      "step": 13872
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0374108571428572e-05,
+      "loss": 1.1532,
+      "step": 13873
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037384e-05,
+      "loss": 1.3734,
+      "step": 13874
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037357142857143e-05,
+      "loss": 0.8988,
+      "step": 13875
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0373302857142858e-05,
+      "loss": 1.5961,
+      "step": 13876
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0373034285714286e-05,
+      "loss": 1.0667,
+      "step": 13877
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0372765714285716e-05,
+      "loss": 1.3475,
+      "step": 13878
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0372497142857143e-05,
+      "loss": 1.4046,
+      "step": 13879
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0372228571428572e-05,
+      "loss": 0.8816,
+      "step": 13880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037196e-05,
+      "loss": 1.6164,
+      "step": 13881
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0371691428571429e-05,
+      "loss": 1.1636,
+      "step": 13882
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0371422857142857e-05,
+      "loss": 1.151,
+      "step": 13883
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0371154285714287e-05,
+      "loss": 1.0607,
+      "step": 13884
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0370885714285713e-05,
+      "loss": 1.5427,
+      "step": 13885
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0370617142857143e-05,
+      "loss": 1.0246,
+      "step": 13886
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0370348571428571e-05,
+      "loss": 1.1306,
+      "step": 13887
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0370080000000001e-05,
+      "loss": 1.3815,
+      "step": 13888
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.036981142857143e-05,
+      "loss": 1.223,
+      "step": 13889
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0369542857142858e-05,
+      "loss": 1.2581,
+      "step": 13890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0369274285714287e-05,
+      "loss": 1.1606,
+      "step": 13891
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0369005714285714e-05,
+      "loss": 1.0691,
+      "step": 13892
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0368737142857144e-05,
+      "loss": 1.2278,
+      "step": 13893
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0368468571428572e-05,
+      "loss": 0.8173,
+      "step": 13894
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.03682e-05,
+      "loss": 1.1656,
+      "step": 13895
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0367931428571428e-05,
+      "loss": 0.9661,
+      "step": 13896
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0367662857142858e-05,
+      "loss": 1.5328,
+      "step": 13897
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0367394285714286e-05,
+      "loss": 1.0108,
+      "step": 13898
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0367125714285715e-05,
+      "loss": 0.9734,
+      "step": 13899
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0366857142857143e-05,
+      "loss": 1.3086,
+      "step": 13900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0366588571428573e-05,
+      "loss": 0.7872,
+      "step": 13901
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.036632e-05,
+      "loss": 1.2741,
+      "step": 13902
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0366051428571429e-05,
+      "loss": 1.0126,
+      "step": 13903
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0365782857142859e-05,
+      "loss": 1.2527,
+      "step": 13904
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0365514285714285e-05,
+      "loss": 1.3256,
+      "step": 13905
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0365245714285715e-05,
+      "loss": 1.3798,
+      "step": 13906
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0364977142857143e-05,
+      "loss": 1.3629,
+      "step": 13907
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0364708571428571e-05,
+      "loss": 1.0762,
+      "step": 13908
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.036444e-05,
+      "loss": 0.9698,
+      "step": 13909
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.036417142857143e-05,
+      "loss": 0.9201,
+      "step": 13910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0363902857142858e-05,
+      "loss": 1.0656,
+      "step": 13911
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0363634285714286e-05,
+      "loss": 1.2926,
+      "step": 13912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0363365714285714e-05,
+      "loss": 0.847,
+      "step": 13913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0363097142857144e-05,
+      "loss": 1.5992,
+      "step": 13914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0362828571428572e-05,
+      "loss": 1.5267,
+      "step": 13915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.036256e-05,
+      "loss": 1.1521,
+      "step": 13916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.036229142857143e-05,
+      "loss": 1.2349,
+      "step": 13917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0362022857142857e-05,
+      "loss": 1.3494,
+      "step": 13918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0361754285714286e-05,
+      "loss": 1.1012,
+      "step": 13919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0361485714285715e-05,
+      "loss": 1.415,
+      "step": 13920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0361217142857145e-05,
+      "loss": 1.1689,
+      "step": 13921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0360948571428571e-05,
+      "loss": 1.4468,
+      "step": 13922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0360680000000001e-05,
+      "loss": 0.8212,
+      "step": 13923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0360411428571429e-05,
+      "loss": 1.1243,
+      "step": 13924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0360142857142857e-05,
+      "loss": 1.2005,
+      "step": 13925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0359874285714285e-05,
+      "loss": 0.904,
+      "step": 13926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0359605714285715e-05,
+      "loss": 1.0882,
+      "step": 13927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0359337142857142e-05,
+      "loss": 0.9028,
+      "step": 13928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0359068571428572e-05,
+      "loss": 0.9105,
+      "step": 13929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0358800000000002e-05,
+      "loss": 1.5144,
+      "step": 13930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0358531428571428e-05,
+      "loss": 1.2863,
+      "step": 13931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0358262857142858e-05,
+      "loss": 1.1035,
+      "step": 13932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0357994285714286e-05,
+      "loss": 1.3409,
+      "step": 13933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0357725714285716e-05,
+      "loss": 1.513,
+      "step": 13934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0357457142857142e-05,
+      "loss": 1.0003,
+      "step": 13935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0357188571428572e-05,
+      "loss": 1.1586,
+      "step": 13936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.035692e-05,
+      "loss": 1.0731,
+      "step": 13937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0356651428571429e-05,
+      "loss": 0.7922,
+      "step": 13938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0356382857142857e-05,
+      "loss": 1.1722,
+      "step": 13939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0356114285714287e-05,
+      "loss": 1.1112,
+      "step": 13940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0355845714285713e-05,
+      "loss": 1.5686,
+      "step": 13941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0355577142857143e-05,
+      "loss": 1.1051,
+      "step": 13942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0355308571428573e-05,
+      "loss": 1.2075,
+      "step": 13943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0355040000000001e-05,
+      "loss": 1.3666,
+      "step": 13944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.035477142857143e-05,
+      "loss": 1.1346,
+      "step": 13945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0354502857142857e-05,
+      "loss": 1.5753,
+      "step": 13946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0354234285714287e-05,
+      "loss": 1.2223,
+      "step": 13947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0353965714285714e-05,
+      "loss": 1.0281,
+      "step": 13948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0353697142857144e-05,
+      "loss": 1.4777,
+      "step": 13949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0353428571428572e-05,
+      "loss": 1.2002,
+      "step": 13950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.035316e-05,
+      "loss": 0.9281,
+      "step": 13951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0352891428571428e-05,
+      "loss": 1.0187,
+      "step": 13952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0352622857142858e-05,
+      "loss": 1.407,
+      "step": 13953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0352354285714285e-05,
+      "loss": 1.1462,
+      "step": 13954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0352085714285714e-05,
+      "loss": 0.9813,
+      "step": 13955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0351817142857144e-05,
+      "loss": 1.4242,
+      "step": 13956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0351548571428572e-05,
+      "loss": 1.1401,
+      "step": 13957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.035128e-05,
+      "loss": 0.9612,
+      "step": 13958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0351011428571429e-05,
+      "loss": 1.2253,
+      "step": 13959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0350742857142859e-05,
+      "loss": 1.1056,
+      "step": 13960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0350474285714285e-05,
+      "loss": 1.8811,
+      "step": 13961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0350205714285715e-05,
+      "loss": 1.1502,
+      "step": 13962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0349937142857143e-05,
+      "loss": 1.0737,
+      "step": 13963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0349668571428571e-05,
+      "loss": 1.5087,
+      "step": 13964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.03494e-05,
+      "loss": 1.2323,
+      "step": 13965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034913142857143e-05,
+      "loss": 1.1348,
+      "step": 13966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0348862857142858e-05,
+      "loss": 1.5625,
+      "step": 13967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0348594285714286e-05,
+      "loss": 0.889,
+      "step": 13968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0348325714285716e-05,
+      "loss": 1.217,
+      "step": 13969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0348057142857144e-05,
+      "loss": 1.7527,
+      "step": 13970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0347788571428572e-05,
+      "loss": 0.7065,
+      "step": 13971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034752e-05,
+      "loss": 1.0993,
+      "step": 13972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034725142857143e-05,
+      "loss": 1.3323,
+      "step": 13973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0346982857142857e-05,
+      "loss": 1.4342,
+      "step": 13974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0346714285714286e-05,
+      "loss": 1.2925,
+      "step": 13975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0346445714285715e-05,
+      "loss": 0.8658,
+      "step": 13976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0346177142857143e-05,
+      "loss": 1.2682,
+      "step": 13977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0345908571428571e-05,
+      "loss": 1.0602,
+      "step": 13978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034564e-05,
+      "loss": 1.1832,
+      "step": 13979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0345371428571429e-05,
+      "loss": 1.3284,
+      "step": 13980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0345102857142857e-05,
+      "loss": 1.4701,
+      "step": 13981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0344834285714287e-05,
+      "loss": 1.0228,
+      "step": 13982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0344565714285715e-05,
+      "loss": 0.9144,
+      "step": 13983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0344297142857143e-05,
+      "loss": 1.0568,
+      "step": 13984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0344028571428572e-05,
+      "loss": 1.3075,
+      "step": 13985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0343760000000001e-05,
+      "loss": 1.1201,
+      "step": 13986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0343491428571428e-05,
+      "loss": 1.3377,
+      "step": 13987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0343222857142858e-05,
+      "loss": 0.9428,
+      "step": 13988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0342954285714286e-05,
+      "loss": 1.061,
+      "step": 13989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0342685714285716e-05,
+      "loss": 1.172,
+      "step": 13990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0342417142857142e-05,
+      "loss": 1.1713,
+      "step": 13991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0342148571428572e-05,
+      "loss": 0.8949,
+      "step": 13992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034188e-05,
+      "loss": 1.5522,
+      "step": 13993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0341611428571429e-05,
+      "loss": 1.3048,
+      "step": 13994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0341342857142858e-05,
+      "loss": 0.7899,
+      "step": 13995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0341074285714287e-05,
+      "loss": 1.7964,
+      "step": 13996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0340805714285715e-05,
+      "loss": 1.3795,
+      "step": 13997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0340537142857143e-05,
+      "loss": 1.2189,
+      "step": 13998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0340268571428573e-05,
+      "loss": 1.2432,
+      "step": 13999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034e-05,
+      "loss": 0.9359,
+      "step": 14000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0339731428571429e-05,
+      "loss": 1.3133,
+      "step": 14001
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0339462857142857e-05,
+      "loss": 1.0045,
+      "step": 14002
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0339194285714287e-05,
+      "loss": 1.2966,
+      "step": 14003
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0338925714285714e-05,
+      "loss": 1.7372,
+      "step": 14004
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0338657142857144e-05,
+      "loss": 0.8617,
+      "step": 14005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0338388571428572e-05,
+      "loss": 1.0045,
+      "step": 14006
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033812e-05,
+      "loss": 0.9896,
+      "step": 14007
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033785142857143e-05,
+      "loss": 1.2656,
+      "step": 14008
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0337582857142858e-05,
+      "loss": 0.9172,
+      "step": 14009
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0337314285714286e-05,
+      "loss": 1.1834,
+      "step": 14010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0337045714285714e-05,
+      "loss": 1.0812,
+      "step": 14011
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0336777142857144e-05,
+      "loss": 1.4899,
+      "step": 14012
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0336508571428572e-05,
+      "loss": 1.1654,
+      "step": 14013
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033624e-05,
+      "loss": 1.3946,
+      "step": 14014
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0335971428571429e-05,
+      "loss": 1.0705,
+      "step": 14015
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0335702857142859e-05,
+      "loss": 1.2497,
+      "step": 14016
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0335434285714285e-05,
+      "loss": 1.3434,
+      "step": 14017
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0335165714285715e-05,
+      "loss": 1.5967,
+      "step": 14018
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0334897142857143e-05,
+      "loss": 1.2462,
+      "step": 14019
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0334628571428571e-05,
+      "loss": 0.9466,
+      "step": 14020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0334360000000001e-05,
+      "loss": 1.1325,
+      "step": 14021
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033409142857143e-05,
+      "loss": 1.4472,
+      "step": 14022
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0333822857142857e-05,
+      "loss": 1.0388,
+      "step": 14023
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0333554285714286e-05,
+      "loss": 1.0092,
+      "step": 14024
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0333285714285716e-05,
+      "loss": 1.3772,
+      "step": 14025
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0333017142857144e-05,
+      "loss": 1.3768,
+      "step": 14026
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0332748571428572e-05,
+      "loss": 1.1186,
+      "step": 14027
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033248e-05,
+      "loss": 1.2287,
+      "step": 14028
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033221142857143e-05,
+      "loss": 0.9826,
+      "step": 14029
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0331942857142856e-05,
+      "loss": 1.2362,
+      "step": 14030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0331674285714286e-05,
+      "loss": 1.4408,
+      "step": 14031
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0331405714285714e-05,
+      "loss": 1.199,
+      "step": 14032
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0331137142857143e-05,
+      "loss": 1.0752,
+      "step": 14033
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033086857142857e-05,
+      "loss": 1.0675,
+      "step": 14034
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.03306e-05,
+      "loss": 0.9891,
+      "step": 14035
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033033142857143e-05,
+      "loss": 1.257,
+      "step": 14036
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0330062857142857e-05,
+      "loss": 1.5886,
+      "step": 14037
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0329794285714287e-05,
+      "loss": 1.3386,
+      "step": 14038
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0329525714285715e-05,
+      "loss": 1.1089,
+      "step": 14039
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0329257142857143e-05,
+      "loss": 1.5126,
+      "step": 14040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0328988571428571e-05,
+      "loss": 0.9014,
+      "step": 14041
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0328720000000001e-05,
+      "loss": 1.3004,
+      "step": 14042
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0328451428571428e-05,
+      "loss": 1.1976,
+      "step": 14043
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0328182857142858e-05,
+      "loss": 1.2505,
+      "step": 14044
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0327914285714286e-05,
+      "loss": 1.3872,
+      "step": 14045
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0327645714285714e-05,
+      "loss": 1.2396,
+      "step": 14046
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0327377142857142e-05,
+      "loss": 1.5656,
+      "step": 14047
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0327108571428572e-05,
+      "loss": 1.0828,
+      "step": 14048
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0326840000000002e-05,
+      "loss": 0.864,
+      "step": 14049
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0326571428571428e-05,
+      "loss": 1.1989,
+      "step": 14050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0326302857142858e-05,
+      "loss": 1.3403,
+      "step": 14051
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0326034285714286e-05,
+      "loss": 1.335,
+      "step": 14052
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0325765714285715e-05,
+      "loss": 1.0772,
+      "step": 14053
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0325497142857143e-05,
+      "loss": 1.712,
+      "step": 14054
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0325228571428573e-05,
+      "loss": 1.3954,
+      "step": 14055
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0324959999999999e-05,
+      "loss": 0.8443,
+      "step": 14056
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0324691428571429e-05,
+      "loss": 1.4299,
+      "step": 14057
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0324422857142857e-05,
+      "loss": 1.2675,
+      "step": 14058
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0324154285714285e-05,
+      "loss": 1.0506,
+      "step": 14059
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0323885714285714e-05,
+      "loss": 1.3789,
+      "step": 14060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0323617142857143e-05,
+      "loss": 1.4691,
+      "step": 14061
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0323348571428573e-05,
+      "loss": 1.0957,
+      "step": 14062
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.032308e-05,
+      "loss": 1.2127,
+      "step": 14063
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.032281142857143e-05,
+      "loss": 1.0662,
+      "step": 14064
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0322542857142858e-05,
+      "loss": 1.0037,
+      "step": 14065
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0322274285714286e-05,
+      "loss": 1.1397,
+      "step": 14066
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0322005714285714e-05,
+      "loss": 1.3562,
+      "step": 14067
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0321737142857144e-05,
+      "loss": 1.5113,
+      "step": 14068
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.032146857142857e-05,
+      "loss": 1.4862,
+      "step": 14069
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.03212e-05,
+      "loss": 1.3393,
+      "step": 14070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0320931428571429e-05,
+      "loss": 1.5277,
+      "step": 14071
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0320662857142858e-05,
+      "loss": 1.6116,
+      "step": 14072
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0320394285714285e-05,
+      "loss": 1.3083,
+      "step": 14073
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0320125714285715e-05,
+      "loss": 0.8503,
+      "step": 14074
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0319857142857145e-05,
+      "loss": 1.4558,
+      "step": 14075
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0319588571428571e-05,
+      "loss": 1.6356,
+      "step": 14076
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0319320000000001e-05,
+      "loss": 0.9218,
+      "step": 14077
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.031905142857143e-05,
+      "loss": 1.3826,
+      "step": 14078
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0318782857142857e-05,
+      "loss": 1.3498,
+      "step": 14079
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0318514285714286e-05,
+      "loss": 0.9202,
+      "step": 14080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0318245714285715e-05,
+      "loss": 1.1051,
+      "step": 14081
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0317977142857142e-05,
+      "loss": 1.4535,
+      "step": 14082
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0317708571428572e-05,
+      "loss": 1.174,
+      "step": 14083
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.031744e-05,
+      "loss": 0.9878,
+      "step": 14084
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.031717142857143e-05,
+      "loss": 1.1368,
+      "step": 14085
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0316902857142856e-05,
+      "loss": 1.3986,
+      "step": 14086
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0316634285714286e-05,
+      "loss": 1.3523,
+      "step": 14087
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0316365714285716e-05,
+      "loss": 1.3747,
+      "step": 14088
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0316097142857143e-05,
+      "loss": 1.0628,
+      "step": 14089
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0315828571428572e-05,
+      "loss": 1.2271,
+      "step": 14090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.031556e-05,
+      "loss": 1.2818,
+      "step": 14091
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0315291428571429e-05,
+      "loss": 0.8727,
+      "step": 14092
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0315022857142857e-05,
+      "loss": 1.0902,
+      "step": 14093
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0314754285714287e-05,
+      "loss": 1.3825,
+      "step": 14094
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0314485714285715e-05,
+      "loss": 1.522,
+      "step": 14095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0314217142857143e-05,
+      "loss": 1.5076,
+      "step": 14096
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0313948571428571e-05,
+      "loss": 1.0544,
+      "step": 14097
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0313680000000001e-05,
+      "loss": 1.2876,
+      "step": 14098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0313411428571428e-05,
+      "loss": 1.2469,
+      "step": 14099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0313142857142858e-05,
+      "loss": 1.1417,
+      "step": 14100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0312874285714287e-05,
+      "loss": 1.4076,
+      "step": 14101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0312605714285714e-05,
+      "loss": 1.0238,
+      "step": 14102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0312337142857144e-05,
+      "loss": 1.3446,
+      "step": 14103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0312068571428572e-05,
+      "loss": 0.9043,
+      "step": 14104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.03118e-05,
+      "loss": 1.1917,
+      "step": 14105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0311531428571428e-05,
+      "loss": 1.2761,
+      "step": 14106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0311262857142858e-05,
+      "loss": 1.008,
+      "step": 14107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0310994285714286e-05,
+      "loss": 1.4762,
+      "step": 14108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0310725714285715e-05,
+      "loss": 1.411,
+      "step": 14109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0310457142857143e-05,
+      "loss": 1.4578,
+      "step": 14110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0310188571428573e-05,
+      "loss": 1.2951,
+      "step": 14111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0309919999999999e-05,
+      "loss": 1.2078,
+      "step": 14112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0309651428571429e-05,
+      "loss": 1.7131,
+      "step": 14113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0309382857142859e-05,
+      "loss": 1.2032,
+      "step": 14114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0309114285714285e-05,
+      "loss": 1.3032,
+      "step": 14115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0308845714285715e-05,
+      "loss": 1.4079,
+      "step": 14116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0308577142857143e-05,
+      "loss": 1.2014,
+      "step": 14117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0308308571428573e-05,
+      "loss": 1.1434,
+      "step": 14118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.030804e-05,
+      "loss": 0.8696,
+      "step": 14119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.030777142857143e-05,
+      "loss": 1.2315,
+      "step": 14120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0307502857142858e-05,
+      "loss": 1.5343,
+      "step": 14121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0307234285714286e-05,
+      "loss": 0.6604,
+      "step": 14122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0306965714285714e-05,
+      "loss": 1.2269,
+      "step": 14123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0306697142857144e-05,
+      "loss": 1.2522,
+      "step": 14124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.030642857142857e-05,
+      "loss": 1.3854,
+      "step": 14125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.030616e-05,
+      "loss": 0.7239,
+      "step": 14126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.030589142857143e-05,
+      "loss": 1.493,
+      "step": 14127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0305622857142857e-05,
+      "loss": 1.7256,
+      "step": 14128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0305354285714286e-05,
+      "loss": 1.0903,
+      "step": 14129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0305085714285715e-05,
+      "loss": 1.4806,
+      "step": 14130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0304817142857145e-05,
+      "loss": 0.7228,
+      "step": 14131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0304548571428571e-05,
+      "loss": 1.3595,
+      "step": 14132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0304280000000001e-05,
+      "loss": 1.1031,
+      "step": 14133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0304011428571429e-05,
+      "loss": 1.1524,
+      "step": 14134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0303742857142857e-05,
+      "loss": 0.9268,
+      "step": 14135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0303474285714285e-05,
+      "loss": 1.1657,
+      "step": 14136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0303205714285715e-05,
+      "loss": 1.221,
+      "step": 14137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0302937142857142e-05,
+      "loss": 2.043,
+      "step": 14138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0302668571428572e-05,
+      "loss": 1.0566,
+      "step": 14139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.03024e-05,
+      "loss": 1.3016,
+      "step": 14140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.030213142857143e-05,
+      "loss": 1.2428,
+      "step": 14141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0301862857142858e-05,
+      "loss": 0.8236,
+      "step": 14142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0301594285714286e-05,
+      "loss": 0.9732,
+      "step": 14143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0301325714285716e-05,
+      "loss": 0.9542,
+      "step": 14144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0301057142857142e-05,
+      "loss": 1.3786,
+      "step": 14145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0300788571428572e-05,
+      "loss": 0.8603,
+      "step": 14146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.030052e-05,
+      "loss": 1.6754,
+      "step": 14147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0300251428571429e-05,
+      "loss": 1.5511,
+      "step": 14148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0299982857142857e-05,
+      "loss": 1.0111,
+      "step": 14149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0299714285714287e-05,
+      "loss": 0.9838,
+      "step": 14150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0299445714285713e-05,
+      "loss": 1.5897,
+      "step": 14151
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0299177142857143e-05,
+      "loss": 1.8233,
+      "step": 14152
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0298908571428571e-05,
+      "loss": 1.2424,
+      "step": 14153
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0298640000000001e-05,
+      "loss": 1.1936,
+      "step": 14154
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.029837142857143e-05,
+      "loss": 1.3845,
+      "step": 14155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0298102857142857e-05,
+      "loss": 1.5674,
+      "step": 14156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0297834285714287e-05,
+      "loss": 1.5417,
+      "step": 14157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0297565714285714e-05,
+      "loss": 1.0022,
+      "step": 14158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0297297142857144e-05,
+      "loss": 1.2925,
+      "step": 14159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0297028571428572e-05,
+      "loss": 1.0833,
+      "step": 14160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.029676e-05,
+      "loss": 1.758,
+      "step": 14161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0296491428571428e-05,
+      "loss": 1.4536,
+      "step": 14162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0296222857142858e-05,
+      "loss": 0.9513,
+      "step": 14163
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0295954285714286e-05,
+      "loss": 1.1208,
+      "step": 14164
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0295685714285714e-05,
+      "loss": 1.132,
+      "step": 14165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0295417142857143e-05,
+      "loss": 1.0777,
+      "step": 14166
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0295148571428572e-05,
+      "loss": 1.0375,
+      "step": 14167
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.029488e-05,
+      "loss": 1.4755,
+      "step": 14168
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0294611428571429e-05,
+      "loss": 1.1961,
+      "step": 14169
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0294342857142859e-05,
+      "loss": 1.4657,
+      "step": 14170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0294074285714285e-05,
+      "loss": 1.4074,
+      "step": 14171
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0293805714285715e-05,
+      "loss": 1.2329,
+      "step": 14172
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0293537142857143e-05,
+      "loss": 1.1487,
+      "step": 14173
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0293268571428571e-05,
+      "loss": 0.9252,
+      "step": 14174
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0293e-05,
+      "loss": 0.9511,
+      "step": 14175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.029273142857143e-05,
+      "loss": 1.111,
+      "step": 14176
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0292462857142858e-05,
+      "loss": 1.224,
+      "step": 14177
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0292194285714286e-05,
+      "loss": 0.9991,
+      "step": 14178
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0291925714285714e-05,
+      "loss": 1.3544,
+      "step": 14179
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0291657142857144e-05,
+      "loss": 1.3743,
+      "step": 14180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0291388571428572e-05,
+      "loss": 1.1702,
+      "step": 14181
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.029112e-05,
+      "loss": 1.558,
+      "step": 14182
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.029085142857143e-05,
+      "loss": 1.2478,
+      "step": 14183
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0290582857142857e-05,
+      "loss": 0.9967,
+      "step": 14184
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0290314285714286e-05,
+      "loss": 1.2451,
+      "step": 14185
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0290045714285715e-05,
+      "loss": 1.196,
+      "step": 14186
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0289777142857144e-05,
+      "loss": 1.4205,
+      "step": 14187
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0289508571428571e-05,
+      "loss": 1.0999,
+      "step": 14188
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.028924e-05,
+      "loss": 0.9536,
+      "step": 14189
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0288971428571429e-05,
+      "loss": 1.0636,
+      "step": 14190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0288702857142857e-05,
+      "loss": 1.5583,
+      "step": 14191
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0288434285714285e-05,
+      "loss": 1.3785,
+      "step": 14192
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0288165714285715e-05,
+      "loss": 1.3627,
+      "step": 14193
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0287897142857143e-05,
+      "loss": 0.4728,
+      "step": 14194
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0287628571428572e-05,
+      "loss": 0.9516,
+      "step": 14195
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0287360000000001e-05,
+      "loss": 1.0881,
+      "step": 14196
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0287091428571428e-05,
+      "loss": 1.1439,
+      "step": 14197
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0286822857142858e-05,
+      "loss": 1.5407,
+      "step": 14198
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0286554285714286e-05,
+      "loss": 1.4476,
+      "step": 14199
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0286285714285716e-05,
+      "loss": 1.5179,
+      "step": 14200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0286017142857142e-05,
+      "loss": 1.4527,
+      "step": 14201
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0285748571428572e-05,
+      "loss": 1.5786,
+      "step": 14202
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.028548e-05,
+      "loss": 1.4156,
+      "step": 14203
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0285211428571429e-05,
+      "loss": 1.351,
+      "step": 14204
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0284942857142857e-05,
+      "loss": 0.9421,
+      "step": 14205
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0284674285714287e-05,
+      "loss": 1.1861,
+      "step": 14206
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0284405714285715e-05,
+      "loss": 1.476,
+      "step": 14207
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0284137142857143e-05,
+      "loss": 0.88,
+      "step": 14208
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0283868571428573e-05,
+      "loss": 1.2115,
+      "step": 14209
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.02836e-05,
+      "loss": 1.2409,
+      "step": 14210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0283331428571429e-05,
+      "loss": 0.802,
+      "step": 14211
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0283062857142857e-05,
+      "loss": 1.1139,
+      "step": 14212
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0282794285714287e-05,
+      "loss": 1.3721,
+      "step": 14213
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0282525714285714e-05,
+      "loss": 1.4509,
+      "step": 14214
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0282257142857144e-05,
+      "loss": 1.2096,
+      "step": 14215
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0281988571428572e-05,
+      "loss": 1.4574,
+      "step": 14216
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.028172e-05,
+      "loss": 0.8673,
+      "step": 14217
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0281451428571428e-05,
+      "loss": 1.1383,
+      "step": 14218
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0281182857142858e-05,
+      "loss": 1.6419,
+      "step": 14219
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0280914285714286e-05,
+      "loss": 1.0558,
+      "step": 14220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0280645714285714e-05,
+      "loss": 1.3009,
+      "step": 14221
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0280377142857144e-05,
+      "loss": 1.0996,
+      "step": 14222
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0280108571428572e-05,
+      "loss": 1.0853,
+      "step": 14223
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.027984e-05,
+      "loss": 1.4985,
+      "step": 14224
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0279571428571429e-05,
+      "loss": 0.9616,
+      "step": 14225
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0279302857142859e-05,
+      "loss": 1.0014,
+      "step": 14226
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0279034285714285e-05,
+      "loss": 1.49,
+      "step": 14227
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0278765714285715e-05,
+      "loss": 1.4817,
+      "step": 14228
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0278497142857143e-05,
+      "loss": 1.6262,
+      "step": 14229
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0278228571428571e-05,
+      "loss": 1.1169,
+      "step": 14230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.027796e-05,
+      "loss": 1.0293,
+      "step": 14231
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.027769142857143e-05,
+      "loss": 1.4243,
+      "step": 14232
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0277422857142857e-05,
+      "loss": 1.6847,
+      "step": 14233
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0277154285714286e-05,
+      "loss": 1.245,
+      "step": 14234
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0276885714285716e-05,
+      "loss": 0.9541,
+      "step": 14235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0276617142857144e-05,
+      "loss": 1.1391,
+      "step": 14236
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0276348571428572e-05,
+      "loss": 1.479,
+      "step": 14237
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.027608e-05,
+      "loss": 1.2736,
+      "step": 14238
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.027581142857143e-05,
+      "loss": 1.1261,
+      "step": 14239
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0275542857142856e-05,
+      "loss": 1.272,
+      "step": 14240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0275274285714286e-05,
+      "loss": 1.2881,
+      "step": 14241
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0275005714285714e-05,
+      "loss": 1.1936,
+      "step": 14242
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0274737142857143e-05,
+      "loss": 0.9695,
+      "step": 14243
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.027446857142857e-05,
+      "loss": 0.7175,
+      "step": 14244
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.02742e-05,
+      "loss": 1.0943,
+      "step": 14245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0273931428571429e-05,
+      "loss": 1.2359,
+      "step": 14246
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0273662857142857e-05,
+      "loss": 1.5041,
+      "step": 14247
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0273394285714287e-05,
+      "loss": 1.2865,
+      "step": 14248
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0273125714285715e-05,
+      "loss": 1.2059,
+      "step": 14249
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0272857142857143e-05,
+      "loss": 1.4739,
+      "step": 14250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0272588571428571e-05,
+      "loss": 1.1154,
+      "step": 14251
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0272320000000001e-05,
+      "loss": 1.3503,
+      "step": 14252
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0272051428571428e-05,
+      "loss": 0.9447,
+      "step": 14253
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0271782857142858e-05,
+      "loss": 1.2235,
+      "step": 14254
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0271514285714286e-05,
+      "loss": 1.1494,
+      "step": 14255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0271245714285714e-05,
+      "loss": 1.1929,
+      "step": 14256
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0270977142857142e-05,
+      "loss": 1.2057,
+      "step": 14257
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0270708571428572e-05,
+      "loss": 0.7986,
+      "step": 14258
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.027044e-05,
+      "loss": 1.2065,
+      "step": 14259
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0270171428571428e-05,
+      "loss": 1.6358,
+      "step": 14260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0269902857142858e-05,
+      "loss": 1.3349,
+      "step": 14261
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0269634285714286e-05,
+      "loss": 1.7154,
+      "step": 14262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0269365714285715e-05,
+      "loss": 1.4452,
+      "step": 14263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0269097142857143e-05,
+      "loss": 1.1297,
+      "step": 14264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0268828571428573e-05,
+      "loss": 1.4707,
+      "step": 14265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0268559999999999e-05,
+      "loss": 1.1919,
+      "step": 14266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0268291428571429e-05,
+      "loss": 0.8391,
+      "step": 14267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0268022857142857e-05,
+      "loss": 1.2106,
+      "step": 14268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0267754285714287e-05,
+      "loss": 0.8929,
+      "step": 14269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0267485714285714e-05,
+      "loss": 1.25,
+      "step": 14270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0267217142857143e-05,
+      "loss": 1.5206,
+      "step": 14271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0266948571428572e-05,
+      "loss": 1.4177,
+      "step": 14272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.026668e-05,
+      "loss": 1.3912,
+      "step": 14273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.026641142857143e-05,
+      "loss": 0.9463,
+      "step": 14274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0266142857142858e-05,
+      "loss": 1.4382,
+      "step": 14275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0265874285714286e-05,
+      "loss": 1.1008,
+      "step": 14276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0265605714285714e-05,
+      "loss": 1.1777,
+      "step": 14277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0265337142857144e-05,
+      "loss": 1.2192,
+      "step": 14278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.026506857142857e-05,
+      "loss": 1.2065,
+      "step": 14279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.02648e-05,
+      "loss": 0.7179,
+      "step": 14280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0264531428571429e-05,
+      "loss": 1.7207,
+      "step": 14281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0264262857142858e-05,
+      "loss": 1.4655,
+      "step": 14282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0263994285714285e-05,
+      "loss": 1.4344,
+      "step": 14283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0263725714285715e-05,
+      "loss": 1.2955,
+      "step": 14284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0263457142857143e-05,
+      "loss": 1.2703,
+      "step": 14285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0263188571428571e-05,
+      "loss": 1.0992,
+      "step": 14286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0262920000000001e-05,
+      "loss": 1.1646,
+      "step": 14287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.026265142857143e-05,
+      "loss": 1.1452,
+      "step": 14288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0262382857142857e-05,
+      "loss": 0.8875,
+      "step": 14289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0262114285714286e-05,
+      "loss": 1.0227,
+      "step": 14290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0261845714285715e-05,
+      "loss": 1.1894,
+      "step": 14291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0261577142857144e-05,
+      "loss": 1.1303,
+      "step": 14292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0261308571428572e-05,
+      "loss": 1.1897,
+      "step": 14293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.026104e-05,
+      "loss": 1.3162,
+      "step": 14294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.026077142857143e-05,
+      "loss": 1.4434,
+      "step": 14295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0260502857142856e-05,
+      "loss": 1.4209,
+      "step": 14296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0260234285714286e-05,
+      "loss": 1.128,
+      "step": 14297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0259965714285714e-05,
+      "loss": 1.2814,
+      "step": 14298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0259697142857143e-05,
+      "loss": 1.4032,
+      "step": 14299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0259428571428572e-05,
+      "loss": 1.1767,
+      "step": 14300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.025916e-05,
+      "loss": 1.2993,
+      "step": 14301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0258891428571429e-05,
+      "loss": 1.1116,
+      "step": 14302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0258622857142857e-05,
+      "loss": 1.2767,
+      "step": 14303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0258354285714287e-05,
+      "loss": 1.4166,
+      "step": 14304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0258085714285715e-05,
+      "loss": 1.0503,
+      "step": 14305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0257817142857143e-05,
+      "loss": 1.4192,
+      "step": 14306
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0257548571428571e-05,
+      "loss": 1.4722,
+      "step": 14307
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0257280000000001e-05,
+      "loss": 1.776,
+      "step": 14308
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0257011428571428e-05,
+      "loss": 1.7054,
+      "step": 14309
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0256742857142858e-05,
+      "loss": 1.0989,
+      "step": 14310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0256474285714286e-05,
+      "loss": 1.0491,
+      "step": 14311
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0256205714285714e-05,
+      "loss": 1.0673,
+      "step": 14312
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0255937142857144e-05,
+      "loss": 1.4531,
+      "step": 14313
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0255668571428572e-05,
+      "loss": 0.9682,
+      "step": 14314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0255400000000002e-05,
+      "loss": 1.2298,
+      "step": 14315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0255131428571428e-05,
+      "loss": 1.6484,
+      "step": 14316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0254862857142858e-05,
+      "loss": 1.1371,
+      "step": 14317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0254594285714286e-05,
+      "loss": 1.4307,
+      "step": 14318
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0254325714285714e-05,
+      "loss": 1.2765,
+      "step": 14319
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0254057142857143e-05,
+      "loss": 1.147,
+      "step": 14320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0253788571428573e-05,
+      "loss": 1.0927,
+      "step": 14321
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0253519999999999e-05,
+      "loss": 1.4444,
+      "step": 14322
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0253251428571429e-05,
+      "loss": 1.236,
+      "step": 14323
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0252982857142857e-05,
+      "loss": 1.2955,
+      "step": 14324
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0252714285714285e-05,
+      "loss": 1.4363,
+      "step": 14325
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0252445714285715e-05,
+      "loss": 0.9161,
+      "step": 14326
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0252177142857143e-05,
+      "loss": 1.2638,
+      "step": 14327
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0251908571428573e-05,
+      "loss": 1.29,
+      "step": 14328
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.025164e-05,
+      "loss": 1.2299,
+      "step": 14329
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.025137142857143e-05,
+      "loss": 1.1382,
+      "step": 14330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0251102857142858e-05,
+      "loss": 1.4311,
+      "step": 14331
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0250834285714286e-05,
+      "loss": 0.7433,
+      "step": 14332
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0250565714285714e-05,
+      "loss": 1.1283,
+      "step": 14333
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0250297142857144e-05,
+      "loss": 1.3595,
+      "step": 14334
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.025002857142857e-05,
+      "loss": 1.3874,
+      "step": 14335
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.024976e-05,
+      "loss": 1.6368,
+      "step": 14336
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0249491428571428e-05,
+      "loss": 0.7109,
+      "step": 14337
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0249222857142858e-05,
+      "loss": 1.3893,
+      "step": 14338
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0248954285714286e-05,
+      "loss": 1.0977,
+      "step": 14339
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0248685714285715e-05,
+      "loss": 0.7851,
+      "step": 14340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0248417142857145e-05,
+      "loss": 1.3254,
+      "step": 14341
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0248148571428571e-05,
+      "loss": 1.308,
+      "step": 14342
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0247880000000001e-05,
+      "loss": 1.7299,
+      "step": 14343
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0247611428571429e-05,
+      "loss": 0.9822,
+      "step": 14344
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0247342857142857e-05,
+      "loss": 1.4769,
+      "step": 14345
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0247074285714285e-05,
+      "loss": 1.5222,
+      "step": 14346
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0246805714285715e-05,
+      "loss": 1.221,
+      "step": 14347
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0246537142857142e-05,
+      "loss": 0.9788,
+      "step": 14348
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0246268571428572e-05,
+      "loss": 1.7539,
+      "step": 14349
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0246e-05,
+      "loss": 1.1535,
+      "step": 14350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.024573142857143e-05,
+      "loss": 1.3546,
+      "step": 14351
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0245462857142858e-05,
+      "loss": 1.084,
+      "step": 14352
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0245194285714286e-05,
+      "loss": 1.1118,
+      "step": 14353
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0244925714285716e-05,
+      "loss": 0.7164,
+      "step": 14354
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0244657142857142e-05,
+      "loss": 1.3532,
+      "step": 14355
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0244388571428572e-05,
+      "loss": 1.1547,
+      "step": 14356
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.024412e-05,
+      "loss": 1.2022,
+      "step": 14357
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0243851428571429e-05,
+      "loss": 1.5353,
+      "step": 14358
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0243582857142857e-05,
+      "loss": 1.6969,
+      "step": 14359
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0243314285714287e-05,
+      "loss": 0.9175,
+      "step": 14360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0243045714285715e-05,
+      "loss": 0.9895,
+      "step": 14361
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0242777142857143e-05,
+      "loss": 1.0026,
+      "step": 14362
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0242508571428571e-05,
+      "loss": 0.7743,
+      "step": 14363
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0242240000000001e-05,
+      "loss": 1.2025,
+      "step": 14364
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.024197142857143e-05,
+      "loss": 1.3102,
+      "step": 14365
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0241702857142857e-05,
+      "loss": 1.0962,
+      "step": 14366
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0241434285714287e-05,
+      "loss": 1.2449,
+      "step": 14367
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0241165714285714e-05,
+      "loss": 1.1556,
+      "step": 14368
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0240897142857144e-05,
+      "loss": 1.2653,
+      "step": 14369
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0240628571428572e-05,
+      "loss": 1.0019,
+      "step": 14370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.024036e-05,
+      "loss": 1.3656,
+      "step": 14371
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0240091428571428e-05,
+      "loss": 0.8701,
+      "step": 14372
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0239822857142858e-05,
+      "loss": 1.2754,
+      "step": 14373
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0239554285714286e-05,
+      "loss": 1.2523,
+      "step": 14374
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0239285714285714e-05,
+      "loss": 1.2389,
+      "step": 14375
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0239017142857143e-05,
+      "loss": 1.3745,
+      "step": 14376
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0238748571428572e-05,
+      "loss": 1.0992,
+      "step": 14377
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.023848e-05,
+      "loss": 1.324,
+      "step": 14378
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0238211428571429e-05,
+      "loss": 1.3331,
+      "step": 14379
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0237942857142859e-05,
+      "loss": 1.2753,
+      "step": 14380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0237674285714285e-05,
+      "loss": 0.9347,
+      "step": 14381
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0237405714285715e-05,
+      "loss": 1.1663,
+      "step": 14382
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0237137142857143e-05,
+      "loss": 1.1095,
+      "step": 14383
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0236868571428571e-05,
+      "loss": 1.089,
+      "step": 14384
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.02366e-05,
+      "loss": 1.3139,
+      "step": 14385
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.023633142857143e-05,
+      "loss": 1.083,
+      "step": 14386
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0236062857142858e-05,
+      "loss": 1.385,
+      "step": 14387
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0235794285714286e-05,
+      "loss": 1.4559,
+      "step": 14388
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0235525714285714e-05,
+      "loss": 0.92,
+      "step": 14389
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0235257142857144e-05,
+      "loss": 0.9135,
+      "step": 14390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0234988571428572e-05,
+      "loss": 1.1074,
+      "step": 14391
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.023472e-05,
+      "loss": 1.3895,
+      "step": 14392
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.023445142857143e-05,
+      "loss": 1.3574,
+      "step": 14393
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0234182857142857e-05,
+      "loss": 1.5033,
+      "step": 14394
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0233914285714286e-05,
+      "loss": 1.158,
+      "step": 14395
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0233645714285715e-05,
+      "loss": 1.2708,
+      "step": 14396
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0233377142857144e-05,
+      "loss": 1.0015,
+      "step": 14397
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0233108571428571e-05,
+      "loss": 1.2288,
+      "step": 14398
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.023284e-05,
+      "loss": 1.4587,
+      "step": 14399
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0232571428571429e-05,
+      "loss": 0.9374,
+      "step": 14400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0232302857142857e-05,
+      "loss": 1.3705,
+      "step": 14401
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0232034285714285e-05,
+      "loss": 1.1174,
+      "step": 14402
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0231765714285715e-05,
+      "loss": 1.2878,
+      "step": 14403
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0231497142857143e-05,
+      "loss": 1.0729,
+      "step": 14404
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0231228571428572e-05,
+      "loss": 1.1559,
+      "step": 14405
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0230960000000001e-05,
+      "loss": 1.127,
+      "step": 14406
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0230691428571428e-05,
+      "loss": 1.2023,
+      "step": 14407
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0230422857142858e-05,
+      "loss": 1.0955,
+      "step": 14408
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0230154285714286e-05,
+      "loss": 1.3519,
+      "step": 14409
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0229885714285716e-05,
+      "loss": 1.1268,
+      "step": 14410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0229617142857142e-05,
+      "loss": 1.2902,
+      "step": 14411
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0229348571428572e-05,
+      "loss": 0.9314,
+      "step": 14412
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.022908e-05,
+      "loss": 0.8193,
+      "step": 14413
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0228811428571429e-05,
+      "loss": 1.0781,
+      "step": 14414
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0228542857142857e-05,
+      "loss": 1.2948,
+      "step": 14415
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0228274285714287e-05,
+      "loss": 1.402,
+      "step": 14416
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0228005714285715e-05,
+      "loss": 1.0353,
+      "step": 14417
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0227737142857143e-05,
+      "loss": 1.0894,
+      "step": 14418
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0227468571428573e-05,
+      "loss": 0.9286,
+      "step": 14419
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0227200000000001e-05,
+      "loss": 1.0232,
+      "step": 14420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0226931428571429e-05,
+      "loss": 1.2612,
+      "step": 14421
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0226662857142857e-05,
+      "loss": 1.0781,
+      "step": 14422
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0226394285714287e-05,
+      "loss": 1.193,
+      "step": 14423
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0226125714285714e-05,
+      "loss": 1.1842,
+      "step": 14424
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0225857142857144e-05,
+      "loss": 1.366,
+      "step": 14425
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0225588571428572e-05,
+      "loss": 1.5705,
+      "step": 14426
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.022532e-05,
+      "loss": 1.5197,
+      "step": 14427
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0225051428571428e-05,
+      "loss": 1.161,
+      "step": 14428
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0224782857142858e-05,
+      "loss": 1.1806,
+      "step": 14429
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0224514285714286e-05,
+      "loss": 1.646,
+      "step": 14430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0224245714285714e-05,
+      "loss": 1.4327,
+      "step": 14431
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0223977142857144e-05,
+      "loss": 1.4368,
+      "step": 14432
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0223708571428572e-05,
+      "loss": 1.066,
+      "step": 14433
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.022344e-05,
+      "loss": 1.4411,
+      "step": 14434
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0223171428571429e-05,
+      "loss": 1.1245,
+      "step": 14435
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0222902857142859e-05,
+      "loss": 1.0015,
+      "step": 14436
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0222634285714285e-05,
+      "loss": 1.0788,
+      "step": 14437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0222365714285715e-05,
+      "loss": 1.5609,
+      "step": 14438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0222097142857143e-05,
+      "loss": 1.5557,
+      "step": 14439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0221828571428571e-05,
+      "loss": 0.9273,
+      "step": 14440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.022156e-05,
+      "loss": 0.7717,
+      "step": 14441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.022129142857143e-05,
+      "loss": 1.4838,
+      "step": 14442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0221022857142857e-05,
+      "loss": 1.5162,
+      "step": 14443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0220754285714286e-05,
+      "loss": 1.2149,
+      "step": 14444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0220485714285716e-05,
+      "loss": 0.8111,
+      "step": 14445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0220217142857144e-05,
+      "loss": 1.2176,
+      "step": 14446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0219948571428572e-05,
+      "loss": 1.2051,
+      "step": 14447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.021968e-05,
+      "loss": 0.7976,
+      "step": 14448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.021941142857143e-05,
+      "loss": 1.2167,
+      "step": 14449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0219142857142856e-05,
+      "loss": 1.0254,
+      "step": 14450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0218874285714286e-05,
+      "loss": 0.9935,
+      "step": 14451
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0218605714285714e-05,
+      "loss": 1.5872,
+      "step": 14452
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0218337142857143e-05,
+      "loss": 0.8333,
+      "step": 14453
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.021806857142857e-05,
+      "loss": 1.1911,
+      "step": 14454
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.02178e-05,
+      "loss": 1.5087,
+      "step": 14455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0217531428571429e-05,
+      "loss": 1.2125,
+      "step": 14456
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0217262857142857e-05,
+      "loss": 1.3717,
+      "step": 14457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0216994285714287e-05,
+      "loss": 1.2859,
+      "step": 14458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0216725714285715e-05,
+      "loss": 1.0959,
+      "step": 14459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0216457142857143e-05,
+      "loss": 1.0305,
+      "step": 14460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0216188571428571e-05,
+      "loss": 0.8738,
+      "step": 14461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0215920000000001e-05,
+      "loss": 1.0834,
+      "step": 14462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0215651428571428e-05,
+      "loss": 0.9405,
+      "step": 14463
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0215382857142858e-05,
+      "loss": 1.3494,
+      "step": 14464
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0215114285714286e-05,
+      "loss": 1.3173,
+      "step": 14465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0214845714285716e-05,
+      "loss": 1.0949,
+      "step": 14466
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0214577142857142e-05,
+      "loss": 1.1993,
+      "step": 14467
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0214308571428572e-05,
+      "loss": 1.3808,
+      "step": 14468
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.021404e-05,
+      "loss": 1.9357,
+      "step": 14469
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0213771428571428e-05,
+      "loss": 1.0995,
+      "step": 14470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0213502857142858e-05,
+      "loss": 0.9917,
+      "step": 14471
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0213234285714286e-05,
+      "loss": 0.9709,
+      "step": 14472
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0212965714285715e-05,
+      "loss": 1.4139,
+      "step": 14473
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0212697142857143e-05,
+      "loss": 1.2293,
+      "step": 14474
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0212428571428573e-05,
+      "loss": 1.3025,
+      "step": 14475
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0212159999999999e-05,
+      "loss": 1.185,
+      "step": 14476
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0211891428571429e-05,
+      "loss": 1.6334,
+      "step": 14477
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0211622857142857e-05,
+      "loss": 1.1531,
+      "step": 14478
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0211354285714287e-05,
+      "loss": 1.3803,
+      "step": 14479
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0211085714285714e-05,
+      "loss": 1.191,
+      "step": 14480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0210817142857143e-05,
+      "loss": 1.0131,
+      "step": 14481
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0210548571428572e-05,
+      "loss": 1.0178,
+      "step": 14482
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.021028e-05,
+      "loss": 1.1764,
+      "step": 14483
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.021001142857143e-05,
+      "loss": 1.0646,
+      "step": 14484
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0209742857142858e-05,
+      "loss": 1.2668,
+      "step": 14485
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0209474285714286e-05,
+      "loss": 1.1597,
+      "step": 14486
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0209205714285714e-05,
+      "loss": 1.6217,
+      "step": 14487
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0208937142857144e-05,
+      "loss": 1.2187,
+      "step": 14488
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0208668571428572e-05,
+      "loss": 1.3864,
+      "step": 14489
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.02084e-05,
+      "loss": 1.2192,
+      "step": 14490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0208131428571429e-05,
+      "loss": 0.9373,
+      "step": 14491
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0207862857142858e-05,
+      "loss": 0.919,
+      "step": 14492
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0207594285714285e-05,
+      "loss": 1.4313,
+      "step": 14493
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0207325714285715e-05,
+      "loss": 1.693,
+      "step": 14494
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0207057142857143e-05,
+      "loss": 1.4676,
+      "step": 14495
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0206788571428571e-05,
+      "loss": 1.143,
+      "step": 14496
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0206520000000001e-05,
+      "loss": 1.2318,
+      "step": 14497
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.020625142857143e-05,
+      "loss": 1.5035,
+      "step": 14498
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0205982857142857e-05,
+      "loss": 1.1749,
+      "step": 14499
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0205714285714286e-05,
+      "loss": 0.9084,
+      "step": 14500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0205445714285715e-05,
+      "loss": 1.7467,
+      "step": 14501
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0205177142857144e-05,
+      "loss": 1.4007,
+      "step": 14502
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0204908571428572e-05,
+      "loss": 0.8626,
+      "step": 14503
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.020464e-05,
+      "loss": 1.5303,
+      "step": 14504
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.020437142857143e-05,
+      "loss": 1.5145,
+      "step": 14505
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0204102857142856e-05,
+      "loss": 0.9726,
+      "step": 14506
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0203834285714286e-05,
+      "loss": 1.4422,
+      "step": 14507
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0203565714285714e-05,
+      "loss": 1.6487,
+      "step": 14508
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0203297142857143e-05,
+      "loss": 1.3721,
+      "step": 14509
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0203028571428572e-05,
+      "loss": 1.104,
+      "step": 14510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.020276e-05,
+      "loss": 1.3285,
+      "step": 14511
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.020249142857143e-05,
+      "loss": 1.2405,
+      "step": 14512
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0202222857142857e-05,
+      "loss": 1.2461,
+      "step": 14513
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0201954285714287e-05,
+      "loss": 1.4201,
+      "step": 14514
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0201685714285715e-05,
+      "loss": 1.3454,
+      "step": 14515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0201417142857143e-05,
+      "loss": 0.9075,
+      "step": 14516
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0201148571428571e-05,
+      "loss": 1.2152,
+      "step": 14517
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0200880000000001e-05,
+      "loss": 0.7138,
+      "step": 14518
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0200611428571428e-05,
+      "loss": 0.7515,
+      "step": 14519
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0200342857142858e-05,
+      "loss": 0.8483,
+      "step": 14520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0200074285714286e-05,
+      "loss": 0.9603,
+      "step": 14521
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0199805714285714e-05,
+      "loss": 1.3221,
+      "step": 14522
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0199537142857144e-05,
+      "loss": 0.9421,
+      "step": 14523
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0199268571428572e-05,
+      "loss": 1.2838,
+      "step": 14524
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0199000000000002e-05,
+      "loss": 1.0516,
+      "step": 14525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0198731428571428e-05,
+      "loss": 0.9592,
+      "step": 14526
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0198462857142858e-05,
+      "loss": 1.6391,
+      "step": 14527
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0198194285714286e-05,
+      "loss": 1.5664,
+      "step": 14528
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0197925714285714e-05,
+      "loss": 1.0522,
+      "step": 14529
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0197657142857143e-05,
+      "loss": 1.1653,
+      "step": 14530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0197388571428573e-05,
+      "loss": 1.1521,
+      "step": 14531
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0197119999999999e-05,
+      "loss": 1.055,
+      "step": 14532
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0196851428571429e-05,
+      "loss": 1.2644,
+      "step": 14533
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0196582857142857e-05,
+      "loss": 1.0962,
+      "step": 14534
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0196314285714287e-05,
+      "loss": 1.2616,
+      "step": 14535
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0196045714285715e-05,
+      "loss": 1.3591,
+      "step": 14536
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0195777142857143e-05,
+      "loss": 0.9483,
+      "step": 14537
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0195508571428573e-05,
+      "loss": 1.0757,
+      "step": 14538
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.019524e-05,
+      "loss": 1.2639,
+      "step": 14539
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.019497142857143e-05,
+      "loss": 1.2184,
+      "step": 14540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0194702857142858e-05,
+      "loss": 1.6258,
+      "step": 14541
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0194434285714286e-05,
+      "loss": 1.7188,
+      "step": 14542
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0194165714285714e-05,
+      "loss": 1.1473,
+      "step": 14543
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0193897142857144e-05,
+      "loss": 1.4805,
+      "step": 14544
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.019362857142857e-05,
+      "loss": 1.4046,
+      "step": 14545
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.019336e-05,
+      "loss": 0.8737,
+      "step": 14546
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0193091428571428e-05,
+      "loss": 0.9317,
+      "step": 14547
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0192822857142858e-05,
+      "loss": 1.5508,
+      "step": 14548
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0192554285714286e-05,
+      "loss": 1.2824,
+      "step": 14549
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0192285714285715e-05,
+      "loss": 1.0878,
+      "step": 14550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0192017142857145e-05,
+      "loss": 1.3799,
+      "step": 14551
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0191748571428571e-05,
+      "loss": 1.023,
+      "step": 14552
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0191480000000001e-05,
+      "loss": 1.6045,
+      "step": 14553
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0191211428571429e-05,
+      "loss": 1.6609,
+      "step": 14554
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0190942857142857e-05,
+      "loss": 1.3884,
+      "step": 14555
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0190674285714285e-05,
+      "loss": 1.1387,
+      "step": 14556
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0190405714285715e-05,
+      "loss": 1.7902,
+      "step": 14557
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0190137142857142e-05,
+      "loss": 1.1767,
+      "step": 14558
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0189868571428572e-05,
+      "loss": 1.3359,
+      "step": 14559
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.01896e-05,
+      "loss": 0.9626,
+      "step": 14560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.018933142857143e-05,
+      "loss": 1.0214,
+      "step": 14561
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0189062857142858e-05,
+      "loss": 1.2467,
+      "step": 14562
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0188794285714286e-05,
+      "loss": 1.0488,
+      "step": 14563
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0188525714285716e-05,
+      "loss": 1.3921,
+      "step": 14564
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0188257142857142e-05,
+      "loss": 1.5012,
+      "step": 14565
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0187988571428572e-05,
+      "loss": 0.7018,
+      "step": 14566
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.018772e-05,
+      "loss": 1.3339,
+      "step": 14567
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0187451428571429e-05,
+      "loss": 1.1469,
+      "step": 14568
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0187182857142857e-05,
+      "loss": 1.2106,
+      "step": 14569
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0186914285714287e-05,
+      "loss": 1.1175,
+      "step": 14570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0186645714285715e-05,
+      "loss": 1.3284,
+      "step": 14571
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0186377142857143e-05,
+      "loss": 1.2205,
+      "step": 14572
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0186108571428571e-05,
+      "loss": 1.0593,
+      "step": 14573
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0185840000000001e-05,
+      "loss": 1.2402,
+      "step": 14574
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.018557142857143e-05,
+      "loss": 1.4445,
+      "step": 14575
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0185302857142857e-05,
+      "loss": 0.9671,
+      "step": 14576
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0185034285714287e-05,
+      "loss": 1.4493,
+      "step": 14577
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0184765714285714e-05,
+      "loss": 1.1423,
+      "step": 14578
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0184497142857144e-05,
+      "loss": 1.1681,
+      "step": 14579
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0184228571428572e-05,
+      "loss": 1.3166,
+      "step": 14580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.018396e-05,
+      "loss": 0.8218,
+      "step": 14581
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0183691428571428e-05,
+      "loss": 1.4718,
+      "step": 14582
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0183422857142858e-05,
+      "loss": 1.444,
+      "step": 14583
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0183154285714286e-05,
+      "loss": 1.0744,
+      "step": 14584
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0182885714285714e-05,
+      "loss": 1.3592,
+      "step": 14585
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0182617142857143e-05,
+      "loss": 1.1986,
+      "step": 14586
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0182348571428572e-05,
+      "loss": 1.136,
+      "step": 14587
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.018208e-05,
+      "loss": 1.2795,
+      "step": 14588
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0181811428571429e-05,
+      "loss": 0.8188,
+      "step": 14589
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0181542857142859e-05,
+      "loss": 1.2312,
+      "step": 14590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0181274285714285e-05,
+      "loss": 0.8894,
+      "step": 14591
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0181005714285715e-05,
+      "loss": 0.5764,
+      "step": 14592
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0180737142857143e-05,
+      "loss": 1.2721,
+      "step": 14593
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0180468571428573e-05,
+      "loss": 0.9449,
+      "step": 14594
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.01802e-05,
+      "loss": 0.9245,
+      "step": 14595
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.017993142857143e-05,
+      "loss": 1.2681,
+      "step": 14596
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0179662857142858e-05,
+      "loss": 1.2382,
+      "step": 14597
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0179394285714286e-05,
+      "loss": 1.0913,
+      "step": 14598
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0179125714285714e-05,
+      "loss": 1.7777,
+      "step": 14599
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0178857142857144e-05,
+      "loss": 1.09,
+      "step": 14600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0178588571428572e-05,
+      "loss": 1.4471,
+      "step": 14601
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.017832e-05,
+      "loss": 1.3993,
+      "step": 14602
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.017805142857143e-05,
+      "loss": 1.2687,
+      "step": 14603
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0177782857142857e-05,
+      "loss": 1.2367,
+      "step": 14604
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0177514285714286e-05,
+      "loss": 0.8002,
+      "step": 14605
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0177245714285715e-05,
+      "loss": 0.814,
+      "step": 14606
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0176977142857144e-05,
+      "loss": 1.0181,
+      "step": 14607
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0176708571428571e-05,
+      "loss": 1.4694,
+      "step": 14608
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.017644e-05,
+      "loss": 1.3058,
+      "step": 14609
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0176171428571429e-05,
+      "loss": 0.9817,
+      "step": 14610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0175902857142857e-05,
+      "loss": 1.0817,
+      "step": 14611
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0175634285714285e-05,
+      "loss": 1.2523,
+      "step": 14612
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0175365714285715e-05,
+      "loss": 1.0128,
+      "step": 14613
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0175097142857143e-05,
+      "loss": 1.0344,
+      "step": 14614
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0174828571428572e-05,
+      "loss": 1.2524,
+      "step": 14615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0174560000000001e-05,
+      "loss": 1.3743,
+      "step": 14616
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.017429142857143e-05,
+      "loss": 0.6979,
+      "step": 14617
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0174022857142858e-05,
+      "loss": 1.4631,
+      "step": 14618
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0173754285714286e-05,
+      "loss": 0.8216,
+      "step": 14619
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0173485714285716e-05,
+      "loss": 1.3024,
+      "step": 14620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0173217142857142e-05,
+      "loss": 1.1014,
+      "step": 14621
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0172948571428572e-05,
+      "loss": 1.1019,
+      "step": 14622
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.017268e-05,
+      "loss": 1.1904,
+      "step": 14623
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0172411428571428e-05,
+      "loss": 1.1745,
+      "step": 14624
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0172142857142857e-05,
+      "loss": 0.9866,
+      "step": 14625
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0171874285714287e-05,
+      "loss": 1.1039,
+      "step": 14626
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0171605714285715e-05,
+      "loss": 1.0277,
+      "step": 14627
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0171337142857143e-05,
+      "loss": 0.6247,
+      "step": 14628
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0171068571428573e-05,
+      "loss": 1.2922,
+      "step": 14629
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0170800000000001e-05,
+      "loss": 1.0153,
+      "step": 14630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0170531428571429e-05,
+      "loss": 1.0155,
+      "step": 14631
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0170262857142857e-05,
+      "loss": 1.0576,
+      "step": 14632
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0169994285714287e-05,
+      "loss": 1.1346,
+      "step": 14633
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0169725714285714e-05,
+      "loss": 1.1744,
+      "step": 14634
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0169457142857144e-05,
+      "loss": 1.573,
+      "step": 14635
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0169188571428572e-05,
+      "loss": 1.3786,
+      "step": 14636
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.016892e-05,
+      "loss": 0.912,
+      "step": 14637
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0168651428571428e-05,
+      "loss": 1.4312,
+      "step": 14638
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0168382857142858e-05,
+      "loss": 1.3127,
+      "step": 14639
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0168114285714286e-05,
+      "loss": 1.1941,
+      "step": 14640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0167845714285714e-05,
+      "loss": 0.8089,
+      "step": 14641
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0167577142857144e-05,
+      "loss": 1.1829,
+      "step": 14642
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0167308571428572e-05,
+      "loss": 0.95,
+      "step": 14643
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.016704e-05,
+      "loss": 1.4825,
+      "step": 14644
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0166771428571429e-05,
+      "loss": 1.2463,
+      "step": 14645
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0166502857142859e-05,
+      "loss": 1.2023,
+      "step": 14646
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0166234285714285e-05,
+      "loss": 1.2275,
+      "step": 14647
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0165965714285715e-05,
+      "loss": 1.3344,
+      "step": 14648
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0165697142857143e-05,
+      "loss": 1.2434,
+      "step": 14649
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0165428571428571e-05,
+      "loss": 1.5637,
+      "step": 14650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.016516e-05,
+      "loss": 1.0003,
+      "step": 14651
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.016489142857143e-05,
+      "loss": 1.1675,
+      "step": 14652
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0164622857142857e-05,
+      "loss": 1.3987,
+      "step": 14653
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0164354285714286e-05,
+      "loss": 0.8967,
+      "step": 14654
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0164085714285716e-05,
+      "loss": 1.1136,
+      "step": 14655
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0163817142857144e-05,
+      "loss": 1.0695,
+      "step": 14656
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0163548571428572e-05,
+      "loss": 1.0263,
+      "step": 14657
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.016328e-05,
+      "loss": 0.929,
+      "step": 14658
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.016301142857143e-05,
+      "loss": 0.9369,
+      "step": 14659
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0162742857142856e-05,
+      "loss": 1.4719,
+      "step": 14660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0162474285714286e-05,
+      "loss": 1.2054,
+      "step": 14661
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0162205714285714e-05,
+      "loss": 0.6326,
+      "step": 14662
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0161937142857144e-05,
+      "loss": 1.2737,
+      "step": 14663
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.016166857142857e-05,
+      "loss": 1.1323,
+      "step": 14664
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.01614e-05,
+      "loss": 1.3733,
+      "step": 14665
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0161131428571429e-05,
+      "loss": 0.9084,
+      "step": 14666
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0160862857142857e-05,
+      "loss": 1.5166,
+      "step": 14667
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0160594285714287e-05,
+      "loss": 1.1857,
+      "step": 14668
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0160325714285715e-05,
+      "loss": 1.4128,
+      "step": 14669
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0160057142857143e-05,
+      "loss": 1.1707,
+      "step": 14670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0159788571428571e-05,
+      "loss": 0.8519,
+      "step": 14671
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0159520000000001e-05,
+      "loss": 1.2741,
+      "step": 14672
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0159251428571428e-05,
+      "loss": 1.5288,
+      "step": 14673
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0158982857142858e-05,
+      "loss": 0.9456,
+      "step": 14674
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0158714285714286e-05,
+      "loss": 1.2811,
+      "step": 14675
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0158445714285716e-05,
+      "loss": 1.1095,
+      "step": 14676
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0158177142857142e-05,
+      "loss": 1.0987,
+      "step": 14677
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0157908571428572e-05,
+      "loss": 1.4706,
+      "step": 14678
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.015764e-05,
+      "loss": 0.8763,
+      "step": 14679
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0157371428571428e-05,
+      "loss": 1.1894,
+      "step": 14680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0157102857142858e-05,
+      "loss": 0.9995,
+      "step": 14681
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0156834285714286e-05,
+      "loss": 1.341,
+      "step": 14682
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0156565714285715e-05,
+      "loss": 1.088,
+      "step": 14683
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0156297142857143e-05,
+      "loss": 1.2642,
+      "step": 14684
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0156028571428573e-05,
+      "loss": 1.3125,
+      "step": 14685
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.015576e-05,
+      "loss": 1.4138,
+      "step": 14686
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0155491428571429e-05,
+      "loss": 1.2645,
+      "step": 14687
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0155222857142857e-05,
+      "loss": 0.8731,
+      "step": 14688
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0154954285714287e-05,
+      "loss": 1.1384,
+      "step": 14689
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0154685714285714e-05,
+      "loss": 1.1266,
+      "step": 14690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0154417142857143e-05,
+      "loss": 1.166,
+      "step": 14691
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0154148571428572e-05,
+      "loss": 1.1158,
+      "step": 14692
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.015388e-05,
+      "loss": 1.0099,
+      "step": 14693
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.015361142857143e-05,
+      "loss": 1.2789,
+      "step": 14694
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0153342857142858e-05,
+      "loss": 1.0094,
+      "step": 14695
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0153074285714286e-05,
+      "loss": 1.5866,
+      "step": 14696
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0152805714285714e-05,
+      "loss": 1.5321,
+      "step": 14697
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0152537142857144e-05,
+      "loss": 1.4049,
+      "step": 14698
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0152268571428572e-05,
+      "loss": 0.7865,
+      "step": 14699
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0152e-05,
+      "loss": 1.2059,
+      "step": 14700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0151731428571429e-05,
+      "loss": 0.8872,
+      "step": 14701
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0151462857142858e-05,
+      "loss": 1.2639,
+      "step": 14702
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0151194285714285e-05,
+      "loss": 1.1599,
+      "step": 14703
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0150925714285715e-05,
+      "loss": 1.0232,
+      "step": 14704
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0150657142857143e-05,
+      "loss": 1.0733,
+      "step": 14705
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0150388571428571e-05,
+      "loss": 1.3951,
+      "step": 14706
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0150120000000001e-05,
+      "loss": 1.2843,
+      "step": 14707
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.014985142857143e-05,
+      "loss": 1.122,
+      "step": 14708
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0149582857142857e-05,
+      "loss": 1.8047,
+      "step": 14709
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0149314285714286e-05,
+      "loss": 0.8492,
+      "step": 14710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0149045714285715e-05,
+      "loss": 1.0641,
+      "step": 14711
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0148777142857144e-05,
+      "loss": 1.0633,
+      "step": 14712
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0148508571428572e-05,
+      "loss": 1.6005,
+      "step": 14713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.014824e-05,
+      "loss": 1.1146,
+      "step": 14714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.014797142857143e-05,
+      "loss": 0.5832,
+      "step": 14715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0147702857142856e-05,
+      "loss": 1.1278,
+      "step": 14716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0147434285714286e-05,
+      "loss": 1.12,
+      "step": 14717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0147165714285714e-05,
+      "loss": 1.4991,
+      "step": 14718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0146897142857143e-05,
+      "loss": 1.4084,
+      "step": 14719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0146628571428572e-05,
+      "loss": 1.473,
+      "step": 14720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.014636e-05,
+      "loss": 1.096,
+      "step": 14721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.014609142857143e-05,
+      "loss": 0.704,
+      "step": 14722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0145822857142857e-05,
+      "loss": 1.6712,
+      "step": 14723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0145554285714287e-05,
+      "loss": 1.1803,
+      "step": 14724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0145285714285715e-05,
+      "loss": 1.1594,
+      "step": 14725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0145017142857143e-05,
+      "loss": 0.9992,
+      "step": 14726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0144748571428571e-05,
+      "loss": 1.4047,
+      "step": 14727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0144480000000001e-05,
+      "loss": 1.4073,
+      "step": 14728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0144211428571428e-05,
+      "loss": 1.2387,
+      "step": 14729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0143942857142858e-05,
+      "loss": 1.0264,
+      "step": 14730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0143674285714286e-05,
+      "loss": 1.2431,
+      "step": 14731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0143405714285714e-05,
+      "loss": 1.1276,
+      "step": 14732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0143137142857144e-05,
+      "loss": 0.9234,
+      "step": 14733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0142868571428572e-05,
+      "loss": 0.8796,
+      "step": 14734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0142600000000002e-05,
+      "loss": 1.0771,
+      "step": 14735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0142331428571428e-05,
+      "loss": 1.3639,
+      "step": 14736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0142062857142858e-05,
+      "loss": 1.3466,
+      "step": 14737
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0141794285714286e-05,
+      "loss": 1.5042,
+      "step": 14738
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0141525714285714e-05,
+      "loss": 1.34,
+      "step": 14739
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0141257142857143e-05,
+      "loss": 1.5268,
+      "step": 14740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0140988571428573e-05,
+      "loss": 1.0827,
+      "step": 14741
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0140719999999999e-05,
+      "loss": 0.9501,
+      "step": 14742
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0140451428571429e-05,
+      "loss": 1.2678,
+      "step": 14743
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0140182857142857e-05,
+      "loss": 0.7268,
+      "step": 14744
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0139914285714287e-05,
+      "loss": 1.0561,
+      "step": 14745
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0139645714285715e-05,
+      "loss": 0.952,
+      "step": 14746
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0139377142857143e-05,
+      "loss": 0.999,
+      "step": 14747
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0139108571428573e-05,
+      "loss": 0.9218,
+      "step": 14748
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.013884e-05,
+      "loss": 0.6559,
+      "step": 14749
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.013857142857143e-05,
+      "loss": 0.8509,
+      "step": 14750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0138302857142858e-05,
+      "loss": 1.411,
+      "step": 14751
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0138034285714286e-05,
+      "loss": 1.0482,
+      "step": 14752
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0137765714285714e-05,
+      "loss": 1.3419,
+      "step": 14753
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0137497142857144e-05,
+      "loss": 1.2426,
+      "step": 14754
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.013722857142857e-05,
+      "loss": 1.1568,
+      "step": 14755
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.013696e-05,
+      "loss": 1.2826,
+      "step": 14756
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0136691428571428e-05,
+      "loss": 1.3744,
+      "step": 14757
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0136422857142858e-05,
+      "loss": 1.2756,
+      "step": 14758
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0136154285714286e-05,
+      "loss": 0.8378,
+      "step": 14759
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0135885714285715e-05,
+      "loss": 1.1997,
+      "step": 14760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0135617142857145e-05,
+      "loss": 1.3386,
+      "step": 14761
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0135348571428571e-05,
+      "loss": 1.2332,
+      "step": 14762
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0135080000000001e-05,
+      "loss": 1.2337,
+      "step": 14763
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0134811428571429e-05,
+      "loss": 1.2552,
+      "step": 14764
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0134542857142857e-05,
+      "loss": 1.0667,
+      "step": 14765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0134274285714285e-05,
+      "loss": 1.4443,
+      "step": 14766
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0134005714285715e-05,
+      "loss": 0.8703,
+      "step": 14767
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0133737142857143e-05,
+      "loss": 1.1504,
+      "step": 14768
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0133468571428572e-05,
+      "loss": 1.3563,
+      "step": 14769
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.01332e-05,
+      "loss": 0.8159,
+      "step": 14770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.013293142857143e-05,
+      "loss": 0.8106,
+      "step": 14771
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0132662857142858e-05,
+      "loss": 1.3333,
+      "step": 14772
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0132394285714286e-05,
+      "loss": 1.4509,
+      "step": 14773
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0132125714285716e-05,
+      "loss": 0.9727,
+      "step": 14774
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0131857142857142e-05,
+      "loss": 1.135,
+      "step": 14775
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0131588571428572e-05,
+      "loss": 1.0823,
+      "step": 14776
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.013132e-05,
+      "loss": 0.9631,
+      "step": 14777
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0131051428571429e-05,
+      "loss": 1.2094,
+      "step": 14778
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0130782857142857e-05,
+      "loss": 1.3556,
+      "step": 14779
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0130514285714287e-05,
+      "loss": 1.2512,
+      "step": 14780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0130245714285715e-05,
+      "loss": 1.1744,
+      "step": 14781
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0129977142857143e-05,
+      "loss": 1.0466,
+      "step": 14782
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0129708571428571e-05,
+      "loss": 1.218,
+      "step": 14783
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0129440000000001e-05,
+      "loss": 1.4722,
+      "step": 14784
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.012917142857143e-05,
+      "loss": 1.4102,
+      "step": 14785
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0128902857142857e-05,
+      "loss": 1.2567,
+      "step": 14786
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0128634285714287e-05,
+      "loss": 0.8165,
+      "step": 14787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0128365714285714e-05,
+      "loss": 1.6383,
+      "step": 14788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0128097142857144e-05,
+      "loss": 1.2357,
+      "step": 14789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0127828571428572e-05,
+      "loss": 1.3159,
+      "step": 14790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0127560000000002e-05,
+      "loss": 1.212,
+      "step": 14791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0127291428571428e-05,
+      "loss": 1.5425,
+      "step": 14792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0127022857142858e-05,
+      "loss": 1.3483,
+      "step": 14793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0126754285714286e-05,
+      "loss": 1.6428,
+      "step": 14794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0126485714285714e-05,
+      "loss": 1.2098,
+      "step": 14795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0126217142857143e-05,
+      "loss": 1.5157,
+      "step": 14796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0125948571428572e-05,
+      "loss": 1.4155,
+      "step": 14797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.012568e-05,
+      "loss": 1.217,
+      "step": 14798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0125411428571429e-05,
+      "loss": 1.1034,
+      "step": 14799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0125142857142859e-05,
+      "loss": 1.0561,
+      "step": 14800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0124874285714285e-05,
+      "loss": 1.0142,
+      "step": 14801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0124605714285715e-05,
+      "loss": 1.1634,
+      "step": 14802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0124337142857143e-05,
+      "loss": 0.9444,
+      "step": 14803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0124068571428573e-05,
+      "loss": 1.3984,
+      "step": 14804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.01238e-05,
+      "loss": 1.1785,
+      "step": 14805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.012353142857143e-05,
+      "loss": 1.2939,
+      "step": 14806
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0123262857142858e-05,
+      "loss": 1.2928,
+      "step": 14807
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0122994285714286e-05,
+      "loss": 0.9456,
+      "step": 14808
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0122725714285714e-05,
+      "loss": 1.0515,
+      "step": 14809
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0122457142857144e-05,
+      "loss": 1.1028,
+      "step": 14810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0122188571428572e-05,
+      "loss": 0.87,
+      "step": 14811
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.012192e-05,
+      "loss": 1.4237,
+      "step": 14812
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.012165142857143e-05,
+      "loss": 1.1813,
+      "step": 14813
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0121382857142858e-05,
+      "loss": 1.0612,
+      "step": 14814
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0121114285714286e-05,
+      "loss": 0.9403,
+      "step": 14815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0120845714285715e-05,
+      "loss": 1.3269,
+      "step": 14816
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0120577142857144e-05,
+      "loss": 1.1862,
+      "step": 14817
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0120308571428571e-05,
+      "loss": 1.3181,
+      "step": 14818
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.012004e-05,
+      "loss": 0.9045,
+      "step": 14819
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0119771428571429e-05,
+      "loss": 1.0263,
+      "step": 14820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0119502857142857e-05,
+      "loss": 1.0293,
+      "step": 14821
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0119234285714285e-05,
+      "loss": 1.2628,
+      "step": 14822
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0118965714285715e-05,
+      "loss": 1.2996,
+      "step": 14823
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0118697142857143e-05,
+      "loss": 0.8787,
+      "step": 14824
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0118428571428572e-05,
+      "loss": 0.997,
+      "step": 14825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0118160000000001e-05,
+      "loss": 0.8804,
+      "step": 14826
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.011789142857143e-05,
+      "loss": 1.2745,
+      "step": 14827
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0117622857142858e-05,
+      "loss": 1.054,
+      "step": 14828
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0117354285714286e-05,
+      "loss": 1.5465,
+      "step": 14829
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0117085714285716e-05,
+      "loss": 1.2409,
+      "step": 14830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0116817142857142e-05,
+      "loss": 1.145,
+      "step": 14831
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0116548571428572e-05,
+      "loss": 1.247,
+      "step": 14832
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.011628e-05,
+      "loss": 1.0251,
+      "step": 14833
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0116011428571428e-05,
+      "loss": 1.3779,
+      "step": 14834
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0115742857142857e-05,
+      "loss": 1.4359,
+      "step": 14835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0115474285714287e-05,
+      "loss": 1.3402,
+      "step": 14836
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0115205714285715e-05,
+      "loss": 1.08,
+      "step": 14837
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0114937142857143e-05,
+      "loss": 1.1339,
+      "step": 14838
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0114668571428573e-05,
+      "loss": 1.6265,
+      "step": 14839
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0114400000000001e-05,
+      "loss": 1.3318,
+      "step": 14840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0114131428571429e-05,
+      "loss": 1.1628,
+      "step": 14841
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0113862857142857e-05,
+      "loss": 0.9717,
+      "step": 14842
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0113594285714287e-05,
+      "loss": 1.2312,
+      "step": 14843
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0113325714285714e-05,
+      "loss": 1.1405,
+      "step": 14844
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0113057142857144e-05,
+      "loss": 1.1448,
+      "step": 14845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0112788571428572e-05,
+      "loss": 0.9438,
+      "step": 14846
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.011252e-05,
+      "loss": 0.8924,
+      "step": 14847
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0112251428571428e-05,
+      "loss": 1.4508,
+      "step": 14848
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0111982857142858e-05,
+      "loss": 1.5024,
+      "step": 14849
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0111714285714286e-05,
+      "loss": 0.9671,
+      "step": 14850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0111445714285714e-05,
+      "loss": 1.552,
+      "step": 14851
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0111177142857144e-05,
+      "loss": 1.1512,
+      "step": 14852
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0110908571428572e-05,
+      "loss": 1.2546,
+      "step": 14853
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.011064e-05,
+      "loss": 1.5332,
+      "step": 14854
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0110371428571429e-05,
+      "loss": 0.9838,
+      "step": 14855
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0110102857142859e-05,
+      "loss": 1.2969,
+      "step": 14856
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0109834285714285e-05,
+      "loss": 1.2375,
+      "step": 14857
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0109565714285715e-05,
+      "loss": 1.3463,
+      "step": 14858
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0109297142857143e-05,
+      "loss": 1.8685,
+      "step": 14859
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0109028571428573e-05,
+      "loss": 1.083,
+      "step": 14860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.010876e-05,
+      "loss": 0.8539,
+      "step": 14861
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.010849142857143e-05,
+      "loss": 0.8494,
+      "step": 14862
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0108222857142857e-05,
+      "loss": 0.9065,
+      "step": 14863
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0107954285714286e-05,
+      "loss": 1.0893,
+      "step": 14864
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0107685714285716e-05,
+      "loss": 1.4636,
+      "step": 14865
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0107417142857144e-05,
+      "loss": 1.2898,
+      "step": 14866
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0107148571428572e-05,
+      "loss": 1.0426,
+      "step": 14867
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.010688e-05,
+      "loss": 1.0193,
+      "step": 14868
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.010661142857143e-05,
+      "loss": 1.5274,
+      "step": 14869
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0106342857142856e-05,
+      "loss": 1.1205,
+      "step": 14870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0106074285714286e-05,
+      "loss": 1.0395,
+      "step": 14871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0105805714285714e-05,
+      "loss": 1.0676,
+      "step": 14872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0105537142857144e-05,
+      "loss": 0.7153,
+      "step": 14873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.010526857142857e-05,
+      "loss": 1.0843,
+      "step": 14874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0105e-05,
+      "loss": 0.9741,
+      "step": 14875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0104731428571429e-05,
+      "loss": 1.1234,
+      "step": 14876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0104462857142857e-05,
+      "loss": 1.0483,
+      "step": 14877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0104194285714287e-05,
+      "loss": 1.7965,
+      "step": 14878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0103925714285715e-05,
+      "loss": 1.2382,
+      "step": 14879
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0103657142857143e-05,
+      "loss": 1.0707,
+      "step": 14880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0103388571428571e-05,
+      "loss": 0.9928,
+      "step": 14881
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0103120000000001e-05,
+      "loss": 1.1024,
+      "step": 14882
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0102851428571428e-05,
+      "loss": 1.0899,
+      "step": 14883
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0102582857142858e-05,
+      "loss": 1.02,
+      "step": 14884
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0102314285714286e-05,
+      "loss": 1.3196,
+      "step": 14885
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0102045714285716e-05,
+      "loss": 0.8362,
+      "step": 14886
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0101777142857142e-05,
+      "loss": 1.4099,
+      "step": 14887
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0101508571428572e-05,
+      "loss": 0.9926,
+      "step": 14888
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.010124e-05,
+      "loss": 1.1399,
+      "step": 14889
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0100971428571428e-05,
+      "loss": 1.3361,
+      "step": 14890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0100702857142858e-05,
+      "loss": 1.4792,
+      "step": 14891
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0100434285714286e-05,
+      "loss": 1.2152,
+      "step": 14892
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0100165714285715e-05,
+      "loss": 1.2847,
+      "step": 14893
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0099897142857143e-05,
+      "loss": 1.4267,
+      "step": 14894
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0099628571428573e-05,
+      "loss": 1.809,
+      "step": 14895
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.009936e-05,
+      "loss": 0.9716,
+      "step": 14896
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0099091428571429e-05,
+      "loss": 1.5817,
+      "step": 14897
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0098822857142857e-05,
+      "loss": 0.8483,
+      "step": 14898
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0098554285714287e-05,
+      "loss": 1.1279,
+      "step": 14899
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0098285714285714e-05,
+      "loss": 1.0595,
+      "step": 14900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0098017142857143e-05,
+      "loss": 0.9358,
+      "step": 14901
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0097748571428572e-05,
+      "loss": 1.3693,
+      "step": 14902
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.009748e-05,
+      "loss": 1.6446,
+      "step": 14903
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.009721142857143e-05,
+      "loss": 1.6221,
+      "step": 14904
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0096942857142858e-05,
+      "loss": 1.2664,
+      "step": 14905
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0096674285714286e-05,
+      "loss": 1.0218,
+      "step": 14906
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0096405714285714e-05,
+      "loss": 1.0812,
+      "step": 14907
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0096137142857144e-05,
+      "loss": 1.0412,
+      "step": 14908
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0095868571428572e-05,
+      "loss": 1.4981,
+      "step": 14909
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.00956e-05,
+      "loss": 1.1663,
+      "step": 14910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0095331428571429e-05,
+      "loss": 1.1245,
+      "step": 14911
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0095062857142858e-05,
+      "loss": 1.1175,
+      "step": 14912
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0094794285714285e-05,
+      "loss": 1.2594,
+      "step": 14913
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0094525714285715e-05,
+      "loss": 1.2375,
+      "step": 14914
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0094257142857143e-05,
+      "loss": 0.9771,
+      "step": 14915
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0093988571428571e-05,
+      "loss": 1.4505,
+      "step": 14916
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0093720000000001e-05,
+      "loss": 0.9127,
+      "step": 14917
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.009345142857143e-05,
+      "loss": 0.792,
+      "step": 14918
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0093182857142859e-05,
+      "loss": 1.3436,
+      "step": 14919
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0092914285714286e-05,
+      "loss": 1.0715,
+      "step": 14920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0092645714285715e-05,
+      "loss": 1.7878,
+      "step": 14921
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0092377142857144e-05,
+      "loss": 1.0692,
+      "step": 14922
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0092108571428572e-05,
+      "loss": 1.3951,
+      "step": 14923
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.009184e-05,
+      "loss": 0.9475,
+      "step": 14924
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.009157142857143e-05,
+      "loss": 1.2723,
+      "step": 14925
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0091302857142856e-05,
+      "loss": 0.9875,
+      "step": 14926
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0091034285714286e-05,
+      "loss": 0.9548,
+      "step": 14927
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0090765714285714e-05,
+      "loss": 1.4037,
+      "step": 14928
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0090497142857142e-05,
+      "loss": 1.2474,
+      "step": 14929
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0090228571428572e-05,
+      "loss": 0.9377,
+      "step": 14930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.008996e-05,
+      "loss": 1.3545,
+      "step": 14931
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.008969142857143e-05,
+      "loss": 1.2343,
+      "step": 14932
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0089422857142857e-05,
+      "loss": 1.0749,
+      "step": 14933
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0089154285714287e-05,
+      "loss": 1.0942,
+      "step": 14934
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0088885714285715e-05,
+      "loss": 0.7769,
+      "step": 14935
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0088617142857143e-05,
+      "loss": 0.9833,
+      "step": 14936
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0088348571428571e-05,
+      "loss": 0.9772,
+      "step": 14937
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0088080000000001e-05,
+      "loss": 0.9097,
+      "step": 14938
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0087811428571428e-05,
+      "loss": 1.1189,
+      "step": 14939
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0087542857142858e-05,
+      "loss": 1.0693,
+      "step": 14940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0087274285714286e-05,
+      "loss": 1.6035,
+      "step": 14941
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0087005714285716e-05,
+      "loss": 1.3413,
+      "step": 14942
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0086737142857142e-05,
+      "loss": 0.8983,
+      "step": 14943
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0086468571428572e-05,
+      "loss": 1.2818,
+      "step": 14944
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0086200000000002e-05,
+      "loss": 0.8475,
+      "step": 14945
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0085931428571428e-05,
+      "loss": 0.9415,
+      "step": 14946
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0085662857142858e-05,
+      "loss": 1.1493,
+      "step": 14947
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0085394285714286e-05,
+      "loss": 1.2021,
+      "step": 14948
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0085125714285714e-05,
+      "loss": 1.1468,
+      "step": 14949
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0084857142857143e-05,
+      "loss": 1.0356,
+      "step": 14950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0084588571428573e-05,
+      "loss": 1.1817,
+      "step": 14951
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0084319999999999e-05,
+      "loss": 1.7233,
+      "step": 14952
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0084051428571429e-05,
+      "loss": 1.3529,
+      "step": 14953
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0083782857142857e-05,
+      "loss": 1.2764,
+      "step": 14954
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0083514285714287e-05,
+      "loss": 1.1778,
+      "step": 14955
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0083245714285713e-05,
+      "loss": 1.0149,
+      "step": 14956
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0082977142857143e-05,
+      "loss": 1.3565,
+      "step": 14957
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0082708571428573e-05,
+      "loss": 1.0679,
+      "step": 14958
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.008244e-05,
+      "loss": 0.9939,
+      "step": 14959
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.008217142857143e-05,
+      "loss": 1.4368,
+      "step": 14960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0081902857142858e-05,
+      "loss": 1.2195,
+      "step": 14961
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0081634285714286e-05,
+      "loss": 1.2569,
+      "step": 14962
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0081365714285714e-05,
+      "loss": 1.0884,
+      "step": 14963
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0081097142857144e-05,
+      "loss": 1.2031,
+      "step": 14964
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0080828571428572e-05,
+      "loss": 1.0644,
+      "step": 14965
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.008056e-05,
+      "loss": 1.3733,
+      "step": 14966
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0080291428571428e-05,
+      "loss": 1.3772,
+      "step": 14967
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0080022857142858e-05,
+      "loss": 1.1408,
+      "step": 14968
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0079754285714285e-05,
+      "loss": 1.6069,
+      "step": 14969
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0079485714285715e-05,
+      "loss": 1.0464,
+      "step": 14970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0079217142857145e-05,
+      "loss": 0.493,
+      "step": 14971
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0078948571428571e-05,
+      "loss": 1.1043,
+      "step": 14972
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0078680000000001e-05,
+      "loss": 0.8772,
+      "step": 14973
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0078411428571429e-05,
+      "loss": 1.5731,
+      "step": 14974
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0078142857142857e-05,
+      "loss": 1.0566,
+      "step": 14975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0077874285714285e-05,
+      "loss": 1.5744,
+      "step": 14976
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0077605714285715e-05,
+      "loss": 0.8904,
+      "step": 14977
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0077337142857143e-05,
+      "loss": 1.388,
+      "step": 14978
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0077068571428572e-05,
+      "loss": 1.222,
+      "step": 14979
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.00768e-05,
+      "loss": 0.6855,
+      "step": 14980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.007653142857143e-05,
+      "loss": 1.0603,
+      "step": 14981
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0076262857142856e-05,
+      "loss": 1.0074,
+      "step": 14982
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0075994285714286e-05,
+      "loss": 1.0923,
+      "step": 14983
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0075725714285716e-05,
+      "loss": 1.0604,
+      "step": 14984
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0075457142857142e-05,
+      "loss": 1.3276,
+      "step": 14985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0075188571428572e-05,
+      "loss": 1.0146,
+      "step": 14986
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.007492e-05,
+      "loss": 1.7107,
+      "step": 14987
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.007465142857143e-05,
+      "loss": 1.5866,
+      "step": 14988
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0074382857142857e-05,
+      "loss": 0.7078,
+      "step": 14989
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0074114285714287e-05,
+      "loss": 1.3041,
+      "step": 14990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0073845714285715e-05,
+      "loss": 1.136,
+      "step": 14991
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0073577142857143e-05,
+      "loss": 1.1837,
+      "step": 14992
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0073308571428571e-05,
+      "loss": 1.329,
+      "step": 14993
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0073040000000001e-05,
+      "loss": 1.0998,
+      "step": 14994
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0072771428571428e-05,
+      "loss": 0.8739,
+      "step": 14995
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0072502857142857e-05,
+      "loss": 1.0909,
+      "step": 14996
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0072234285714287e-05,
+      "loss": 1.0625,
+      "step": 14997
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0071965714285714e-05,
+      "loss": 1.5749,
+      "step": 14998
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0071697142857144e-05,
+      "loss": 1.2464,
+      "step": 14999
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0071428571428572e-05,
+      "loss": 1.3034,
+      "step": 15000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0071160000000002e-05,
+      "loss": 0.9557,
+      "step": 15001
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0070891428571428e-05,
+      "loss": 1.2557,
+      "step": 15002
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0070622857142858e-05,
+      "loss": 1.1129,
+      "step": 15003
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0070354285714286e-05,
+      "loss": 0.8816,
+      "step": 15004
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0070085714285714e-05,
+      "loss": 1.4548,
+      "step": 15005
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0069817142857143e-05,
+      "loss": 1.3445,
+      "step": 15006
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0069548571428572e-05,
+      "loss": 1.2992,
+      "step": 15007
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0069279999999999e-05,
+      "loss": 1.1147,
+      "step": 15008
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0069011428571429e-05,
+      "loss": 1.4191,
+      "step": 15009
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0068742857142859e-05,
+      "loss": 1.0781,
+      "step": 15010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0068474285714287e-05,
+      "loss": 1.5049,
+      "step": 15011
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0068205714285715e-05,
+      "loss": 0.98,
+      "step": 15012
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0067937142857143e-05,
+      "loss": 1.0793,
+      "step": 15013
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0067668571428573e-05,
+      "loss": 1.0498,
+      "step": 15014
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.00674e-05,
+      "loss": 1.0446,
+      "step": 15015
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.006713142857143e-05,
+      "loss": 1.1252,
+      "step": 15016
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0066862857142858e-05,
+      "loss": 1.1181,
+      "step": 15017
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0066594285714286e-05,
+      "loss": 1.1278,
+      "step": 15018
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0066325714285714e-05,
+      "loss": 1.2026,
+      "step": 15019
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0066057142857144e-05,
+      "loss": 1.3796,
+      "step": 15020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.006578857142857e-05,
+      "loss": 1.0485,
+      "step": 15021
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.006552e-05,
+      "loss": 1.037,
+      "step": 15022
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.006525142857143e-05,
+      "loss": 1.3054,
+      "step": 15023
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0064982857142858e-05,
+      "loss": 0.8976,
+      "step": 15024
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0064714285714286e-05,
+      "loss": 1.7473,
+      "step": 15025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0064445714285715e-05,
+      "loss": 1.2301,
+      "step": 15026
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0064177142857144e-05,
+      "loss": 1.1637,
+      "step": 15027
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0063908571428571e-05,
+      "loss": 1.4093,
+      "step": 15028
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.006364e-05,
+      "loss": 1.4487,
+      "step": 15029
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0063371428571429e-05,
+      "loss": 1.3732,
+      "step": 15030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0063102857142857e-05,
+      "loss": 1.1961,
+      "step": 15031
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0062834285714285e-05,
+      "loss": 1.2344,
+      "step": 15032
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0062565714285715e-05,
+      "loss": 0.9504,
+      "step": 15033
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0062297142857143e-05,
+      "loss": 1.486,
+      "step": 15034
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0062028571428572e-05,
+      "loss": 0.945,
+      "step": 15035
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0061760000000001e-05,
+      "loss": 1.1966,
+      "step": 15036
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.006149142857143e-05,
+      "loss": 0.9901,
+      "step": 15037
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0061222857142858e-05,
+      "loss": 1.0191,
+      "step": 15038
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0060954285714286e-05,
+      "loss": 1.3839,
+      "step": 15039
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0060685714285716e-05,
+      "loss": 1.3707,
+      "step": 15040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0060417142857142e-05,
+      "loss": 0.9119,
+      "step": 15041
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0060148571428572e-05,
+      "loss": 1.2609,
+      "step": 15042
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005988e-05,
+      "loss": 1.1313,
+      "step": 15043
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0059611428571428e-05,
+      "loss": 1.0782,
+      "step": 15044
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0059342857142857e-05,
+      "loss": 1.1183,
+      "step": 15045
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0059074285714287e-05,
+      "loss": 1.0225,
+      "step": 15046
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0058805714285715e-05,
+      "loss": 0.9556,
+      "step": 15047
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0058537142857143e-05,
+      "loss": 0.7862,
+      "step": 15048
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0058268571428571e-05,
+      "loss": 0.9826,
+      "step": 15049
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0058000000000001e-05,
+      "loss": 1.4922,
+      "step": 15050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0057731428571429e-05,
+      "loss": 1.4029,
+      "step": 15051
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0057462857142857e-05,
+      "loss": 1.3187,
+      "step": 15052
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0057194285714287e-05,
+      "loss": 1.1926,
+      "step": 15053
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0056925714285714e-05,
+      "loss": 1.2149,
+      "step": 15054
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0056657142857144e-05,
+      "loss": 1.2299,
+      "step": 15055
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0056388571428572e-05,
+      "loss": 0.6476,
+      "step": 15056
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005612e-05,
+      "loss": 1.5471,
+      "step": 15057
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0055851428571428e-05,
+      "loss": 1.4653,
+      "step": 15058
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0055582857142858e-05,
+      "loss": 1.6668,
+      "step": 15059
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0055314285714286e-05,
+      "loss": 1.2149,
+      "step": 15060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0055045714285714e-05,
+      "loss": 1.0775,
+      "step": 15061
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0054777142857142e-05,
+      "loss": 1.0955,
+      "step": 15062
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0054508571428572e-05,
+      "loss": 0.8989,
+      "step": 15063
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005424e-05,
+      "loss": 1.0006,
+      "step": 15064
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0053971428571429e-05,
+      "loss": 1.6033,
+      "step": 15065
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0053702857142859e-05,
+      "loss": 1.1063,
+      "step": 15066
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0053434285714285e-05,
+      "loss": 1.0343,
+      "step": 15067
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0053165714285715e-05,
+      "loss": 0.8231,
+      "step": 15068
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0052897142857143e-05,
+      "loss": 1.5051,
+      "step": 15069
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0052628571428573e-05,
+      "loss": 1.3497,
+      "step": 15070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005236e-05,
+      "loss": 1.1355,
+      "step": 15071
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005209142857143e-05,
+      "loss": 1.1312,
+      "step": 15072
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0051822857142857e-05,
+      "loss": 1.1934,
+      "step": 15073
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0051554285714286e-05,
+      "loss": 1.1187,
+      "step": 15074
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0051285714285714e-05,
+      "loss": 1.5961,
+      "step": 15075
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0051017142857144e-05,
+      "loss": 1.181,
+      "step": 15076
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0050748571428572e-05,
+      "loss": 1.5476,
+      "step": 15077
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005048e-05,
+      "loss": 1.0652,
+      "step": 15078
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005021142857143e-05,
+      "loss": 1.2336,
+      "step": 15079
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0049942857142856e-05,
+      "loss": 1.3657,
+      "step": 15080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0049674285714286e-05,
+      "loss": 1.5094,
+      "step": 15081
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0049405714285714e-05,
+      "loss": 0.8324,
+      "step": 15082
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0049137142857144e-05,
+      "loss": 1.4471,
+      "step": 15083
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.004886857142857e-05,
+      "loss": 1.2815,
+      "step": 15084
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.00486e-05,
+      "loss": 1.3131,
+      "step": 15085
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0048331428571429e-05,
+      "loss": 1.0987,
+      "step": 15086
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0048062857142857e-05,
+      "loss": 1.4977,
+      "step": 15087
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0047794285714285e-05,
+      "loss": 1.2999,
+      "step": 15088
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0047525714285715e-05,
+      "loss": 1.033,
+      "step": 15089
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0047257142857143e-05,
+      "loss": 1.1814,
+      "step": 15090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0046988571428571e-05,
+      "loss": 0.852,
+      "step": 15091
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0046720000000001e-05,
+      "loss": 1.1111,
+      "step": 15092
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.004645142857143e-05,
+      "loss": 1.1167,
+      "step": 15093
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0046182857142858e-05,
+      "loss": 0.9728,
+      "step": 15094
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0045914285714286e-05,
+      "loss": 1.402,
+      "step": 15095
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0045645714285716e-05,
+      "loss": 1.474,
+      "step": 15096
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0045377142857142e-05,
+      "loss": 1.1416,
+      "step": 15097
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0045108571428572e-05,
+      "loss": 1.5865,
+      "step": 15098
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.004484e-05,
+      "loss": 1.4466,
+      "step": 15099
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0044571428571428e-05,
+      "loss": 1.3792,
+      "step": 15100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0044302857142857e-05,
+      "loss": 1.1494,
+      "step": 15101
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0044034285714286e-05,
+      "loss": 0.9713,
+      "step": 15102
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0043765714285715e-05,
+      "loss": 0.9329,
+      "step": 15103
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0043497142857143e-05,
+      "loss": 0.9491,
+      "step": 15104
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0043228571428573e-05,
+      "loss": 1.427,
+      "step": 15105
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.004296e-05,
+      "loss": 0.8925,
+      "step": 15106
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0042691428571429e-05,
+      "loss": 0.9221,
+      "step": 15107
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0042422857142857e-05,
+      "loss": 1.4842,
+      "step": 15108
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0042154285714287e-05,
+      "loss": 1.3145,
+      "step": 15109
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0041885714285714e-05,
+      "loss": 1.3329,
+      "step": 15110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0041617142857143e-05,
+      "loss": 0.8425,
+      "step": 15111
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0041348571428572e-05,
+      "loss": 1.3519,
+      "step": 15112
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.004108e-05,
+      "loss": 1.4249,
+      "step": 15113
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0040811428571428e-05,
+      "loss": 1.264,
+      "step": 15114
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0040542857142858e-05,
+      "loss": 1.5414,
+      "step": 15115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0040274285714288e-05,
+      "loss": 1.3683,
+      "step": 15116
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0040005714285714e-05,
+      "loss": 1.6551,
+      "step": 15117
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0039737142857144e-05,
+      "loss": 1.2552,
+      "step": 15118
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0039468571428572e-05,
+      "loss": 1.4268,
+      "step": 15119
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.00392e-05,
+      "loss": 1.1818,
+      "step": 15120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0038931428571429e-05,
+      "loss": 0.9657,
+      "step": 15121
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0038662857142858e-05,
+      "loss": 1.3863,
+      "step": 15122
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0038394285714285e-05,
+      "loss": 1.1804,
+      "step": 15123
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0038125714285715e-05,
+      "loss": 1.4986,
+      "step": 15124
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0037857142857143e-05,
+      "loss": 0.843,
+      "step": 15125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0037588571428571e-05,
+      "loss": 1.3484,
+      "step": 15126
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003732e-05,
+      "loss": 1.2182,
+      "step": 15127
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003705142857143e-05,
+      "loss": 0.9852,
+      "step": 15128
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0036782857142859e-05,
+      "loss": 1.0103,
+      "step": 15129
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0036514285714286e-05,
+      "loss": 1.3225,
+      "step": 15130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0036245714285715e-05,
+      "loss": 1.0984,
+      "step": 15131
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0035977142857144e-05,
+      "loss": 1.0556,
+      "step": 15132
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0035708571428572e-05,
+      "loss": 1.1979,
+      "step": 15133
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003544e-05,
+      "loss": 1.2706,
+      "step": 15134
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003517142857143e-05,
+      "loss": 0.9281,
+      "step": 15135
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0034902857142856e-05,
+      "loss": 1.4588,
+      "step": 15136
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0034634285714286e-05,
+      "loss": 1.3954,
+      "step": 15137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0034365714285714e-05,
+      "loss": 1.6511,
+      "step": 15138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0034097142857144e-05,
+      "loss": 1.0547,
+      "step": 15139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.003382857142857e-05,
+      "loss": 1.0777,
+      "step": 15140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.003356e-05,
+      "loss": 1.528,
+      "step": 15141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.003329142857143e-05,
+      "loss": 1.6192,
+      "step": 15142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0033022857142857e-05,
+      "loss": 1.0826,
+      "step": 15143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0032754285714287e-05,
+      "loss": 1.4168,
+      "step": 15144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0032485714285715e-05,
+      "loss": 1.3919,
+      "step": 15145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0032217142857143e-05,
+      "loss": 1.0911,
+      "step": 15146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0031948571428571e-05,
+      "loss": 1.3308,
+      "step": 15147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0031680000000001e-05,
+      "loss": 1.6296,
+      "step": 15148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0031411428571428e-05,
+      "loss": 0.7027,
+      "step": 15149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0031142857142858e-05,
+      "loss": 1.0656,
+      "step": 15150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0030874285714286e-05,
+      "loss": 0.9174,
+      "step": 15151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0030605714285716e-05,
+      "loss": 1.1509,
+      "step": 15152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0030337142857142e-05,
+      "loss": 1.2777,
+      "step": 15153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0030068571428572e-05,
+      "loss": 1.3004,
+      "step": 15154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.00298e-05,
+      "loss": 0.9289,
+      "step": 15155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0029531428571428e-05,
+      "loss": 0.9728,
+      "step": 15156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0029262857142858e-05,
+      "loss": 1.0067,
+      "step": 15157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0028994285714286e-05,
+      "loss": 1.4836,
+      "step": 15158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0028725714285714e-05,
+      "loss": 1.1726,
+      "step": 15159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0028457142857143e-05,
+      "loss": 1.4832,
+      "step": 15160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0028188571428573e-05,
+      "loss": 1.0962,
+      "step": 15161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.002792e-05,
+      "loss": 1.3053,
+      "step": 15162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0027651428571429e-05,
+      "loss": 1.2273,
+      "step": 15163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0027382857142857e-05,
+      "loss": 1.3105,
+      "step": 15164
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0027114285714287e-05,
+      "loss": 1.2422,
+      "step": 15165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0026845714285713e-05,
+      "loss": 1.7182,
+      "step": 15166
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0026577142857143e-05,
+      "loss": 1.2678,
+      "step": 15167
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0026308571428571e-05,
+      "loss": 1.0492,
+      "step": 15168
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.002604e-05,
+      "loss": 1.0104,
+      "step": 15169
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.002577142857143e-05,
+      "loss": 0.8497,
+      "step": 15170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0025502857142858e-05,
+      "loss": 1.1685,
+      "step": 15171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0025234285714286e-05,
+      "loss": 0.9755,
+      "step": 15172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0024965714285714e-05,
+      "loss": 0.5899,
+      "step": 15173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0024697142857144e-05,
+      "loss": 1.0159,
+      "step": 15174
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0024428571428572e-05,
+      "loss": 1.2663,
+      "step": 15175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.002416e-05,
+      "loss": 1.5042,
+      "step": 15176
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0023891428571428e-05,
+      "loss": 1.5225,
+      "step": 15177
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0023622857142858e-05,
+      "loss": 1.1491,
+      "step": 15178
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0023354285714285e-05,
+      "loss": 1.2123,
+      "step": 15179
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0023085714285715e-05,
+      "loss": 1.2316,
+      "step": 15180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0022817142857143e-05,
+      "loss": 1.1056,
+      "step": 15181
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0022548571428571e-05,
+      "loss": 0.9231,
+      "step": 15182
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0022280000000001e-05,
+      "loss": 1.3325,
+      "step": 15183
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0022011428571429e-05,
+      "loss": 1.3013,
+      "step": 15184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0021742857142859e-05,
+      "loss": 1.5717,
+      "step": 15185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0021474285714285e-05,
+      "loss": 0.9837,
+      "step": 15186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0021205714285715e-05,
+      "loss": 1.6452,
+      "step": 15187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0020937142857143e-05,
+      "loss": 1.3313,
+      "step": 15188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0020668571428572e-05,
+      "loss": 0.931,
+      "step": 15189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.00204e-05,
+      "loss": 1.3259,
+      "step": 15190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.002013142857143e-05,
+      "loss": 1.2285,
+      "step": 15191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0019862857142856e-05,
+      "loss": 1.0561,
+      "step": 15192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0019594285714286e-05,
+      "loss": 0.8687,
+      "step": 15193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0019325714285714e-05,
+      "loss": 1.734,
+      "step": 15194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0019057142857142e-05,
+      "loss": 0.9967,
+      "step": 15195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0018788571428572e-05,
+      "loss": 0.8696,
+      "step": 15196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.001852e-05,
+      "loss": 0.977,
+      "step": 15197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.001825142857143e-05,
+      "loss": 1.5117,
+      "step": 15198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0017982857142857e-05,
+      "loss": 1.3921,
+      "step": 15199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0017714285714287e-05,
+      "loss": 1.0544,
+      "step": 15200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0017445714285715e-05,
+      "loss": 1.48,
+      "step": 15201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0017177142857143e-05,
+      "loss": 1.1083,
+      "step": 15202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0016908571428571e-05,
+      "loss": 0.8529,
+      "step": 15203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0016640000000001e-05,
+      "loss": 1.1108,
+      "step": 15204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0016371428571428e-05,
+      "loss": 1.5235,
+      "step": 15205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0016102857142857e-05,
+      "loss": 1.1734,
+      "step": 15206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0015834285714286e-05,
+      "loss": 1.0668,
+      "step": 15207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0015565714285714e-05,
+      "loss": 0.9412,
+      "step": 15208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0015297142857144e-05,
+      "loss": 1.381,
+      "step": 15209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0015028571428572e-05,
+      "loss": 0.9541,
+      "step": 15210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0014760000000002e-05,
+      "loss": 1.4546,
+      "step": 15211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0014491428571428e-05,
+      "loss": 0.552,
+      "step": 15212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0014222857142858e-05,
+      "loss": 1.0556,
+      "step": 15213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0013954285714286e-05,
+      "loss": 0.9771,
+      "step": 15214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0013685714285714e-05,
+      "loss": 1.1373,
+      "step": 15215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0013417142857143e-05,
+      "loss": 1.0016,
+      "step": 15216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0013148571428572e-05,
+      "loss": 1.3946,
+      "step": 15217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0012879999999999e-05,
+      "loss": 0.6393,
+      "step": 15218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0012611428571429e-05,
+      "loss": 1.0308,
+      "step": 15219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0012342857142857e-05,
+      "loss": 0.8492,
+      "step": 15220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0012074285714287e-05,
+      "loss": 0.8979,
+      "step": 15221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0011805714285715e-05,
+      "loss": 1.4612,
+      "step": 15222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0011537142857143e-05,
+      "loss": 0.708,
+      "step": 15223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0011268571428573e-05,
+      "loss": 0.9414,
+      "step": 15224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0011e-05,
+      "loss": 1.0206,
+      "step": 15225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.001073142857143e-05,
+      "loss": 1.4135,
+      "step": 15226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0010462857142858e-05,
+      "loss": 1.4356,
+      "step": 15227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0010194285714286e-05,
+      "loss": 1.3632,
+      "step": 15228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0009925714285714e-05,
+      "loss": 1.1172,
+      "step": 15229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0009657142857144e-05,
+      "loss": 1.4776,
+      "step": 15230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.000938857142857e-05,
+      "loss": 1.0519,
+      "step": 15231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.000912e-05,
+      "loss": 1.1348,
+      "step": 15232
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0008851428571428e-05,
+      "loss": 1.1182,
+      "step": 15233
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0008582857142858e-05,
+      "loss": 1.298,
+      "step": 15234
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0008314285714286e-05,
+      "loss": 1.3822,
+      "step": 15235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0008045714285715e-05,
+      "loss": 1.4435,
+      "step": 15236
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0007777142857144e-05,
+      "loss": 1.0262,
+      "step": 15237
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0007508571428571e-05,
+      "loss": 1.2168,
+      "step": 15238
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.000724e-05,
+      "loss": 1.299,
+      "step": 15239
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0006971428571429e-05,
+      "loss": 1.1407,
+      "step": 15240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0006702857142857e-05,
+      "loss": 1.1155,
+      "step": 15241
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0006434285714285e-05,
+      "loss": 1.4372,
+      "step": 15242
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0006165714285715e-05,
+      "loss": 1.0012,
+      "step": 15243
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0005897142857143e-05,
+      "loss": 1.0743,
+      "step": 15244
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0005628571428572e-05,
+      "loss": 0.6554,
+      "step": 15245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.000536e-05,
+      "loss": 1.5132,
+      "step": 15246
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.000509142857143e-05,
+      "loss": 1.2465,
+      "step": 15247
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0004822857142858e-05,
+      "loss": 1.2648,
+      "step": 15248
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0004554285714286e-05,
+      "loss": 0.8824,
+      "step": 15249
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0004285714285716e-05,
+      "loss": 1.0776,
+      "step": 15250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0004017142857142e-05,
+      "loss": 0.6032,
+      "step": 15251
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0003748571428572e-05,
+      "loss": 1.3738,
+      "step": 15252
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.000348e-05,
+      "loss": 1.5421,
+      "step": 15253
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0003211428571428e-05,
+      "loss": 1.199,
+      "step": 15254
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0002942857142857e-05,
+      "loss": 1.1062,
+      "step": 15255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0002674285714287e-05,
+      "loss": 1.1409,
+      "step": 15256
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0002405714285715e-05,
+      "loss": 1.2437,
+      "step": 15257
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0002137142857143e-05,
+      "loss": 1.285,
+      "step": 15258
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0001868571428571e-05,
+      "loss": 1.1277,
+      "step": 15259
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0001600000000001e-05,
+      "loss": 1.7535,
+      "step": 15260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0001331428571429e-05,
+      "loss": 1.7419,
+      "step": 15261
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0001062857142857e-05,
+      "loss": 0.8204,
+      "step": 15262
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0000794285714287e-05,
+      "loss": 1.2293,
+      "step": 15263
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0000525714285714e-05,
+      "loss": 0.9125,
+      "step": 15264
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0000257142857144e-05,
+      "loss": 1.0043,
+      "step": 15265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.999988571428572e-06,
+      "loss": 1.0139,
+      "step": 15266
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.999720000000002e-06,
+      "loss": 1.2398,
+      "step": 15267
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.999451428571428e-06,
+      "loss": 1.2056,
+      "step": 15268
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.999182857142858e-06,
+      "loss": 1.1161,
+      "step": 15269
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.998914285714286e-06,
+      "loss": 1.5067,
+      "step": 15270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.998645714285714e-06,
+      "loss": 1.5617,
+      "step": 15271
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.998377142857142e-06,
+      "loss": 1.2433,
+      "step": 15272
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.998108571428572e-06,
+      "loss": 1.2814,
+      "step": 15273
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.99784e-06,
+      "loss": 1.3643,
+      "step": 15274
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.997571428571429e-06,
+      "loss": 1.1756,
+      "step": 15275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.997302857142859e-06,
+      "loss": 0.9331,
+      "step": 15276
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.997034285714285e-06,
+      "loss": 1.1348,
+      "step": 15277
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.996765714285715e-06,
+      "loss": 1.1204,
+      "step": 15278
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.996497142857143e-06,
+      "loss": 1.2639,
+      "step": 15279
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.996228571428573e-06,
+      "loss": 1.2459,
+      "step": 15280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.99596e-06,
+      "loss": 1.2403,
+      "step": 15281
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.99569142857143e-06,
+      "loss": 1.5653,
+      "step": 15282
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.995422857142857e-06,
+      "loss": 1.3647,
+      "step": 15283
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.995154285714286e-06,
+      "loss": 1.2283,
+      "step": 15284
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.994885714285714e-06,
+      "loss": 1.0626,
+      "step": 15285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.994617142857144e-06,
+      "loss": 1.2429,
+      "step": 15286
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.994348571428572e-06,
+      "loss": 1.1795,
+      "step": 15287
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.99408e-06,
+      "loss": 1.1238,
+      "step": 15288
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.99381142857143e-06,
+      "loss": 1.2912,
+      "step": 15289
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.993542857142858e-06,
+      "loss": 0.8708,
+      "step": 15290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.993274285714286e-06,
+      "loss": 1.0636,
+      "step": 15291
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.993005714285714e-06,
+      "loss": 1.5256,
+      "step": 15292
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.992737142857144e-06,
+      "loss": 1.0078,
+      "step": 15293
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.99246857142857e-06,
+      "loss": 1.2837,
+      "step": 15294
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.9922e-06,
+      "loss": 1.4816,
+      "step": 15295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.991931428571429e-06,
+      "loss": 1.3896,
+      "step": 15296
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.991662857142857e-06,
+      "loss": 1.3125,
+      "step": 15297
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.991394285714285e-06,
+      "loss": 1.0946,
+      "step": 15298
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.991125714285715e-06,
+      "loss": 1.0839,
+      "step": 15299
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.990857142857143e-06,
+      "loss": 0.9593,
+      "step": 15300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.990588571428571e-06,
+      "loss": 1.2,
+      "step": 15301
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.990320000000001e-06,
+      "loss": 1.2511,
+      "step": 15302
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.99005142857143e-06,
+      "loss": 1.4152,
+      "step": 15303
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.989782857142858e-06,
+      "loss": 1.2931,
+      "step": 15304
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.989514285714286e-06,
+      "loss": 0.9903,
+      "step": 15305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.989245714285716e-06,
+      "loss": 1.4286,
+      "step": 15306
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.988977142857142e-06,
+      "loss": 1.3517,
+      "step": 15307
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.988708571428572e-06,
+      "loss": 1.0192,
+      "step": 15308
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.98844e-06,
+      "loss": 1.1838,
+      "step": 15309
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.988171428571428e-06,
+      "loss": 0.9972,
+      "step": 15310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.987902857142857e-06,
+      "loss": 0.8881,
+      "step": 15311
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.987634285714286e-06,
+      "loss": 1.0616,
+      "step": 15312
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.987365714285715e-06,
+      "loss": 1.2132,
+      "step": 15313
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.987097142857143e-06,
+      "loss": 1.228,
+      "step": 15314
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.986828571428573e-06,
+      "loss": 1.2649,
+      "step": 15315
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.98656e-06,
+      "loss": 1.2518,
+      "step": 15316
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.986291428571429e-06,
+      "loss": 1.2199,
+      "step": 15317
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.986022857142857e-06,
+      "loss": 1.6296,
+      "step": 15318
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.985754285714287e-06,
+      "loss": 0.815,
+      "step": 15319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.985485714285714e-06,
+      "loss": 1.2123,
+      "step": 15320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.985217142857143e-06,
+      "loss": 1.1446,
+      "step": 15321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.984948571428572e-06,
+      "loss": 1.172,
+      "step": 15322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.98468e-06,
+      "loss": 1.2157,
+      "step": 15323
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.984411428571428e-06,
+      "loss": 0.959,
+      "step": 15324
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.984142857142858e-06,
+      "loss": 1.778,
+      "step": 15325
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.983874285714286e-06,
+      "loss": 1.4623,
+      "step": 15326
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.983605714285714e-06,
+      "loss": 1.2092,
+      "step": 15327
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.983337142857144e-06,
+      "loss": 1.4035,
+      "step": 15328
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.983068571428572e-06,
+      "loss": 1.108,
+      "step": 15329
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.9828e-06,
+      "loss": 1.1659,
+      "step": 15330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.982531428571429e-06,
+      "loss": 1.4001,
+      "step": 15331
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.982262857142858e-06,
+      "loss": 1.387,
+      "step": 15332
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.981994285714285e-06,
+      "loss": 1.3089,
+      "step": 15333
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.981725714285715e-06,
+      "loss": 1.2916,
+      "step": 15334
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.981457142857143e-06,
+      "loss": 1.2978,
+      "step": 15335
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.981188571428573e-06,
+      "loss": 1.0428,
+      "step": 15336
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.98092e-06,
+      "loss": 1.4122,
+      "step": 15337
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.98065142857143e-06,
+      "loss": 1.1522,
+      "step": 15338
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.980382857142857e-06,
+      "loss": 1.3354,
+      "step": 15339
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.980114285714286e-06,
+      "loss": 1.1738,
+      "step": 15340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.979845714285715e-06,
+      "loss": 1.0466,
+      "step": 15341
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.979577142857144e-06,
+      "loss": 0.7429,
+      "step": 15342
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.979308571428572e-06,
+      "loss": 1.2257,
+      "step": 15343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.97904e-06,
+      "loss": 1.1791,
+      "step": 15344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.97877142857143e-06,
+      "loss": 1.2937,
+      "step": 15345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.978502857142856e-06,
+      "loss": 1.1725,
+      "step": 15346
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.978234285714286e-06,
+      "loss": 0.7805,
+      "step": 15347
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.977965714285714e-06,
+      "loss": 1.1755,
+      "step": 15348
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.977697142857144e-06,
+      "loss": 1.2066,
+      "step": 15349
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.97742857142857e-06,
+      "loss": 1.0605,
+      "step": 15350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.97716e-06,
+      "loss": 1.0843,
+      "step": 15351
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.976891428571429e-06,
+      "loss": 0.9559,
+      "step": 15352
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.976622857142857e-06,
+      "loss": 1.0065,
+      "step": 15353
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.976354285714287e-06,
+      "loss": 1.3686,
+      "step": 15354
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.976085714285715e-06,
+      "loss": 0.8079,
+      "step": 15355
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.975817142857143e-06,
+      "loss": 1.0644,
+      "step": 15356
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.975548571428571e-06,
+      "loss": 1.3128,
+      "step": 15357
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.975280000000001e-06,
+      "loss": 0.7073,
+      "step": 15358
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.97501142857143e-06,
+      "loss": 1.1368,
+      "step": 15359
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.974742857142858e-06,
+      "loss": 0.9731,
+      "step": 15360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.974474285714286e-06,
+      "loss": 1.4763,
+      "step": 15361
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.974205714285716e-06,
+      "loss": 0.8427,
+      "step": 15362
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.973937142857142e-06,
+      "loss": 1.0537,
+      "step": 15363
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.973668571428572e-06,
+      "loss": 1.2798,
+      "step": 15364
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.9734e-06,
+      "loss": 1.0507,
+      "step": 15365
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.973131428571428e-06,
+      "loss": 1.6508,
+      "step": 15366
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.972862857142858e-06,
+      "loss": 1.0211,
+      "step": 15367
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.972594285714286e-06,
+      "loss": 1.3844,
+      "step": 15368
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.972325714285714e-06,
+      "loss": 1.5088,
+      "step": 15369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.972057142857143e-06,
+      "loss": 1.7487,
+      "step": 15370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.971788571428573e-06,
+      "loss": 1.4032,
+      "step": 15371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.97152e-06,
+      "loss": 1.0018,
+      "step": 15372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.971251428571429e-06,
+      "loss": 1.3272,
+      "step": 15373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.970982857142857e-06,
+      "loss": 1.2306,
+      "step": 15374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.970714285714287e-06,
+      "loss": 1.25,
+      "step": 15375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.970445714285713e-06,
+      "loss": 1.3031,
+      "step": 15376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.970177142857143e-06,
+      "loss": 1.3874,
+      "step": 15377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.969908571428571e-06,
+      "loss": 1.0366,
+      "step": 15378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.96964e-06,
+      "loss": 1.0527,
+      "step": 15379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.96937142857143e-06,
+      "loss": 0.9548,
+      "step": 15380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.969102857142858e-06,
+      "loss": 1.1803,
+      "step": 15381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.968834285714286e-06,
+      "loss": 0.9833,
+      "step": 15382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.968565714285714e-06,
+      "loss": 1.1946,
+      "step": 15383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.968297142857144e-06,
+      "loss": 1.386,
+      "step": 15384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.968028571428572e-06,
+      "loss": 1.337,
+      "step": 15385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.96776e-06,
+      "loss": 1.0941,
+      "step": 15386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.967491428571428e-06,
+      "loss": 1.2455,
+      "step": 15387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.967222857142858e-06,
+      "loss": 1.216,
+      "step": 15388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.966954285714285e-06,
+      "loss": 1.5513,
+      "step": 15389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.966685714285715e-06,
+      "loss": 1.3705,
+      "step": 15390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.966417142857143e-06,
+      "loss": 0.844,
+      "step": 15391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.966148571428571e-06,
+      "loss": 1.0162,
+      "step": 15392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.965880000000001e-06,
+      "loss": 1.3006,
+      "step": 15393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.965611428571429e-06,
+      "loss": 1.0443,
+      "step": 15394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.965342857142859e-06,
+      "loss": 1.1773,
+      "step": 15395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.965074285714285e-06,
+      "loss": 1.354,
+      "step": 15396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.964805714285715e-06,
+      "loss": 1.4222,
+      "step": 15397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.964537142857143e-06,
+      "loss": 1.3778,
+      "step": 15398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.964268571428572e-06,
+      "loss": 1.3173,
+      "step": 15399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.964e-06,
+      "loss": 1.3447,
+      "step": 15400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.96373142857143e-06,
+      "loss": 1.0978,
+      "step": 15401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.963462857142856e-06,
+      "loss": 1.1862,
+      "step": 15402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.963194285714286e-06,
+      "loss": 1.3533,
+      "step": 15403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.962925714285714e-06,
+      "loss": 1.0441,
+      "step": 15404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.962657142857142e-06,
+      "loss": 1.1263,
+      "step": 15405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.962388571428572e-06,
+      "loss": 0.9712,
+      "step": 15406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.96212e-06,
+      "loss": 0.9237,
+      "step": 15407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.96185142857143e-06,
+      "loss": 1.2641,
+      "step": 15408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.961582857142857e-06,
+      "loss": 1.0528,
+      "step": 15409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.961314285714287e-06,
+      "loss": 0.857,
+      "step": 15410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.961045714285715e-06,
+      "loss": 1.2623,
+      "step": 15411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.960777142857143e-06,
+      "loss": 0.9367,
+      "step": 15412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.960508571428571e-06,
+      "loss": 1.312,
+      "step": 15413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.960240000000001e-06,
+      "loss": 1.467,
+      "step": 15414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.959971428571428e-06,
+      "loss": 1.1697,
+      "step": 15415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.959702857142857e-06,
+      "loss": 1.1861,
+      "step": 15416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.959434285714286e-06,
+      "loss": 1.2563,
+      "step": 15417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.959165714285715e-06,
+      "loss": 1.4908,
+      "step": 15418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.958897142857144e-06,
+      "loss": 1.3163,
+      "step": 15419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.958628571428572e-06,
+      "loss": 1.0318,
+      "step": 15420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.958360000000002e-06,
+      "loss": 1.5612,
+      "step": 15421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.958091428571428e-06,
+      "loss": 1.0851,
+      "step": 15422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.957822857142858e-06,
+      "loss": 1.4743,
+      "step": 15423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.957554285714286e-06,
+      "loss": 1.241,
+      "step": 15424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.957285714285714e-06,
+      "loss": 1.3944,
+      "step": 15425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.957017142857143e-06,
+      "loss": 0.8993,
+      "step": 15426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.956748571428572e-06,
+      "loss": 0.7771,
+      "step": 15427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.956479999999999e-06,
+      "loss": 1.2346,
+      "step": 15428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.956211428571429e-06,
+      "loss": 0.9805,
+      "step": 15429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.955942857142857e-06,
+      "loss": 1.4948,
+      "step": 15430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.955674285714287e-06,
+      "loss": 1.2755,
+      "step": 15431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.955405714285715e-06,
+      "loss": 1.2328,
+      "step": 15432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.955137142857143e-06,
+      "loss": 1.6333,
+      "step": 15433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.954868571428573e-06,
+      "loss": 1.4305,
+      "step": 15434
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.9546e-06,
+      "loss": 1.3488,
+      "step": 15435
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.95433142857143e-06,
+      "loss": 1.0847,
+      "step": 15436
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.954062857142858e-06,
+      "loss": 1.0549,
+      "step": 15437
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.953794285714286e-06,
+      "loss": 1.3012,
+      "step": 15438
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.953525714285714e-06,
+      "loss": 1.0439,
+      "step": 15439
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.953257142857144e-06,
+      "loss": 1.0755,
+      "step": 15440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.952988571428572e-06,
+      "loss": 1.6075,
+      "step": 15441
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.95272e-06,
+      "loss": 1.4446,
+      "step": 15442
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.952451428571428e-06,
+      "loss": 1.2943,
+      "step": 15443
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.952182857142858e-06,
+      "loss": 1.155,
+      "step": 15444
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.951914285714286e-06,
+      "loss": 1.0005,
+      "step": 15445
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.951645714285715e-06,
+      "loss": 1.0702,
+      "step": 15446
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.951377142857144e-06,
+      "loss": 1.3968,
+      "step": 15447
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.951108571428571e-06,
+      "loss": 1.5896,
+      "step": 15448
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.95084e-06,
+      "loss": 1.3247,
+      "step": 15449
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.950571428571429e-06,
+      "loss": 1.2534,
+      "step": 15450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.950302857142857e-06,
+      "loss": 1.663,
+      "step": 15451
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.950034285714285e-06,
+      "loss": 0.9864,
+      "step": 15452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.949765714285715e-06,
+      "loss": 1.0329,
+      "step": 15453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.949497142857143e-06,
+      "loss": 1.127,
+      "step": 15454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.949228571428572e-06,
+      "loss": 1.3609,
+      "step": 15455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.94896e-06,
+      "loss": 1.0099,
+      "step": 15456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.94869142857143e-06,
+      "loss": 1.5687,
+      "step": 15457
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.948422857142858e-06,
+      "loss": 0.989,
+      "step": 15458
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.948154285714286e-06,
+      "loss": 1.3692,
+      "step": 15459
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.947885714285716e-06,
+      "loss": 1.1252,
+      "step": 15460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.947617142857142e-06,
+      "loss": 1.0825,
+      "step": 15461
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.947348571428572e-06,
+      "loss": 1.3548,
+      "step": 15462
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.94708e-06,
+      "loss": 1.3322,
+      "step": 15463
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.94681142857143e-06,
+      "loss": 1.6708,
+      "step": 15464
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.946542857142857e-06,
+      "loss": 1.3092,
+      "step": 15465
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.946274285714287e-06,
+      "loss": 0.7029,
+      "step": 15466
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.946005714285715e-06,
+      "loss": 1.2058,
+      "step": 15467
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.945737142857143e-06,
+      "loss": 1.2999,
+      "step": 15468
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.945468571428571e-06,
+      "loss": 0.9928,
+      "step": 15469
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.945200000000001e-06,
+      "loss": 1.1061,
+      "step": 15470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.944931428571429e-06,
+      "loss": 1.3071,
+      "step": 15471
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.944662857142857e-06,
+      "loss": 1.1802,
+      "step": 15472
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.944394285714287e-06,
+      "loss": 1.1445,
+      "step": 15473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.944125714285714e-06,
+      "loss": 1.5228,
+      "step": 15474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.943857142857144e-06,
+      "loss": 1.2703,
+      "step": 15475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.943588571428572e-06,
+      "loss": 1.3804,
+      "step": 15476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.943320000000002e-06,
+      "loss": 1.1308,
+      "step": 15477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.943051428571428e-06,
+      "loss": 1.4183,
+      "step": 15478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.942782857142858e-06,
+      "loss": 1.0973,
+      "step": 15479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.942514285714286e-06,
+      "loss": 1.1083,
+      "step": 15480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.942245714285714e-06,
+      "loss": 1.0429,
+      "step": 15481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.941977142857142e-06,
+      "loss": 1.0583,
+      "step": 15482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.941708571428572e-06,
+      "loss": 1.1751,
+      "step": 15483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.94144e-06,
+      "loss": 1.3765,
+      "step": 15484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.941171428571429e-06,
+      "loss": 1.1748,
+      "step": 15485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.940902857142859e-06,
+      "loss": 1.3818,
+      "step": 15486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.940634285714287e-06,
+      "loss": 1.456,
+      "step": 15487
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.940365714285715e-06,
+      "loss": 1.0233,
+      "step": 15488
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.940097142857143e-06,
+      "loss": 1.4537,
+      "step": 15489
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.939828571428573e-06,
+      "loss": 1.1444,
+      "step": 15490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93956e-06,
+      "loss": 1.2866,
+      "step": 15491
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93929142857143e-06,
+      "loss": 1.3719,
+      "step": 15492
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.939022857142857e-06,
+      "loss": 1.3824,
+      "step": 15493
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.938754285714286e-06,
+      "loss": 1.1041,
+      "step": 15494
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.938485714285714e-06,
+      "loss": 0.8543,
+      "step": 15495
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.938217142857144e-06,
+      "loss": 1.3175,
+      "step": 15496
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.937948571428572e-06,
+      "loss": 1.2067,
+      "step": 15497
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93768e-06,
+      "loss": 1.1017,
+      "step": 15498
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93741142857143e-06,
+      "loss": 1.1669,
+      "step": 15499
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.937142857142858e-06,
+      "loss": 0.934,
+      "step": 15500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.936874285714286e-06,
+      "loss": 1.2698,
+      "step": 15501
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.936605714285714e-06,
+      "loss": 0.9615,
+      "step": 15502
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.936337142857144e-06,
+      "loss": 1.3115,
+      "step": 15503
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93606857142857e-06,
+      "loss": 1.3829,
+      "step": 15504
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.9358e-06,
+      "loss": 0.8947,
+      "step": 15505
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.935531428571429e-06,
+      "loss": 1.2754,
+      "step": 15506
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.935262857142857e-06,
+      "loss": 1.4286,
+      "step": 15507
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.934994285714285e-06,
+      "loss": 1.2844,
+      "step": 15508
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.934725714285715e-06,
+      "loss": 1.4273,
+      "step": 15509
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.934457142857143e-06,
+      "loss": 1.3167,
+      "step": 15510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.934188571428571e-06,
+      "loss": 1.4367,
+      "step": 15511
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.933920000000001e-06,
+      "loss": 1.0882,
+      "step": 15512
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93365142857143e-06,
+      "loss": 0.9467,
+      "step": 15513
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.933382857142858e-06,
+      "loss": 1.3667,
+      "step": 15514
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.933114285714286e-06,
+      "loss": 1.2648,
+      "step": 15515
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.932845714285716e-06,
+      "loss": 0.8985,
+      "step": 15516
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.932577142857142e-06,
+      "loss": 1.2851,
+      "step": 15517
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.932308571428572e-06,
+      "loss": 1.0943,
+      "step": 15518
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93204e-06,
+      "loss": 1.2612,
+      "step": 15519
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.931771428571428e-06,
+      "loss": 1.0176,
+      "step": 15520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.931502857142857e-06,
+      "loss": 1.1945,
+      "step": 15521
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.931234285714286e-06,
+      "loss": 1.2837,
+      "step": 15522
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.930965714285715e-06,
+      "loss": 1.2829,
+      "step": 15523
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.930697142857143e-06,
+      "loss": 1.3192,
+      "step": 15524
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.930428571428573e-06,
+      "loss": 1.4012,
+      "step": 15525
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.93016e-06,
+      "loss": 1.1775,
+      "step": 15526
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.929891428571429e-06,
+      "loss": 0.9932,
+      "step": 15527
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.929622857142857e-06,
+      "loss": 1.4705,
+      "step": 15528
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.929354285714287e-06,
+      "loss": 1.5275,
+      "step": 15529
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.929085714285714e-06,
+      "loss": 0.9756,
+      "step": 15530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.928817142857143e-06,
+      "loss": 1.0638,
+      "step": 15531
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.928548571428572e-06,
+      "loss": 1.3097,
+      "step": 15532
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.928280000000001e-06,
+      "loss": 1.1557,
+      "step": 15533
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.928011428571428e-06,
+      "loss": 1.4658,
+      "step": 15534
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.927742857142858e-06,
+      "loss": 0.9937,
+      "step": 15535
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.927474285714286e-06,
+      "loss": 1.1656,
+      "step": 15536
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.927205714285714e-06,
+      "loss": 1.6176,
+      "step": 15537
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.926937142857144e-06,
+      "loss": 1.3288,
+      "step": 15538
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.926668571428572e-06,
+      "loss": 1.2186,
+      "step": 15539
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.9264e-06,
+      "loss": 1.1587,
+      "step": 15540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.926131428571429e-06,
+      "loss": 1.3425,
+      "step": 15541
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.925862857142858e-06,
+      "loss": 1.2261,
+      "step": 15542
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.925594285714285e-06,
+      "loss": 1.1171,
+      "step": 15543
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.925325714285715e-06,
+      "loss": 1.3983,
+      "step": 15544
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.925057142857143e-06,
+      "loss": 1.0692,
+      "step": 15545
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.924788571428573e-06,
+      "loss": 1.049,
+      "step": 15546
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.92452e-06,
+      "loss": 1.0034,
+      "step": 15547
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.92425142857143e-06,
+      "loss": 1.2223,
+      "step": 15548
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.923982857142857e-06,
+      "loss": 1.3076,
+      "step": 15549
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.923714285714286e-06,
+      "loss": 1.2839,
+      "step": 15550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.923445714285715e-06,
+      "loss": 1.1669,
+      "step": 15551
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.923177142857144e-06,
+      "loss": 1.3404,
+      "step": 15552
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.922908571428572e-06,
+      "loss": 1.0045,
+      "step": 15553
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.92264e-06,
+      "loss": 0.9019,
+      "step": 15554
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.92237142857143e-06,
+      "loss": 1.5426,
+      "step": 15555
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.922102857142856e-06,
+      "loss": 1.3513,
+      "step": 15556
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.921834285714286e-06,
+      "loss": 1.1441,
+      "step": 15557
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.921565714285714e-06,
+      "loss": 1.2796,
+      "step": 15558
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.921297142857144e-06,
+      "loss": 1.3936,
+      "step": 15559
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.92102857142857e-06,
+      "loss": 1.2302,
+      "step": 15560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.92076e-06,
+      "loss": 1.2679,
+      "step": 15561
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.920491428571429e-06,
+      "loss": 0.7706,
+      "step": 15562
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.920222857142857e-06,
+      "loss": 1.1835,
+      "step": 15563
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.919954285714287e-06,
+      "loss": 1.3463,
+      "step": 15564
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.919685714285715e-06,
+      "loss": 0.6487,
+      "step": 15565
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.919417142857143e-06,
+      "loss": 1.5207,
+      "step": 15566
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.919148571428571e-06,
+      "loss": 0.8974,
+      "step": 15567
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.918880000000001e-06,
+      "loss": 0.8822,
+      "step": 15568
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.91861142857143e-06,
+      "loss": 1.673,
+      "step": 15569
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.918342857142858e-06,
+      "loss": 1.2316,
+      "step": 15570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.918074285714286e-06,
+      "loss": 1.3155,
+      "step": 15571
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.917805714285716e-06,
+      "loss": 0.7626,
+      "step": 15572
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.917537142857142e-06,
+      "loss": 1.2528,
+      "step": 15573
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.917268571428572e-06,
+      "loss": 1.2561,
+      "step": 15574
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.917e-06,
+      "loss": 1.029,
+      "step": 15575
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.916731428571428e-06,
+      "loss": 0.9554,
+      "step": 15576
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.916462857142858e-06,
+      "loss": 1.1847,
+      "step": 15577
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.916194285714286e-06,
+      "loss": 1.315,
+      "step": 15578
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.915925714285714e-06,
+      "loss": 1.6452,
+      "step": 15579
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.915657142857143e-06,
+      "loss": 1.5724,
+      "step": 15580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.915388571428573e-06,
+      "loss": 1.1777,
+      "step": 15581
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.91512e-06,
+      "loss": 0.9988,
+      "step": 15582
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.914851428571429e-06,
+      "loss": 1.2731,
+      "step": 15583
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.914582857142857e-06,
+      "loss": 1.4034,
+      "step": 15584
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.914314285714287e-06,
+      "loss": 0.896,
+      "step": 15585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.914045714285713e-06,
+      "loss": 1.3816,
+      "step": 15586
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.913777142857143e-06,
+      "loss": 1.1718,
+      "step": 15587
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.913508571428571e-06,
+      "loss": 1.2144,
+      "step": 15588
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.91324e-06,
+      "loss": 0.8408,
+      "step": 15589
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.91297142857143e-06,
+      "loss": 1.0911,
+      "step": 15590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.912702857142858e-06,
+      "loss": 1.4065,
+      "step": 15591
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.912434285714288e-06,
+      "loss": 1.1912,
+      "step": 15592
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.912165714285714e-06,
+      "loss": 0.8612,
+      "step": 15593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.911897142857144e-06,
+      "loss": 1.388,
+      "step": 15594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.911628571428572e-06,
+      "loss": 1.3225,
+      "step": 15595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.91136e-06,
+      "loss": 1.1657,
+      "step": 15596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.911091428571428e-06,
+      "loss": 1.0506,
+      "step": 15597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.910822857142858e-06,
+      "loss": 0.9127,
+      "step": 15598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.910554285714285e-06,
+      "loss": 1.5889,
+      "step": 15599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.910285714285715e-06,
+      "loss": 1.4767,
+      "step": 15600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.910017142857143e-06,
+      "loss": 1.3785,
+      "step": 15601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.909748571428571e-06,
+      "loss": 0.8417,
+      "step": 15602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.909480000000001e-06,
+      "loss": 1.0253,
+      "step": 15603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.909211428571429e-06,
+      "loss": 1.016,
+      "step": 15604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.908942857142859e-06,
+      "loss": 1.0904,
+      "step": 15605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.908674285714285e-06,
+      "loss": 1.1168,
+      "step": 15606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.908405714285715e-06,
+      "loss": 1.2051,
+      "step": 15607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.908137142857143e-06,
+      "loss": 1.3082,
+      "step": 15608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.907868571428572e-06,
+      "loss": 1.3239,
+      "step": 15609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.9076e-06,
+      "loss": 1.0404,
+      "step": 15610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.90733142857143e-06,
+      "loss": 1.0753,
+      "step": 15611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.907062857142856e-06,
+      "loss": 1.3004,
+      "step": 15612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.906794285714286e-06,
+      "loss": 1.1452,
+      "step": 15613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.906525714285714e-06,
+      "loss": 1.6098,
+      "step": 15614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.906257142857144e-06,
+      "loss": 1.1698,
+      "step": 15615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.905988571428572e-06,
+      "loss": 1.1978,
+      "step": 15616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.90572e-06,
+      "loss": 1.373,
+      "step": 15617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.90545142857143e-06,
+      "loss": 1.4952,
+      "step": 15618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.905182857142857e-06,
+      "loss": 1.4451,
+      "step": 15619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.904914285714287e-06,
+      "loss": 1.0453,
+      "step": 15620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.904645714285715e-06,
+      "loss": 1.4258,
+      "step": 15621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.904377142857143e-06,
+      "loss": 1.4361,
+      "step": 15622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.904108571428571e-06,
+      "loss": 1.4282,
+      "step": 15623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.903840000000001e-06,
+      "loss": 1.6713,
+      "step": 15624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.903571428571428e-06,
+      "loss": 1.5583,
+      "step": 15625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.903302857142857e-06,
+      "loss": 1.0886,
+      "step": 15626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.903034285714286e-06,
+      "loss": 1.4424,
+      "step": 15627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.902765714285715e-06,
+      "loss": 1.4825,
+      "step": 15628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.902497142857144e-06,
+      "loss": 1.1529,
+      "step": 15629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.902228571428572e-06,
+      "loss": 1.1391,
+      "step": 15630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.901960000000002e-06,
+      "loss": 1.3676,
+      "step": 15631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.901691428571428e-06,
+      "loss": 1.0899,
+      "step": 15632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.901422857142858e-06,
+      "loss": 1.1462,
+      "step": 15633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.901154285714286e-06,
+      "loss": 1.0752,
+      "step": 15634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.900885714285714e-06,
+      "loss": 1.0913,
+      "step": 15635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.900617142857143e-06,
+      "loss": 0.9608,
+      "step": 15636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.900348571428572e-06,
+      "loss": 1.2022,
+      "step": 15637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.90008e-06,
+      "loss": 0.7095,
+      "step": 15638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.899811428571429e-06,
+      "loss": 1.1087,
+      "step": 15639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.899542857142857e-06,
+      "loss": 1.4617,
+      "step": 15640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.899274285714287e-06,
+      "loss": 1.1032,
+      "step": 15641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.899005714285715e-06,
+      "loss": 1.4304,
+      "step": 15642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.898737142857143e-06,
+      "loss": 1.4317,
+      "step": 15643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.898468571428573e-06,
+      "loss": 1.0385,
+      "step": 15644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.8982e-06,
+      "loss": 1.0195,
+      "step": 15645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.89793142857143e-06,
+      "loss": 1.2603,
+      "step": 15646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.897662857142858e-06,
+      "loss": 0.9626,
+      "step": 15647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.897394285714286e-06,
+      "loss": 1.1137,
+      "step": 15648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.897125714285714e-06,
+      "loss": 1.5862,
+      "step": 15649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.896857142857144e-06,
+      "loss": 1.0918,
+      "step": 15650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.896588571428572e-06,
+      "loss": 1.0673,
+      "step": 15651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.89632e-06,
+      "loss": 1.4479,
+      "step": 15652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.896051428571428e-06,
+      "loss": 1.1542,
+      "step": 15653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.895782857142858e-06,
+      "loss": 1.2248,
+      "step": 15654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.895514285714286e-06,
+      "loss": 1.9498,
+      "step": 15655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.895245714285715e-06,
+      "loss": 1.0345,
+      "step": 15656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.894977142857144e-06,
+      "loss": 1.219,
+      "step": 15657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.894708571428571e-06,
+      "loss": 1.4498,
+      "step": 15658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.89444e-06,
+      "loss": 1.4511,
+      "step": 15659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.894171428571429e-06,
+      "loss": 1.0141,
+      "step": 15660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.893902857142859e-06,
+      "loss": 1.1485,
+      "step": 15661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.893634285714285e-06,
+      "loss": 1.2194,
+      "step": 15662
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.893365714285715e-06,
+      "loss": 1.1404,
+      "step": 15663
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.893097142857143e-06,
+      "loss": 1.1352,
+      "step": 15664
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.892828571428572e-06,
+      "loss": 1.6831,
+      "step": 15665
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.89256e-06,
+      "loss": 1.1808,
+      "step": 15666
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.89229142857143e-06,
+      "loss": 1.5193,
+      "step": 15667
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.892022857142858e-06,
+      "loss": 0.8883,
+      "step": 15668
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.891754285714286e-06,
+      "loss": 1.4265,
+      "step": 15669
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.891485714285716e-06,
+      "loss": 0.9012,
+      "step": 15670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.891217142857142e-06,
+      "loss": 1.1737,
+      "step": 15671
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.890948571428572e-06,
+      "loss": 1.3242,
+      "step": 15672
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.89068e-06,
+      "loss": 0.9892,
+      "step": 15673
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.89041142857143e-06,
+      "loss": 0.9513,
+      "step": 15674
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.890142857142857e-06,
+      "loss": 1.2412,
+      "step": 15675
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.889874285714287e-06,
+      "loss": 1.15,
+      "step": 15676
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.889605714285715e-06,
+      "loss": 1.415,
+      "step": 15677
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.889337142857143e-06,
+      "loss": 1.2693,
+      "step": 15678
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.889068571428571e-06,
+      "loss": 1.3388,
+      "step": 15679
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.888800000000001e-06,
+      "loss": 1.5783,
+      "step": 15680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.888531428571429e-06,
+      "loss": 1.1085,
+      "step": 15681
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.888262857142857e-06,
+      "loss": 1.5224,
+      "step": 15682
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.887994285714287e-06,
+      "loss": 1.0558,
+      "step": 15683
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.887725714285715e-06,
+      "loss": 0.8775,
+      "step": 15684
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.887457142857144e-06,
+      "loss": 1.3167,
+      "step": 15685
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.887188571428572e-06,
+      "loss": 0.8973,
+      "step": 15686
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.886920000000002e-06,
+      "loss": 0.9082,
+      "step": 15687
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.886651428571428e-06,
+      "loss": 0.9938,
+      "step": 15688
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.886382857142858e-06,
+      "loss": 1.0958,
+      "step": 15689
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.886114285714286e-06,
+      "loss": 1.1204,
+      "step": 15690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.885845714285714e-06,
+      "loss": 1.2393,
+      "step": 15691
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.885577142857142e-06,
+      "loss": 1.2548,
+      "step": 15692
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.885308571428572e-06,
+      "loss": 0.9984,
+      "step": 15693
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.88504e-06,
+      "loss": 0.9152,
+      "step": 15694
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.884771428571429e-06,
+      "loss": 1.3535,
+      "step": 15695
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.884502857142859e-06,
+      "loss": 1.2228,
+      "step": 15696
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.884234285714287e-06,
+      "loss": 1.2074,
+      "step": 15697
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.883965714285715e-06,
+      "loss": 1.4624,
+      "step": 15698
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.883697142857143e-06,
+      "loss": 1.0609,
+      "step": 15699
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.883428571428573e-06,
+      "loss": 1.1507,
+      "step": 15700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.88316e-06,
+      "loss": 1.2758,
+      "step": 15701
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.88289142857143e-06,
+      "loss": 1.1983,
+      "step": 15702
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.882622857142857e-06,
+      "loss": 1.2244,
+      "step": 15703
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.882354285714286e-06,
+      "loss": 1.2267,
+      "step": 15704
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.882085714285714e-06,
+      "loss": 1.6028,
+      "step": 15705
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.881817142857144e-06,
+      "loss": 0.9378,
+      "step": 15706
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.881548571428572e-06,
+      "loss": 1.288,
+      "step": 15707
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.88128e-06,
+      "loss": 1.1948,
+      "step": 15708
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.88101142857143e-06,
+      "loss": 0.9661,
+      "step": 15709
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.880742857142858e-06,
+      "loss": 0.919,
+      "step": 15710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.880474285714286e-06,
+      "loss": 0.5744,
+      "step": 15711
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.880205714285714e-06,
+      "loss": 0.957,
+      "step": 15712
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.879937142857144e-06,
+      "loss": 1.2511,
+      "step": 15713
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.87966857142857e-06,
+      "loss": 1.3633,
+      "step": 15714
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.8794e-06,
+      "loss": 1.2389,
+      "step": 15715
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.879131428571429e-06,
+      "loss": 1.1778,
+      "step": 15716
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.878862857142857e-06,
+      "loss": 0.9596,
+      "step": 15717
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.878594285714285e-06,
+      "loss": 1.5896,
+      "step": 15718
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.878325714285715e-06,
+      "loss": 1.2314,
+      "step": 15719
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.878057142857143e-06,
+      "loss": 1.2658,
+      "step": 15720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.877788571428571e-06,
+      "loss": 0.9379,
+      "step": 15721
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.877520000000001e-06,
+      "loss": 1.6054,
+      "step": 15722
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.87725142857143e-06,
+      "loss": 1.2797,
+      "step": 15723
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.876982857142858e-06,
+      "loss": 1.1361,
+      "step": 15724
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.876714285714286e-06,
+      "loss": 0.8693,
+      "step": 15725
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.876445714285716e-06,
+      "loss": 1.2621,
+      "step": 15726
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.876177142857142e-06,
+      "loss": 1.2239,
+      "step": 15727
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.875908571428572e-06,
+      "loss": 1.1935,
+      "step": 15728
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.87564e-06,
+      "loss": 1.1025,
+      "step": 15729
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.875371428571428e-06,
+      "loss": 1.1482,
+      "step": 15730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.875102857142857e-06,
+      "loss": 1.1236,
+      "step": 15731
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.874834285714286e-06,
+      "loss": 1.0059,
+      "step": 15732
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.874565714285715e-06,
+      "loss": 1.4472,
+      "step": 15733
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.874297142857143e-06,
+      "loss": 1.0783,
+      "step": 15734
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.874028571428573e-06,
+      "loss": 0.8227,
+      "step": 15735
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.87376e-06,
+      "loss": 1.2875,
+      "step": 15736
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.873491428571429e-06,
+      "loss": 1.4154,
+      "step": 15737
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.873222857142857e-06,
+      "loss": 0.8384,
+      "step": 15738
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.872954285714287e-06,
+      "loss": 1.271,
+      "step": 15739
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.872685714285714e-06,
+      "loss": 0.9053,
+      "step": 15740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.872417142857143e-06,
+      "loss": 1.3308,
+      "step": 15741
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.872148571428572e-06,
+      "loss": 1.2499,
+      "step": 15742
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.871880000000001e-06,
+      "loss": 1.2774,
+      "step": 15743
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.871611428571428e-06,
+      "loss": 1.1256,
+      "step": 15744
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.871342857142858e-06,
+      "loss": 1.076,
+      "step": 15745
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.871074285714286e-06,
+      "loss": 1.4749,
+      "step": 15746
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.870805714285714e-06,
+      "loss": 1.428,
+      "step": 15747
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.870537142857144e-06,
+      "loss": 1.3248,
+      "step": 15748
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.870268571428572e-06,
+      "loss": 0.8891,
+      "step": 15749
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.87e-06,
+      "loss": 0.8706,
+      "step": 15750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.869731428571429e-06,
+      "loss": 1.5771,
+      "step": 15751
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.869462857142858e-06,
+      "loss": 1.0968,
+      "step": 15752
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.869194285714285e-06,
+      "loss": 1.2727,
+      "step": 15753
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.868925714285715e-06,
+      "loss": 1.0724,
+      "step": 15754
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.868657142857143e-06,
+      "loss": 1.2304,
+      "step": 15755
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.868388571428573e-06,
+      "loss": 1.5041,
+      "step": 15756
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.86812e-06,
+      "loss": 1.1307,
+      "step": 15757
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.86785142857143e-06,
+      "loss": 1.1251,
+      "step": 15758
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.867582857142857e-06,
+      "loss": 1.3345,
+      "step": 15759
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.867314285714286e-06,
+      "loss": 1.125,
+      "step": 15760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.867045714285715e-06,
+      "loss": 1.3828,
+      "step": 15761
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.866777142857144e-06,
+      "loss": 1.2102,
+      "step": 15762
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.866508571428572e-06,
+      "loss": 1.5042,
+      "step": 15763
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.86624e-06,
+      "loss": 1.1941,
+      "step": 15764
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.86597142857143e-06,
+      "loss": 1.4188,
+      "step": 15765
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.865702857142858e-06,
+      "loss": 0.8538,
+      "step": 15766
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.865434285714286e-06,
+      "loss": 1.1298,
+      "step": 15767
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.865165714285714e-06,
+      "loss": 1.1383,
+      "step": 15768
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.864897142857144e-06,
+      "loss": 1.2368,
+      "step": 15769
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.86462857142857e-06,
+      "loss": 1.2483,
+      "step": 15770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.86436e-06,
+      "loss": 1.4554,
+      "step": 15771
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.864091428571429e-06,
+      "loss": 1.0868,
+      "step": 15772
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.863822857142857e-06,
+      "loss": 1.3292,
+      "step": 15773
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.863554285714287e-06,
+      "loss": 0.9367,
+      "step": 15774
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.863285714285715e-06,
+      "loss": 1.595,
+      "step": 15775
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.863017142857143e-06,
+      "loss": 1.3674,
+      "step": 15776
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.862748571428571e-06,
+      "loss": 0.8738,
+      "step": 15777
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.862480000000001e-06,
+      "loss": 1.5706,
+      "step": 15778
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.86221142857143e-06,
+      "loss": 1.1776,
+      "step": 15779
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.861942857142858e-06,
+      "loss": 0.8773,
+      "step": 15780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.861674285714286e-06,
+      "loss": 0.7576,
+      "step": 15781
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.861405714285716e-06,
+      "loss": 1.0433,
+      "step": 15782
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.861137142857142e-06,
+      "loss": 1.413,
+      "step": 15783
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.860868571428572e-06,
+      "loss": 1.0214,
+      "step": 15784
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.8606e-06,
+      "loss": 1.1118,
+      "step": 15785
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.860331428571428e-06,
+      "loss": 1.4484,
+      "step": 15786
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.860062857142858e-06,
+      "loss": 1.1935,
+      "step": 15787
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.859794285714286e-06,
+      "loss": 1.0156,
+      "step": 15788
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.859525714285716e-06,
+      "loss": 1.172,
+      "step": 15789
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.859257142857143e-06,
+      "loss": 1.2591,
+      "step": 15790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.858988571428573e-06,
+      "loss": 1.3044,
+      "step": 15791
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.85872e-06,
+      "loss": 1.3035,
+      "step": 15792
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.858451428571429e-06,
+      "loss": 1.3001,
+      "step": 15793
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.858182857142857e-06,
+      "loss": 1.6241,
+      "step": 15794
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.857914285714287e-06,
+      "loss": 1.3269,
+      "step": 15795
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.857645714285713e-06,
+      "loss": 1.5739,
+      "step": 15796
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.857377142857143e-06,
+      "loss": 1.6393,
+      "step": 15797
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.857108571428571e-06,
+      "loss": 1.2574,
+      "step": 15798
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.85684e-06,
+      "loss": 1.2826,
+      "step": 15799
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.85657142857143e-06,
+      "loss": 1.1497,
+      "step": 15800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.856302857142858e-06,
+      "loss": 1.5942,
+      "step": 15801
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.856034285714288e-06,
+      "loss": 1.513,
+      "step": 15802
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.855765714285714e-06,
+      "loss": 1.1439,
+      "step": 15803
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.855497142857144e-06,
+      "loss": 1.5602,
+      "step": 15804
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.855228571428572e-06,
+      "loss": 1.1875,
+      "step": 15805
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.85496e-06,
+      "loss": 1.2919,
+      "step": 15806
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.854691428571428e-06,
+      "loss": 0.9995,
+      "step": 15807
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.854422857142858e-06,
+      "loss": 1.0058,
+      "step": 15808
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.854154285714285e-06,
+      "loss": 1.1262,
+      "step": 15809
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.853885714285715e-06,
+      "loss": 1.1854,
+      "step": 15810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.853617142857143e-06,
+      "loss": 1.1502,
+      "step": 15811
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.853348571428573e-06,
+      "loss": 1.257,
+      "step": 15812
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.853080000000001e-06,
+      "loss": 1.0888,
+      "step": 15813
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.852811428571429e-06,
+      "loss": 1.5579,
+      "step": 15814
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.852542857142859e-06,
+      "loss": 1.1072,
+      "step": 15815
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.852274285714285e-06,
+      "loss": 0.8554,
+      "step": 15816
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.852005714285715e-06,
+      "loss": 1.2838,
+      "step": 15817
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.851737142857143e-06,
+      "loss": 0.82,
+      "step": 15818
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.851468571428572e-06,
+      "loss": 1.3916,
+      "step": 15819
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.8512e-06,
+      "loss": 1.2859,
+      "step": 15820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.85093142857143e-06,
+      "loss": 1.1176,
+      "step": 15821
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.850662857142856e-06,
+      "loss": 1.1749,
+      "step": 15822
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.850394285714286e-06,
+      "loss": 1.4577,
+      "step": 15823
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.850125714285714e-06,
+      "loss": 1.3938,
+      "step": 15824
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.849857142857144e-06,
+      "loss": 1.3202,
+      "step": 15825
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.849588571428572e-06,
+      "loss": 1.4248,
+      "step": 15826
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.84932e-06,
+      "loss": 1.2738,
+      "step": 15827
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.84905142857143e-06,
+      "loss": 0.846,
+      "step": 15828
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.848782857142857e-06,
+      "loss": 1.418,
+      "step": 15829
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.848514285714287e-06,
+      "loss": 1.3512,
+      "step": 15830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.848245714285715e-06,
+      "loss": 0.8333,
+      "step": 15831
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.847977142857143e-06,
+      "loss": 0.7763,
+      "step": 15832
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.847708571428571e-06,
+      "loss": 1.1133,
+      "step": 15833
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.847440000000001e-06,
+      "loss": 0.9046,
+      "step": 15834
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.84717142857143e-06,
+      "loss": 1.6596,
+      "step": 15835
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.846902857142857e-06,
+      "loss": 1.1503,
+      "step": 15836
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.846634285714286e-06,
+      "loss": 0.7295,
+      "step": 15837
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.846365714285715e-06,
+      "loss": 0.6819,
+      "step": 15838
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.846097142857142e-06,
+      "loss": 1.2557,
+      "step": 15839
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.845828571428572e-06,
+      "loss": 1.4288,
+      "step": 15840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.845560000000002e-06,
+      "loss": 1.4528,
+      "step": 15841
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.845291428571428e-06,
+      "loss": 1.3404,
+      "step": 15842
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.845022857142858e-06,
+      "loss": 0.8923,
+      "step": 15843
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.844754285714286e-06,
+      "loss": 1.2857,
+      "step": 15844
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.844485714285714e-06,
+      "loss": 1.1958,
+      "step": 15845
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.844217142857143e-06,
+      "loss": 1.1065,
+      "step": 15846
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.843948571428572e-06,
+      "loss": 1.2569,
+      "step": 15847
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.84368e-06,
+      "loss": 1.0371,
+      "step": 15848
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.843411428571429e-06,
+      "loss": 1.201,
+      "step": 15849
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.843142857142857e-06,
+      "loss": 1.0899,
+      "step": 15850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.842874285714287e-06,
+      "loss": 0.94,
+      "step": 15851
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.842605714285713e-06,
+      "loss": 0.9145,
+      "step": 15852
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.842337142857143e-06,
+      "loss": 0.9823,
+      "step": 15853
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.842068571428573e-06,
+      "loss": 1.3495,
+      "step": 15854
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.8418e-06,
+      "loss": 1.1373,
+      "step": 15855
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.84153142857143e-06,
+      "loss": 1.6944,
+      "step": 15856
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.841262857142858e-06,
+      "loss": 1.1046,
+      "step": 15857
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.840994285714287e-06,
+      "loss": 1.0205,
+      "step": 15858
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.840725714285714e-06,
+      "loss": 1.2696,
+      "step": 15859
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.840457142857144e-06,
+      "loss": 1.5425,
+      "step": 15860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.840188571428572e-06,
+      "loss": 1.2953,
+      "step": 15861
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.83992e-06,
+      "loss": 1.6168,
+      "step": 15862
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.839651428571428e-06,
+      "loss": 0.8507,
+      "step": 15863
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.839382857142858e-06,
+      "loss": 1.2036,
+      "step": 15864
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.839114285714285e-06,
+      "loss": 0.9344,
+      "step": 15865
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.838845714285715e-06,
+      "loss": 1.0413,
+      "step": 15866
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.838577142857144e-06,
+      "loss": 0.9967,
+      "step": 15867
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.838308571428571e-06,
+      "loss": 0.7883,
+      "step": 15868
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.83804e-06,
+      "loss": 0.99,
+      "step": 15869
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.837771428571429e-06,
+      "loss": 0.8414,
+      "step": 15870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.837502857142859e-06,
+      "loss": 1.5014,
+      "step": 15871
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.837234285714285e-06,
+      "loss": 1.2922,
+      "step": 15872
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.836965714285715e-06,
+      "loss": 1.0664,
+      "step": 15873
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.836697142857143e-06,
+      "loss": 0.893,
+      "step": 15874
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.836428571428572e-06,
+      "loss": 1.6125,
+      "step": 15875
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.83616e-06,
+      "loss": 1.1824,
+      "step": 15876
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.83589142857143e-06,
+      "loss": 0.9405,
+      "step": 15877
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.835622857142856e-06,
+      "loss": 1.1778,
+      "step": 15878
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.835354285714286e-06,
+      "loss": 1.1375,
+      "step": 15879
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.835085714285716e-06,
+      "loss": 1.2416,
+      "step": 15880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.834817142857142e-06,
+      "loss": 1.142,
+      "step": 15881
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.834548571428572e-06,
+      "loss": 1.1616,
+      "step": 15882
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.83428e-06,
+      "loss": 1.3107,
+      "step": 15883
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.83401142857143e-06,
+      "loss": 1.3837,
+      "step": 15884
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.833742857142857e-06,
+      "loss": 1.2549,
+      "step": 15885
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.833474285714287e-06,
+      "loss": 1.296,
+      "step": 15886
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.833205714285715e-06,
+      "loss": 1.5963,
+      "step": 15887
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.832937142857143e-06,
+      "loss": 1.2467,
+      "step": 15888
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.832668571428571e-06,
+      "loss": 1.8841,
+      "step": 15889
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.832400000000001e-06,
+      "loss": 1.4858,
+      "step": 15890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.832131428571427e-06,
+      "loss": 1.3302,
+      "step": 15891
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.831862857142857e-06,
+      "loss": 0.9754,
+      "step": 15892
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.831594285714287e-06,
+      "loss": 1.2148,
+      "step": 15893
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.831325714285715e-06,
+      "loss": 1.154,
+      "step": 15894
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.831057142857144e-06,
+      "loss": 1.3251,
+      "step": 15895
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.830788571428572e-06,
+      "loss": 1.2547,
+      "step": 15896
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.830520000000002e-06,
+      "loss": 0.9955,
+      "step": 15897
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.830251428571428e-06,
+      "loss": 1.2901,
+      "step": 15898
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.829982857142858e-06,
+      "loss": 1.2358,
+      "step": 15899
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.829714285714286e-06,
+      "loss": 1.7579,
+      "step": 15900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.829445714285714e-06,
+      "loss": 1.6544,
+      "step": 15901
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.829177142857142e-06,
+      "loss": 0.8399,
+      "step": 15902
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.828908571428572e-06,
+      "loss": 1.247,
+      "step": 15903
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.828639999999999e-06,
+      "loss": 0.8077,
+      "step": 15904
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.828371428571429e-06,
+      "loss": 1.0887,
+      "step": 15905
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.828102857142859e-06,
+      "loss": 1.4055,
+      "step": 15906
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.827834285714287e-06,
+      "loss": 0.9951,
+      "step": 15907
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.827565714285715e-06,
+      "loss": 0.8268,
+      "step": 15908
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.827297142857143e-06,
+      "loss": 1.1997,
+      "step": 15909
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.827028571428573e-06,
+      "loss": 0.9792,
+      "step": 15910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.82676e-06,
+      "loss": 1.0448,
+      "step": 15911
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.82649142857143e-06,
+      "loss": 0.9356,
+      "step": 15912
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.826222857142857e-06,
+      "loss": 1.265,
+      "step": 15913
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.825954285714286e-06,
+      "loss": 1.4924,
+      "step": 15914
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.825685714285714e-06,
+      "loss": 1.4403,
+      "step": 15915
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.825417142857144e-06,
+      "loss": 1.2239,
+      "step": 15916
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.825148571428572e-06,
+      "loss": 1.3487,
+      "step": 15917
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.82488e-06,
+      "loss": 1.4235,
+      "step": 15918
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.82461142857143e-06,
+      "loss": 0.8514,
+      "step": 15919
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.824342857142858e-06,
+      "loss": 0.8512,
+      "step": 15920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.824074285714286e-06,
+      "loss": 1.0443,
+      "step": 15921
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.823805714285714e-06,
+      "loss": 1.4079,
+      "step": 15922
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.823537142857144e-06,
+      "loss": 1.4196,
+      "step": 15923
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.82326857142857e-06,
+      "loss": 1.6256,
+      "step": 15924
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.823e-06,
+      "loss": 1.1246,
+      "step": 15925
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.822731428571429e-06,
+      "loss": 1.2333,
+      "step": 15926
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.822462857142857e-06,
+      "loss": 0.8336,
+      "step": 15927
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.822194285714285e-06,
+      "loss": 1.0392,
+      "step": 15928
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.821925714285715e-06,
+      "loss": 1.1404,
+      "step": 15929
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.821657142857143e-06,
+      "loss": 1.4424,
+      "step": 15930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.821388571428571e-06,
+      "loss": 1.3395,
+      "step": 15931
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.821120000000001e-06,
+      "loss": 1.3044,
+      "step": 15932
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.82085142857143e-06,
+      "loss": 1.0752,
+      "step": 15933
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.820582857142858e-06,
+      "loss": 1.7128,
+      "step": 15934
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.820314285714286e-06,
+      "loss": 1.4266,
+      "step": 15935
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.820045714285716e-06,
+      "loss": 1.1167,
+      "step": 15936
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.819777142857142e-06,
+      "loss": 1.1105,
+      "step": 15937
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.819508571428572e-06,
+      "loss": 1.3484,
+      "step": 15938
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.81924e-06,
+      "loss": 1.4594,
+      "step": 15939
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.81897142857143e-06,
+      "loss": 1.266,
+      "step": 15940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.818702857142857e-06,
+      "loss": 1.0132,
+      "step": 15941
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.818434285714286e-06,
+      "loss": 1.0028,
+      "step": 15942
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.818165714285715e-06,
+      "loss": 1.1475,
+      "step": 15943
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.817897142857143e-06,
+      "loss": 1.072,
+      "step": 15944
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.817628571428571e-06,
+      "loss": 1.4945,
+      "step": 15945
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.81736e-06,
+      "loss": 1.0539,
+      "step": 15946
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.817091428571429e-06,
+      "loss": 0.8585,
+      "step": 15947
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.816822857142857e-06,
+      "loss": 0.7827,
+      "step": 15948
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.816554285714287e-06,
+      "loss": 1.421,
+      "step": 15949
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.816285714285714e-06,
+      "loss": 1.2496,
+      "step": 15950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.816017142857143e-06,
+      "loss": 1.4563,
+      "step": 15951
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.815748571428572e-06,
+      "loss": 1.1407,
+      "step": 15952
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.815480000000001e-06,
+      "loss": 0.907,
+      "step": 15953
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.815211428571428e-06,
+      "loss": 1.124,
+      "step": 15954
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.814942857142858e-06,
+      "loss": 1.2513,
+      "step": 15955
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.814674285714286e-06,
+      "loss": 1.1512,
+      "step": 15956
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.814405714285714e-06,
+      "loss": 1.1767,
+      "step": 15957
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.814137142857142e-06,
+      "loss": 1.2592,
+      "step": 15958
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.813868571428572e-06,
+      "loss": 1.1638,
+      "step": 15959
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.8136e-06,
+      "loss": 1.1204,
+      "step": 15960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.813331428571429e-06,
+      "loss": 1.2812,
+      "step": 15961
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.813062857142858e-06,
+      "loss": 0.8722,
+      "step": 15962
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.812794285714287e-06,
+      "loss": 1.3502,
+      "step": 15963
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.812525714285715e-06,
+      "loss": 1.188,
+      "step": 15964
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.812257142857143e-06,
+      "loss": 1.2845,
+      "step": 15965
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.811988571428573e-06,
+      "loss": 1.4323,
+      "step": 15966
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.81172e-06,
+      "loss": 1.0442,
+      "step": 15967
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.81145142857143e-06,
+      "loss": 0.9395,
+      "step": 15968
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.811182857142857e-06,
+      "loss": 1.5008,
+      "step": 15969
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.810914285714286e-06,
+      "loss": 1.333,
+      "step": 15970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.810645714285714e-06,
+      "loss": 1.6268,
+      "step": 15971
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.810377142857144e-06,
+      "loss": 1.0735,
+      "step": 15972
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.810108571428572e-06,
+      "loss": 1.186,
+      "step": 15973
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.80984e-06,
+      "loss": 0.7008,
+      "step": 15974
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.80957142857143e-06,
+      "loss": 1.7367,
+      "step": 15975
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.809302857142858e-06,
+      "loss": 1.4433,
+      "step": 15976
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.809034285714286e-06,
+      "loss": 1.3783,
+      "step": 15977
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.808765714285714e-06,
+      "loss": 1.2796,
+      "step": 15978
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.808497142857144e-06,
+      "loss": 1.3989,
+      "step": 15979
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.80822857142857e-06,
+      "loss": 1.4073,
+      "step": 15980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.80796e-06,
+      "loss": 1.3333,
+      "step": 15981
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.807691428571429e-06,
+      "loss": 1.3225,
+      "step": 15982
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.807422857142857e-06,
+      "loss": 1.5056,
+      "step": 15983
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.807154285714285e-06,
+      "loss": 1.4482,
+      "step": 15984
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.806885714285715e-06,
+      "loss": 1.4706,
+      "step": 15985
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.806617142857145e-06,
+      "loss": 1.5443,
+      "step": 15986
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.806348571428571e-06,
+      "loss": 1.1646,
+      "step": 15987
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.806080000000001e-06,
+      "loss": 1.2713,
+      "step": 15988
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.80581142857143e-06,
+      "loss": 0.9925,
+      "step": 15989
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.805542857142858e-06,
+      "loss": 1.2718,
+      "step": 15990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.805274285714286e-06,
+      "loss": 1.2006,
+      "step": 15991
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.805005714285716e-06,
+      "loss": 1.1871,
+      "step": 15992
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.804737142857142e-06,
+      "loss": 1.3593,
+      "step": 15993
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.804468571428572e-06,
+      "loss": 1.2583,
+      "step": 15994
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.8042e-06,
+      "loss": 1.142,
+      "step": 15995
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.803931428571428e-06,
+      "loss": 1.0731,
+      "step": 15996
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.803662857142856e-06,
+      "loss": 0.8995,
+      "step": 15997
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.803394285714286e-06,
+      "loss": 1.1061,
+      "step": 15998
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.803125714285716e-06,
+      "loss": 1.0103,
+      "step": 15999
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.802857142857143e-06,
+      "loss": 1.3358,
+      "step": 16000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.802588571428573e-06,
+      "loss": 1.1745,
+      "step": 16001
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.80232e-06,
+      "loss": 1.3499,
+      "step": 16002
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.802051428571429e-06,
+      "loss": 1.3259,
+      "step": 16003
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.801782857142857e-06,
+      "loss": 1.2729,
+      "step": 16004
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.801514285714287e-06,
+      "loss": 0.8644,
+      "step": 16005
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.801245714285713e-06,
+      "loss": 1.0385,
+      "step": 16006
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.800977142857143e-06,
+      "loss": 1.432,
+      "step": 16007
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.800708571428571e-06,
+      "loss": 1.0964,
+      "step": 16008
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.800440000000001e-06,
+      "loss": 1.171,
+      "step": 16009
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.800171428571428e-06,
+      "loss": 1.453,
+      "step": 16010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.799902857142858e-06,
+      "loss": 1.0669,
+      "step": 16011
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.799634285714288e-06,
+      "loss": 0.7508,
+      "step": 16012
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.799365714285714e-06,
+      "loss": 1.2552,
+      "step": 16013
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.799097142857144e-06,
+      "loss": 1.4737,
+      "step": 16014
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.798828571428572e-06,
+      "loss": 1.3771,
+      "step": 16015
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.79856e-06,
+      "loss": 1.3893,
+      "step": 16016
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.798291428571428e-06,
+      "loss": 1.4686,
+      "step": 16017
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.798022857142858e-06,
+      "loss": 0.9646,
+      "step": 16018
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.797754285714285e-06,
+      "loss": 1.5155,
+      "step": 16019
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.797485714285715e-06,
+      "loss": 1.4887,
+      "step": 16020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.797217142857143e-06,
+      "loss": 1.0506,
+      "step": 16021
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.796948571428573e-06,
+      "loss": 1.1948,
+      "step": 16022
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.79668e-06,
+      "loss": 0.8643,
+      "step": 16023
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.796411428571429e-06,
+      "loss": 0.877,
+      "step": 16024
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.796142857142859e-06,
+      "loss": 1.1231,
+      "step": 16025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.795874285714285e-06,
+      "loss": 1.3351,
+      "step": 16026
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.795605714285715e-06,
+      "loss": 1.0376,
+      "step": 16027
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.795337142857143e-06,
+      "loss": 1.2179,
+      "step": 16028
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.795068571428572e-06,
+      "loss": 1.6361,
+      "step": 16029
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.7948e-06,
+      "loss": 1.003,
+      "step": 16030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.79453142857143e-06,
+      "loss": 1.4025,
+      "step": 16031
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.794262857142858e-06,
+      "loss": 1.1925,
+      "step": 16032
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.793994285714286e-06,
+      "loss": 1.1173,
+      "step": 16033
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.793725714285714e-06,
+      "loss": 1.0756,
+      "step": 16034
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.793457142857144e-06,
+      "loss": 0.9827,
+      "step": 16035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.79318857142857e-06,
+      "loss": 1.3879,
+      "step": 16036
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.79292e-06,
+      "loss": 1.1808,
+      "step": 16037
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.79265142857143e-06,
+      "loss": 1.431,
+      "step": 16038
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.792382857142857e-06,
+      "loss": 0.8422,
+      "step": 16039
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.792114285714287e-06,
+      "loss": 1.0709,
+      "step": 16040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.791845714285715e-06,
+      "loss": 1.6084,
+      "step": 16041
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.791577142857143e-06,
+      "loss": 0.8954,
+      "step": 16042
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.791308571428571e-06,
+      "loss": 1.2384,
+      "step": 16043
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.791040000000001e-06,
+      "loss": 1.059,
+      "step": 16044
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.79077142857143e-06,
+      "loss": 1.1706,
+      "step": 16045
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.790502857142857e-06,
+      "loss": 1.3364,
+      "step": 16046
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.790234285714286e-06,
+      "loss": 0.9647,
+      "step": 16047
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.789965714285715e-06,
+      "loss": 0.8526,
+      "step": 16048
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.789697142857142e-06,
+      "loss": 1.3862,
+      "step": 16049
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.789428571428572e-06,
+      "loss": 0.6655,
+      "step": 16050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.78916e-06,
+      "loss": 1.1896,
+      "step": 16051
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.788891428571428e-06,
+      "loss": 1.1928,
+      "step": 16052
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.788622857142858e-06,
+      "loss": 1.2735,
+      "step": 16053
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.788354285714286e-06,
+      "loss": 1.1733,
+      "step": 16054
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.788085714285714e-06,
+      "loss": 0.6727,
+      "step": 16055
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.787817142857143e-06,
+      "loss": 1.2642,
+      "step": 16056
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.787548571428572e-06,
+      "loss": 0.9027,
+      "step": 16057
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.78728e-06,
+      "loss": 0.7215,
+      "step": 16058
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.787011428571429e-06,
+      "loss": 1.0393,
+      "step": 16059
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.786742857142857e-06,
+      "loss": 1.04,
+      "step": 16060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.786474285714287e-06,
+      "loss": 1.1395,
+      "step": 16061
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.786205714285713e-06,
+      "loss": 0.896,
+      "step": 16062
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.785937142857143e-06,
+      "loss": 1.2118,
+      "step": 16063
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.785668571428571e-06,
+      "loss": 1.0701,
+      "step": 16064
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.7854e-06,
+      "loss": 0.9541,
+      "step": 16065
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.78513142857143e-06,
+      "loss": 1.1421,
+      "step": 16066
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.784862857142858e-06,
+      "loss": 1.1378,
+      "step": 16067
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.784594285714287e-06,
+      "loss": 1.3941,
+      "step": 16068
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.784325714285714e-06,
+      "loss": 1.1935,
+      "step": 16069
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.784057142857144e-06,
+      "loss": 1.2871,
+      "step": 16070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.783788571428572e-06,
+      "loss": 1.2091,
+      "step": 16071
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.78352e-06,
+      "loss": 1.4661,
+      "step": 16072
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.783251428571428e-06,
+      "loss": 0.9582,
+      "step": 16073
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.782982857142858e-06,
+      "loss": 1.0808,
+      "step": 16074
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.782714285714285e-06,
+      "loss": 1.0156,
+      "step": 16075
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.782445714285715e-06,
+      "loss": 0.997,
+      "step": 16076
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.782177142857143e-06,
+      "loss": 1.3983,
+      "step": 16077
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.781908571428571e-06,
+      "loss": 0.5948,
+      "step": 16078
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.78164e-06,
+      "loss": 0.6988,
+      "step": 16079
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.781371428571429e-06,
+      "loss": 1.1521,
+      "step": 16080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.781102857142859e-06,
+      "loss": 1.2682,
+      "step": 16081
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.780834285714285e-06,
+      "loss": 1.014,
+      "step": 16082
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.780565714285715e-06,
+      "loss": 1.2219,
+      "step": 16083
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.780297142857143e-06,
+      "loss": 0.9109,
+      "step": 16084
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.780028571428572e-06,
+      "loss": 1.5349,
+      "step": 16085
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.77976e-06,
+      "loss": 1.0414,
+      "step": 16086
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.77949142857143e-06,
+      "loss": 1.1545,
+      "step": 16087
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.779222857142856e-06,
+      "loss": 0.906,
+      "step": 16088
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.778954285714286e-06,
+      "loss": 1.5545,
+      "step": 16089
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.778685714285714e-06,
+      "loss": 1.0894,
+      "step": 16090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.778417142857144e-06,
+      "loss": 0.8735,
+      "step": 16091
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.778148571428572e-06,
+      "loss": 1.3976,
+      "step": 16092
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.77788e-06,
+      "loss": 1.4346,
+      "step": 16093
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.77761142857143e-06,
+      "loss": 1.0164,
+      "step": 16094
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.777342857142857e-06,
+      "loss": 0.9214,
+      "step": 16095
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.777074285714287e-06,
+      "loss": 0.7876,
+      "step": 16096
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.776805714285715e-06,
+      "loss": 0.9463,
+      "step": 16097
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.776537142857143e-06,
+      "loss": 0.9058,
+      "step": 16098
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.776268571428571e-06,
+      "loss": 0.8109,
+      "step": 16099
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.776000000000001e-06,
+      "loss": 1.4441,
+      "step": 16100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.775731428571427e-06,
+      "loss": 1.0051,
+      "step": 16101
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.775462857142857e-06,
+      "loss": 1.0183,
+      "step": 16102
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.775194285714285e-06,
+      "loss": 1.4327,
+      "step": 16103
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.774925714285715e-06,
+      "loss": 1.3887,
+      "step": 16104
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.774657142857144e-06,
+      "loss": 1.2058,
+      "step": 16105
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.774388571428572e-06,
+      "loss": 1.1155,
+      "step": 16106
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.774120000000002e-06,
+      "loss": 1.1609,
+      "step": 16107
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.773851428571428e-06,
+      "loss": 1.1356,
+      "step": 16108
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.773582857142858e-06,
+      "loss": 1.7492,
+      "step": 16109
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.773314285714286e-06,
+      "loss": 1.1776,
+      "step": 16110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.773045714285714e-06,
+      "loss": 0.9462,
+      "step": 16111
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.772777142857142e-06,
+      "loss": 1.3613,
+      "step": 16112
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.772508571428572e-06,
+      "loss": 1.3773,
+      "step": 16113
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.77224e-06,
+      "loss": 1.3659,
+      "step": 16114
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.771971428571429e-06,
+      "loss": 1.1215,
+      "step": 16115
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.771702857142857e-06,
+      "loss": 1.0053,
+      "step": 16116
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.771434285714287e-06,
+      "loss": 1.1206,
+      "step": 16117
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.771165714285715e-06,
+      "loss": 1.2804,
+      "step": 16118
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.770897142857143e-06,
+      "loss": 1.518,
+      "step": 16119
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.770628571428573e-06,
+      "loss": 1.0934,
+      "step": 16120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.77036e-06,
+      "loss": 1.2799,
+      "step": 16121
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.77009142857143e-06,
+      "loss": 1.27,
+      "step": 16122
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.769822857142857e-06,
+      "loss": 0.9388,
+      "step": 16123
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.769554285714286e-06,
+      "loss": 1.1506,
+      "step": 16124
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.769285714285714e-06,
+      "loss": 1.3396,
+      "step": 16125
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.769017142857144e-06,
+      "loss": 1.5244,
+      "step": 16126
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.768748571428572e-06,
+      "loss": 1.1795,
+      "step": 16127
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.76848e-06,
+      "loss": 0.9827,
+      "step": 16128
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.768211428571428e-06,
+      "loss": 1.5265,
+      "step": 16129
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.767942857142858e-06,
+      "loss": 1.3543,
+      "step": 16130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.767674285714286e-06,
+      "loss": 1.4623,
+      "step": 16131
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.767405714285714e-06,
+      "loss": 1.1142,
+      "step": 16132
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.767137142857144e-06,
+      "loss": 1.0534,
+      "step": 16133
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.76686857142857e-06,
+      "loss": 1.279,
+      "step": 16134
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.7666e-06,
+      "loss": 1.3127,
+      "step": 16135
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.766331428571429e-06,
+      "loss": 1.3276,
+      "step": 16136
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.766062857142859e-06,
+      "loss": 1.172,
+      "step": 16137
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.765794285714285e-06,
+      "loss": 0.9675,
+      "step": 16138
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.765525714285715e-06,
+      "loss": 1.6617,
+      "step": 16139
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.765257142857143e-06,
+      "loss": 0.9748,
+      "step": 16140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.764988571428571e-06,
+      "loss": 1.2987,
+      "step": 16141
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.76472e-06,
+      "loss": 1.2069,
+      "step": 16142
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.76445142857143e-06,
+      "loss": 1.492,
+      "step": 16143
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.764182857142858e-06,
+      "loss": 0.908,
+      "step": 16144
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.763914285714286e-06,
+      "loss": 1.2325,
+      "step": 16145
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.763645714285716e-06,
+      "loss": 0.7517,
+      "step": 16146
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.763377142857142e-06,
+      "loss": 0.6701,
+      "step": 16147
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.763108571428572e-06,
+      "loss": 1.1751,
+      "step": 16148
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.76284e-06,
+      "loss": 1.0584,
+      "step": 16149
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.76257142857143e-06,
+      "loss": 1.1143,
+      "step": 16150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.762302857142857e-06,
+      "loss": 1.1435,
+      "step": 16151
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.762034285714286e-06,
+      "loss": 1.4414,
+      "step": 16152
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.761765714285715e-06,
+      "loss": 0.8987,
+      "step": 16153
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.761497142857143e-06,
+      "loss": 1.2052,
+      "step": 16154
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.761228571428571e-06,
+      "loss": 0.7773,
+      "step": 16155
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.76096e-06,
+      "loss": 1.1394,
+      "step": 16156
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.760691428571429e-06,
+      "loss": 1.1281,
+      "step": 16157
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.760422857142857e-06,
+      "loss": 1.0086,
+      "step": 16158
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.760154285714287e-06,
+      "loss": 1.2388,
+      "step": 16159
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.759885714285715e-06,
+      "loss": 1.2854,
+      "step": 16160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.759617142857143e-06,
+      "loss": 1.0208,
+      "step": 16161
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.759348571428572e-06,
+      "loss": 1.0377,
+      "step": 16162
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.759080000000001e-06,
+      "loss": 1.405,
+      "step": 16163
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.758811428571428e-06,
+      "loss": 1.297,
+      "step": 16164
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.758542857142858e-06,
+      "loss": 1.1796,
+      "step": 16165
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.758274285714286e-06,
+      "loss": 0.9683,
+      "step": 16166
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.758005714285714e-06,
+      "loss": 0.9536,
+      "step": 16167
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.757737142857142e-06,
+      "loss": 1.4139,
+      "step": 16168
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.757468571428572e-06,
+      "loss": 0.8838,
+      "step": 16169
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.7572e-06,
+      "loss": 1.1211,
+      "step": 16170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.756931428571429e-06,
+      "loss": 1.2053,
+      "step": 16171
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.756662857142858e-06,
+      "loss": 1.6652,
+      "step": 16172
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.756394285714287e-06,
+      "loss": 1.1796,
+      "step": 16173
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.756125714285715e-06,
+      "loss": 1.0871,
+      "step": 16174
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.755857142857143e-06,
+      "loss": 1.0403,
+      "step": 16175
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.755588571428573e-06,
+      "loss": 1.2237,
+      "step": 16176
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.75532e-06,
+      "loss": 1.3,
+      "step": 16177
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.75505142857143e-06,
+      "loss": 1.2904,
+      "step": 16178
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.754782857142857e-06,
+      "loss": 1.1776,
+      "step": 16179
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.754514285714286e-06,
+      "loss": 1.074,
+      "step": 16180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.754245714285714e-06,
+      "loss": 1.0281,
+      "step": 16181
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.753977142857144e-06,
+      "loss": 0.801,
+      "step": 16182
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.753708571428572e-06,
+      "loss": 1.384,
+      "step": 16183
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.75344e-06,
+      "loss": 1.2292,
+      "step": 16184
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.75317142857143e-06,
+      "loss": 1.3616,
+      "step": 16185
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.752902857142858e-06,
+      "loss": 1.1677,
+      "step": 16186
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.752634285714286e-06,
+      "loss": 1.0127,
+      "step": 16187
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.752365714285714e-06,
+      "loss": 0.9835,
+      "step": 16188
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.752097142857144e-06,
+      "loss": 1.4954,
+      "step": 16189
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.75182857142857e-06,
+      "loss": 1.3056,
+      "step": 16190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.75156e-06,
+      "loss": 1.2586,
+      "step": 16191
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.751291428571429e-06,
+      "loss": 1.5838,
+      "step": 16192
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.751022857142857e-06,
+      "loss": 0.9989,
+      "step": 16193
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.750754285714285e-06,
+      "loss": 0.9535,
+      "step": 16194
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.750485714285715e-06,
+      "loss": 1.2719,
+      "step": 16195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.750217142857143e-06,
+      "loss": 0.9099,
+      "step": 16196
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.749948571428571e-06,
+      "loss": 1.342,
+      "step": 16197
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.749680000000001e-06,
+      "loss": 1.0665,
+      "step": 16198
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.74941142857143e-06,
+      "loss": 1.5784,
+      "step": 16199
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.749142857142858e-06,
+      "loss": 0.9424,
+      "step": 16200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.748874285714286e-06,
+      "loss": 1.2474,
+      "step": 16201
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.748605714285716e-06,
+      "loss": 1.1952,
+      "step": 16202
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.748337142857142e-06,
+      "loss": 1.4942,
+      "step": 16203
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.748068571428572e-06,
+      "loss": 1.1475,
+      "step": 16204
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.7478e-06,
+      "loss": 1.3359,
+      "step": 16205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.747531428571428e-06,
+      "loss": 1.5629,
+      "step": 16206
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.747262857142856e-06,
+      "loss": 1.302,
+      "step": 16207
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.746994285714286e-06,
+      "loss": 1.0712,
+      "step": 16208
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.746725714285714e-06,
+      "loss": 0.7044,
+      "step": 16209
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.746457142857143e-06,
+      "loss": 1.1771,
+      "step": 16210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.746188571428573e-06,
+      "loss": 1.0922,
+      "step": 16211
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.74592e-06,
+      "loss": 0.714,
+      "step": 16212
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.745651428571429e-06,
+      "loss": 1.0258,
+      "step": 16213
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.745382857142857e-06,
+      "loss": 0.8355,
+      "step": 16214
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.745114285714287e-06,
+      "loss": 1.42,
+      "step": 16215
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.744845714285713e-06,
+      "loss": 1.471,
+      "step": 16216
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.744577142857143e-06,
+      "loss": 0.7587,
+      "step": 16217
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.744308571428571e-06,
+      "loss": 1.6207,
+      "step": 16218
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.744040000000001e-06,
+      "loss": 1.4247,
+      "step": 16219
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.743771428571428e-06,
+      "loss": 1.1344,
+      "step": 16220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.743502857142858e-06,
+      "loss": 1.1911,
+      "step": 16221
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.743234285714286e-06,
+      "loss": 1.2042,
+      "step": 16222
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.742965714285714e-06,
+      "loss": 1.0014,
+      "step": 16223
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.742697142857144e-06,
+      "loss": 1.1815,
+      "step": 16224
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.742428571428572e-06,
+      "loss": 1.3957,
+      "step": 16225
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.74216e-06,
+      "loss": 1.3691,
+      "step": 16226
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.741891428571428e-06,
+      "loss": 1.0523,
+      "step": 16227
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.741622857142858e-06,
+      "loss": 1.1454,
+      "step": 16228
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.741354285714285e-06,
+      "loss": 1.5875,
+      "step": 16229
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.741085714285715e-06,
+      "loss": 1.2968,
+      "step": 16230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.740817142857143e-06,
+      "loss": 1.3082,
+      "step": 16231
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.740548571428573e-06,
+      "loss": 1.6271,
+      "step": 16232
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.74028e-06,
+      "loss": 1.1863,
+      "step": 16233
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.740011428571429e-06,
+      "loss": 1.0276,
+      "step": 16234
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.739742857142857e-06,
+      "loss": 1.6264,
+      "step": 16235
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.739474285714285e-06,
+      "loss": 1.2818,
+      "step": 16236
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.739205714285715e-06,
+      "loss": 0.9992,
+      "step": 16237
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.738937142857143e-06,
+      "loss": 0.804,
+      "step": 16238
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.738668571428572e-06,
+      "loss": 1.2382,
+      "step": 16239
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.7384e-06,
+      "loss": 1.0639,
+      "step": 16240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.73813142857143e-06,
+      "loss": 0.9194,
+      "step": 16241
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.737862857142858e-06,
+      "loss": 1.3359,
+      "step": 16242
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.737594285714286e-06,
+      "loss": 1.148,
+      "step": 16243
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.737325714285714e-06,
+      "loss": 1.7258,
+      "step": 16244
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.737057142857144e-06,
+      "loss": 1.2033,
+      "step": 16245
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.73678857142857e-06,
+      "loss": 0.9186,
+      "step": 16246
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.73652e-06,
+      "loss": 0.9807,
+      "step": 16247
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.736251428571429e-06,
+      "loss": 1.2698,
+      "step": 16248
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.735982857142857e-06,
+      "loss": 1.4538,
+      "step": 16249
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.735714285714287e-06,
+      "loss": 1.1633,
+      "step": 16250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.735445714285715e-06,
+      "loss": 1.1555,
+      "step": 16251
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.735177142857143e-06,
+      "loss": 1.5759,
+      "step": 16252
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.734908571428571e-06,
+      "loss": 1.1653,
+      "step": 16253
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.734640000000001e-06,
+      "loss": 1.1894,
+      "step": 16254
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.73437142857143e-06,
+      "loss": 1.4638,
+      "step": 16255
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.734102857142857e-06,
+      "loss": 1.0998,
+      "step": 16256
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.733834285714286e-06,
+      "loss": 1.2147,
+      "step": 16257
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.733565714285715e-06,
+      "loss": 1.4802,
+      "step": 16258
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.733297142857142e-06,
+      "loss": 1.085,
+      "step": 16259
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.733028571428572e-06,
+      "loss": 0.8951,
+      "step": 16260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.73276e-06,
+      "loss": 0.9839,
+      "step": 16261
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.732491428571428e-06,
+      "loss": 1.1486,
+      "step": 16262
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.732222857142858e-06,
+      "loss": 1.1262,
+      "step": 16263
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.731954285714286e-06,
+      "loss": 1.1455,
+      "step": 16264
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.731685714285716e-06,
+      "loss": 1.3109,
+      "step": 16265
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.731417142857143e-06,
+      "loss": 1.4879,
+      "step": 16266
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.731148571428572e-06,
+      "loss": 1.0336,
+      "step": 16267
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.73088e-06,
+      "loss": 0.8245,
+      "step": 16268
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.730611428571429e-06,
+      "loss": 1.2316,
+      "step": 16269
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.730342857142857e-06,
+      "loss": 1.1524,
+      "step": 16270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.730074285714287e-06,
+      "loss": 1.3597,
+      "step": 16271
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.729805714285713e-06,
+      "loss": 0.8408,
+      "step": 16272
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.729537142857143e-06,
+      "loss": 1.2049,
+      "step": 16273
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.729268571428571e-06,
+      "loss": 1.213,
+      "step": 16274
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.729e-06,
+      "loss": 1.2811,
+      "step": 16275
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.72873142857143e-06,
+      "loss": 1.3539,
+      "step": 16276
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.728462857142858e-06,
+      "loss": 1.0274,
+      "step": 16277
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.728194285714287e-06,
+      "loss": 1.0728,
+      "step": 16278
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.727925714285714e-06,
+      "loss": 1.2345,
+      "step": 16279
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.727657142857144e-06,
+      "loss": 1.4475,
+      "step": 16280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.727388571428572e-06,
+      "loss": 1.461,
+      "step": 16281
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.72712e-06,
+      "loss": 1.1572,
+      "step": 16282
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.726851428571428e-06,
+      "loss": 1.0457,
+      "step": 16283
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.726582857142858e-06,
+      "loss": 1.1123,
+      "step": 16284
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.726314285714285e-06,
+      "loss": 1.1342,
+      "step": 16285
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.726045714285715e-06,
+      "loss": 1.0402,
+      "step": 16286
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.725777142857143e-06,
+      "loss": 1.5609,
+      "step": 16287
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.725508571428573e-06,
+      "loss": 1.491,
+      "step": 16288
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.72524e-06,
+      "loss": 1.2654,
+      "step": 16289
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.724971428571429e-06,
+      "loss": 1.2337,
+      "step": 16290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.724702857142859e-06,
+      "loss": 1.269,
+      "step": 16291
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.724434285714285e-06,
+      "loss": 0.9727,
+      "step": 16292
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.724165714285715e-06,
+      "loss": 1.3448,
+      "step": 16293
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.723897142857143e-06,
+      "loss": 1.2013,
+      "step": 16294
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.723628571428572e-06,
+      "loss": 1.2409,
+      "step": 16295
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.72336e-06,
+      "loss": 0.925,
+      "step": 16296
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.72309142857143e-06,
+      "loss": 1.1166,
+      "step": 16297
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.722822857142856e-06,
+      "loss": 1.3282,
+      "step": 16298
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.722554285714286e-06,
+      "loss": 1.6159,
+      "step": 16299
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.722285714285714e-06,
+      "loss": 1.2306,
+      "step": 16300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.722017142857144e-06,
+      "loss": 1.0194,
+      "step": 16301
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.721748571428572e-06,
+      "loss": 1.5033,
+      "step": 16302
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.72148e-06,
+      "loss": 1.159,
+      "step": 16303
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.72121142857143e-06,
+      "loss": 1.2162,
+      "step": 16304
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.720942857142857e-06,
+      "loss": 1.3436,
+      "step": 16305
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.720674285714287e-06,
+      "loss": 1.3675,
+      "step": 16306
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.720405714285715e-06,
+      "loss": 1.4311,
+      "step": 16307
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.720137142857143e-06,
+      "loss": 1.0007,
+      "step": 16308
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.719868571428571e-06,
+      "loss": 1.5602,
+      "step": 16309
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.719600000000001e-06,
+      "loss": 1.3999,
+      "step": 16310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.719331428571429e-06,
+      "loss": 1.4165,
+      "step": 16311
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.719062857142857e-06,
+      "loss": 1.3501,
+      "step": 16312
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.718794285714285e-06,
+      "loss": 1.1718,
+      "step": 16313
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.718525714285715e-06,
+      "loss": 1.2866,
+      "step": 16314
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.718257142857144e-06,
+      "loss": 0.946,
+      "step": 16315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.717988571428572e-06,
+      "loss": 1.6365,
+      "step": 16316
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.717720000000002e-06,
+      "loss": 1.242,
+      "step": 16317
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.717451428571428e-06,
+      "loss": 1.2691,
+      "step": 16318
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.717182857142858e-06,
+      "loss": 1.3543,
+      "step": 16319
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.716914285714286e-06,
+      "loss": 1.3082,
+      "step": 16320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.716645714285714e-06,
+      "loss": 1.2409,
+      "step": 16321
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.716377142857142e-06,
+      "loss": 1.2481,
+      "step": 16322
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.716108571428572e-06,
+      "loss": 1.3327,
+      "step": 16323
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.71584e-06,
+      "loss": 1.5265,
+      "step": 16324
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.715571428571429e-06,
+      "loss": 0.9563,
+      "step": 16325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.715302857142857e-06,
+      "loss": 1.5373,
+      "step": 16326
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.715034285714287e-06,
+      "loss": 1.7004,
+      "step": 16327
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.714765714285715e-06,
+      "loss": 0.9972,
+      "step": 16328
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.714497142857143e-06,
+      "loss": 1.4439,
+      "step": 16329
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.714228571428573e-06,
+      "loss": 0.7498,
+      "step": 16330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.71396e-06,
+      "loss": 0.9134,
+      "step": 16331
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.71369142857143e-06,
+      "loss": 1.4724,
+      "step": 16332
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.713422857142857e-06,
+      "loss": 0.962,
+      "step": 16333
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.713154285714287e-06,
+      "loss": 1.0198,
+      "step": 16334
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.712885714285714e-06,
+      "loss": 1.1389,
+      "step": 16335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.712617142857144e-06,
+      "loss": 0.8593,
+      "step": 16336
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.712348571428572e-06,
+      "loss": 1.4565,
+      "step": 16337
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.71208e-06,
+      "loss": 1.5168,
+      "step": 16338
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.711811428571428e-06,
+      "loss": 1.3258,
+      "step": 16339
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.711542857142858e-06,
+      "loss": 1.3723,
+      "step": 16340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.711274285714286e-06,
+      "loss": 1.2498,
+      "step": 16341
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.711005714285714e-06,
+      "loss": 1.4636,
+      "step": 16342
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.710737142857144e-06,
+      "loss": 1.833,
+      "step": 16343
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.71046857142857e-06,
+      "loss": 1.1211,
+      "step": 16344
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.7102e-06,
+      "loss": 1.3128,
+      "step": 16345
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.709931428571429e-06,
+      "loss": 0.9392,
+      "step": 16346
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.709662857142859e-06,
+      "loss": 1.5265,
+      "step": 16347
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.709394285714285e-06,
+      "loss": 0.9284,
+      "step": 16348
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.709125714285715e-06,
+      "loss": 1.1763,
+      "step": 16349
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.708857142857143e-06,
+      "loss": 1.0857,
+      "step": 16350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.708588571428571e-06,
+      "loss": 1.1337,
+      "step": 16351
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.70832e-06,
+      "loss": 1.5603,
+      "step": 16352
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.70805142857143e-06,
+      "loss": 1.1734,
+      "step": 16353
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.707782857142858e-06,
+      "loss": 1.0826,
+      "step": 16354
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.707514285714286e-06,
+      "loss": 1.1441,
+      "step": 16355
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.707245714285716e-06,
+      "loss": 0.8326,
+      "step": 16356
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.706977142857144e-06,
+      "loss": 1.5722,
+      "step": 16357
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.706708571428572e-06,
+      "loss": 1.4321,
+      "step": 16358
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.70644e-06,
+      "loss": 1.1804,
+      "step": 16359
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.70617142857143e-06,
+      "loss": 1.0576,
+      "step": 16360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.705902857142857e-06,
+      "loss": 1.2131,
+      "step": 16361
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.705634285714286e-06,
+      "loss": 0.9864,
+      "step": 16362
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.705365714285715e-06,
+      "loss": 1.018,
+      "step": 16363
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.705097142857143e-06,
+      "loss": 1.5136,
+      "step": 16364
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.704828571428571e-06,
+      "loss": 1.2323,
+      "step": 16365
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.70456e-06,
+      "loss": 1.2681,
+      "step": 16366
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.704291428571429e-06,
+      "loss": 1.253,
+      "step": 16367
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.704022857142857e-06,
+      "loss": 1.3259,
+      "step": 16368
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.703754285714287e-06,
+      "loss": 1.2991,
+      "step": 16369
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.703485714285715e-06,
+      "loss": 1.5921,
+      "step": 16370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.703217142857143e-06,
+      "loss": 1.4916,
+      "step": 16371
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.702948571428572e-06,
+      "loss": 1.0752,
+      "step": 16372
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.702680000000001e-06,
+      "loss": 1.0076,
+      "step": 16373
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.702411428571428e-06,
+      "loss": 0.8734,
+      "step": 16374
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.702142857142858e-06,
+      "loss": 1.3882,
+      "step": 16375
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.701874285714286e-06,
+      "loss": 1.5735,
+      "step": 16376
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.701605714285714e-06,
+      "loss": 1.0553,
+      "step": 16377
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.701337142857142e-06,
+      "loss": 1.1539,
+      "step": 16378
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.701068571428572e-06,
+      "loss": 1.4987,
+      "step": 16379
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.7008e-06,
+      "loss": 1.2125,
+      "step": 16380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.700531428571429e-06,
+      "loss": 0.8298,
+      "step": 16381
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.700262857142858e-06,
+      "loss": 0.8703,
+      "step": 16382
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.699994285714287e-06,
+      "loss": 1.3714,
+      "step": 16383
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.699725714285715e-06,
+      "loss": 1.4435,
+      "step": 16384
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.699457142857143e-06,
+      "loss": 1.2991,
+      "step": 16385
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.699188571428573e-06,
+      "loss": 1.2187,
+      "step": 16386
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69892e-06,
+      "loss": 1.0797,
+      "step": 16387
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69865142857143e-06,
+      "loss": 1.2005,
+      "step": 16388
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.698382857142857e-06,
+      "loss": 1.1466,
+      "step": 16389
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.698114285714286e-06,
+      "loss": 1.0589,
+      "step": 16390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.697845714285714e-06,
+      "loss": 1.6402,
+      "step": 16391
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.697577142857144e-06,
+      "loss": 1.2792,
+      "step": 16392
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.697308571428572e-06,
+      "loss": 1.0767,
+      "step": 16393
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69704e-06,
+      "loss": 1.2483,
+      "step": 16394
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69677142857143e-06,
+      "loss": 1.4549,
+      "step": 16395
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.696502857142858e-06,
+      "loss": 1.0903,
+      "step": 16396
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.696234285714286e-06,
+      "loss": 1.1587,
+      "step": 16397
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.695965714285714e-06,
+      "loss": 1.2412,
+      "step": 16398
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.695697142857144e-06,
+      "loss": 0.8616,
+      "step": 16399
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69542857142857e-06,
+      "loss": 1.1079,
+      "step": 16400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69516e-06,
+      "loss": 1.441,
+      "step": 16401
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.694891428571429e-06,
+      "loss": 1.1147,
+      "step": 16402
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.694622857142857e-06,
+      "loss": 0.9282,
+      "step": 16403
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.694354285714285e-06,
+      "loss": 1.2389,
+      "step": 16404
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.694085714285715e-06,
+      "loss": 0.8827,
+      "step": 16405
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.693817142857143e-06,
+      "loss": 1.2386,
+      "step": 16406
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.693548571428571e-06,
+      "loss": 1.1911,
+      "step": 16407
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.693280000000001e-06,
+      "loss": 1.2793,
+      "step": 16408
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69301142857143e-06,
+      "loss": 1.2372,
+      "step": 16409
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.692742857142858e-06,
+      "loss": 1.0524,
+      "step": 16410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.692474285714286e-06,
+      "loss": 1.0016,
+      "step": 16411
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.692205714285716e-06,
+      "loss": 1.2715,
+      "step": 16412
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.691937142857142e-06,
+      "loss": 1.2839,
+      "step": 16413
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.691668571428572e-06,
+      "loss": 1.247,
+      "step": 16414
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.6914e-06,
+      "loss": 1.4688,
+      "step": 16415
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69113142857143e-06,
+      "loss": 1.0849,
+      "step": 16416
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.690862857142856e-06,
+      "loss": 1.3748,
+      "step": 16417
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.690594285714286e-06,
+      "loss": 1.1383,
+      "step": 16418
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.690325714285714e-06,
+      "loss": 1.1776,
+      "step": 16419
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.690057142857143e-06,
+      "loss": 1.3741,
+      "step": 16420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.689788571428573e-06,
+      "loss": 0.8676,
+      "step": 16421
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.68952e-06,
+      "loss": 1.1104,
+      "step": 16422
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.689251428571429e-06,
+      "loss": 1.006,
+      "step": 16423
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.688982857142857e-06,
+      "loss": 1.5352,
+      "step": 16424
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.688714285714287e-06,
+      "loss": 1.2602,
+      "step": 16425
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.688445714285713e-06,
+      "loss": 1.5726,
+      "step": 16426
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.688177142857143e-06,
+      "loss": 1.1148,
+      "step": 16427
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.687908571428571e-06,
+      "loss": 0.8642,
+      "step": 16428
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.687640000000001e-06,
+      "loss": 1.1054,
+      "step": 16429
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.687371428571428e-06,
+      "loss": 1.0633,
+      "step": 16430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.687102857142858e-06,
+      "loss": 1.5457,
+      "step": 16431
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.686834285714286e-06,
+      "loss": 1.2978,
+      "step": 16432
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.686565714285714e-06,
+      "loss": 0.8915,
+      "step": 16433
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.686297142857144e-06,
+      "loss": 1.3448,
+      "step": 16434
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.686028571428572e-06,
+      "loss": 1.0856,
+      "step": 16435
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.68576e-06,
+      "loss": 1.0479,
+      "step": 16436
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.685491428571428e-06,
+      "loss": 0.9356,
+      "step": 16437
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.685222857142858e-06,
+      "loss": 1.4163,
+      "step": 16438
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.684954285714286e-06,
+      "loss": 1.335,
+      "step": 16439
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.684685714285715e-06,
+      "loss": 1.2474,
+      "step": 16440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.684417142857143e-06,
+      "loss": 1.0395,
+      "step": 16441
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.684148571428573e-06,
+      "loss": 0.9733,
+      "step": 16442
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.68388e-06,
+      "loss": 1.5134,
+      "step": 16443
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.683611428571429e-06,
+      "loss": 0.9846,
+      "step": 16444
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.683342857142857e-06,
+      "loss": 1.0102,
+      "step": 16445
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.683074285714285e-06,
+      "loss": 0.9937,
+      "step": 16446
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.682805714285715e-06,
+      "loss": 1.523,
+      "step": 16447
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.682537142857143e-06,
+      "loss": 1.0319,
+      "step": 16448
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.682268571428572e-06,
+      "loss": 1.0449,
+      "step": 16449
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.682e-06,
+      "loss": 1.3611,
+      "step": 16450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.68173142857143e-06,
+      "loss": 1.3366,
+      "step": 16451
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.681462857142858e-06,
+      "loss": 1.2622,
+      "step": 16452
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.681194285714286e-06,
+      "loss": 1.1667,
+      "step": 16453
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.680925714285714e-06,
+      "loss": 1.1765,
+      "step": 16454
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.680657142857144e-06,
+      "loss": 1.2574,
+      "step": 16455
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.68038857142857e-06,
+      "loss": 0.8806,
+      "step": 16456
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.68012e-06,
+      "loss": 1.4872,
+      "step": 16457
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.679851428571429e-06,
+      "loss": 1.3703,
+      "step": 16458
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.679582857142857e-06,
+      "loss": 1.014,
+      "step": 16459
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.679314285714287e-06,
+      "loss": 0.74,
+      "step": 16460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.679045714285715e-06,
+      "loss": 1.6046,
+      "step": 16461
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.678777142857145e-06,
+      "loss": 1.3111,
+      "step": 16462
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.678508571428571e-06,
+      "loss": 1.2387,
+      "step": 16463
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.678240000000001e-06,
+      "loss": 1.0761,
+      "step": 16464
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.67797142857143e-06,
+      "loss": 1.1951,
+      "step": 16465
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.677702857142857e-06,
+      "loss": 1.1107,
+      "step": 16466
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.677434285714286e-06,
+      "loss": 1.2523,
+      "step": 16467
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.677165714285715e-06,
+      "loss": 1.4353,
+      "step": 16468
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.676897142857142e-06,
+      "loss": 0.6028,
+      "step": 16469
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.676628571428572e-06,
+      "loss": 1.2248,
+      "step": 16470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.67636e-06,
+      "loss": 0.7481,
+      "step": 16471
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.676091428571428e-06,
+      "loss": 1.3774,
+      "step": 16472
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.675822857142858e-06,
+      "loss": 1.4715,
+      "step": 16473
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.675554285714286e-06,
+      "loss": 1.1747,
+      "step": 16474
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.675285714285716e-06,
+      "loss": 1.3912,
+      "step": 16475
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.675017142857143e-06,
+      "loss": 1.3765,
+      "step": 16476
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.674748571428572e-06,
+      "loss": 1.3548,
+      "step": 16477
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.67448e-06,
+      "loss": 1.467,
+      "step": 16478
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.674211428571429e-06,
+      "loss": 1.2919,
+      "step": 16479
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.673942857142857e-06,
+      "loss": 0.8411,
+      "step": 16480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.673674285714287e-06,
+      "loss": 1.3159,
+      "step": 16481
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.673405714285713e-06,
+      "loss": 1.1898,
+      "step": 16482
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.673137142857143e-06,
+      "loss": 1.3391,
+      "step": 16483
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.672868571428571e-06,
+      "loss": 1.5392,
+      "step": 16484
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.672600000000001e-06,
+      "loss": 1.2122,
+      "step": 16485
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.67233142857143e-06,
+      "loss": 0.6251,
+      "step": 16486
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.672062857142858e-06,
+      "loss": 0.9456,
+      "step": 16487
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.671794285714287e-06,
+      "loss": 1.6857,
+      "step": 16488
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.671525714285714e-06,
+      "loss": 1.312,
+      "step": 16489
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.671257142857144e-06,
+      "loss": 1.531,
+      "step": 16490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.670988571428572e-06,
+      "loss": 0.9988,
+      "step": 16491
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.67072e-06,
+      "loss": 1.1284,
+      "step": 16492
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.670451428571428e-06,
+      "loss": 0.9136,
+      "step": 16493
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.670182857142858e-06,
+      "loss": 0.8921,
+      "step": 16494
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.669914285714285e-06,
+      "loss": 0.7039,
+      "step": 16495
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.669645714285715e-06,
+      "loss": 1.0371,
+      "step": 16496
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.669377142857143e-06,
+      "loss": 1.204,
+      "step": 16497
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.669108571428573e-06,
+      "loss": 1.0332,
+      "step": 16498
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.66884e-06,
+      "loss": 1.3134,
+      "step": 16499
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.668571428571429e-06,
+      "loss": 1.2574,
+      "step": 16500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.668302857142859e-06,
+      "loss": 1.2304,
+      "step": 16501
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.668034285714285e-06,
+      "loss": 0.8747,
+      "step": 16502
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.667765714285715e-06,
+      "loss": 1.3766,
+      "step": 16503
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.667497142857143e-06,
+      "loss": 1.3043,
+      "step": 16504
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.667228571428572e-06,
+      "loss": 1.1718,
+      "step": 16505
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.66696e-06,
+      "loss": 1.3287,
+      "step": 16506
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.66669142857143e-06,
+      "loss": 1.286,
+      "step": 16507
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.666422857142858e-06,
+      "loss": 1.2803,
+      "step": 16508
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.666154285714286e-06,
+      "loss": 1.0002,
+      "step": 16509
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.665885714285714e-06,
+      "loss": 0.7851,
+      "step": 16510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.665617142857144e-06,
+      "loss": 1.348,
+      "step": 16511
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.665348571428572e-06,
+      "loss": 1.2588,
+      "step": 16512
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.66508e-06,
+      "loss": 1.068,
+      "step": 16513
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.66481142857143e-06,
+      "loss": 1.5036,
+      "step": 16514
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.664542857142857e-06,
+      "loss": 1.2379,
+      "step": 16515
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.664274285714287e-06,
+      "loss": 1.3842,
+      "step": 16516
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.664005714285715e-06,
+      "loss": 1.0685,
+      "step": 16517
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.663737142857143e-06,
+      "loss": 1.095,
+      "step": 16518
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.663468571428571e-06,
+      "loss": 2.0025,
+      "step": 16519
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.663200000000001e-06,
+      "loss": 1.065,
+      "step": 16520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.662931428571429e-06,
+      "loss": 1.0638,
+      "step": 16521
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.662662857142857e-06,
+      "loss": 1.5076,
+      "step": 16522
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.662394285714285e-06,
+      "loss": 0.9709,
+      "step": 16523
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.662125714285715e-06,
+      "loss": 0.8975,
+      "step": 16524
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.661857142857144e-06,
+      "loss": 1.1766,
+      "step": 16525
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.661588571428572e-06,
+      "loss": 1.6087,
+      "step": 16526
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.661320000000002e-06,
+      "loss": 0.9934,
+      "step": 16527
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.661051428571428e-06,
+      "loss": 1.5391,
+      "step": 16528
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.660782857142858e-06,
+      "loss": 1.257,
+      "step": 16529
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.660514285714286e-06,
+      "loss": 1.3063,
+      "step": 16530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.660245714285714e-06,
+      "loss": 1.3591,
+      "step": 16531
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.659977142857142e-06,
+      "loss": 0.9077,
+      "step": 16532
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.659708571428572e-06,
+      "loss": 1.1478,
+      "step": 16533
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.65944e-06,
+      "loss": 1.1643,
+      "step": 16534
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.659171428571429e-06,
+      "loss": 1.337,
+      "step": 16535
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.658902857142857e-06,
+      "loss": 1.3099,
+      "step": 16536
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.658634285714287e-06,
+      "loss": 1.3848,
+      "step": 16537
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.658365714285715e-06,
+      "loss": 1.222,
+      "step": 16538
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.658097142857143e-06,
+      "loss": 0.9781,
+      "step": 16539
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.657828571428573e-06,
+      "loss": 1.2395,
+      "step": 16540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.65756e-06,
+      "loss": 1.1327,
+      "step": 16541
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.65729142857143e-06,
+      "loss": 1.4102,
+      "step": 16542
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.657022857142857e-06,
+      "loss": 1.1225,
+      "step": 16543
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.656754285714287e-06,
+      "loss": 1.2202,
+      "step": 16544
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.656485714285714e-06,
+      "loss": 0.9802,
+      "step": 16545
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.656217142857144e-06,
+      "loss": 1.1887,
+      "step": 16546
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.655948571428572e-06,
+      "loss": 1.2513,
+      "step": 16547
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.65568e-06,
+      "loss": 1.2789,
+      "step": 16548
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.655411428571428e-06,
+      "loss": 1.1768,
+      "step": 16549
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.655142857142858e-06,
+      "loss": 1.4999,
+      "step": 16550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.654874285714286e-06,
+      "loss": 1.1951,
+      "step": 16551
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.654605714285714e-06,
+      "loss": 1.0629,
+      "step": 16552
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.654337142857144e-06,
+      "loss": 1.2005,
+      "step": 16553
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.65406857142857e-06,
+      "loss": 1.3091,
+      "step": 16554
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.6538e-06,
+      "loss": 1.2218,
+      "step": 16555
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.653531428571429e-06,
+      "loss": 1.3183,
+      "step": 16556
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.653262857142859e-06,
+      "loss": 1.409,
+      "step": 16557
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.652994285714285e-06,
+      "loss": 1.2179,
+      "step": 16558
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.652725714285715e-06,
+      "loss": 0.9118,
+      "step": 16559
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.652457142857143e-06,
+      "loss": 1.0772,
+      "step": 16560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.652188571428571e-06,
+      "loss": 1.0342,
+      "step": 16561
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.65192e-06,
+      "loss": 1.2626,
+      "step": 16562
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.65165142857143e-06,
+      "loss": 1.3854,
+      "step": 16563
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.651382857142858e-06,
+      "loss": 1.6284,
+      "step": 16564
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.651114285714286e-06,
+      "loss": 1.3743,
+      "step": 16565
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.650845714285716e-06,
+      "loss": 0.9464,
+      "step": 16566
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.650577142857144e-06,
+      "loss": 1.2005,
+      "step": 16567
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.650308571428572e-06,
+      "loss": 0.8831,
+      "step": 16568
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.65004e-06,
+      "loss": 1.3206,
+      "step": 16569
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.64977142857143e-06,
+      "loss": 0.9329,
+      "step": 16570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.649502857142857e-06,
+      "loss": 1.2886,
+      "step": 16571
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.649234285714286e-06,
+      "loss": 1.1465,
+      "step": 16572
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.648965714285715e-06,
+      "loss": 1.0598,
+      "step": 16573
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.648697142857143e-06,
+      "loss": 1.3006,
+      "step": 16574
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.648428571428571e-06,
+      "loss": 0.9502,
+      "step": 16575
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.64816e-06,
+      "loss": 1.3492,
+      "step": 16576
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.647891428571429e-06,
+      "loss": 1.4692,
+      "step": 16577
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.647622857142857e-06,
+      "loss": 1.2245,
+      "step": 16578
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.647354285714287e-06,
+      "loss": 0.8496,
+      "step": 16579
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.647085714285715e-06,
+      "loss": 0.7481,
+      "step": 16580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.646817142857143e-06,
+      "loss": 1.5631,
+      "step": 16581
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.646548571428572e-06,
+      "loss": 0.9244,
+      "step": 16582
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.646280000000001e-06,
+      "loss": 1.029,
+      "step": 16583
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.646011428571428e-06,
+      "loss": 1.2403,
+      "step": 16584
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.645742857142858e-06,
+      "loss": 1.682,
+      "step": 16585
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.645474285714286e-06,
+      "loss": 1.2982,
+      "step": 16586
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.645205714285714e-06,
+      "loss": 0.8253,
+      "step": 16587
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.644937142857142e-06,
+      "loss": 1.2682,
+      "step": 16588
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.644668571428572e-06,
+      "loss": 1.4127,
+      "step": 16589
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.6444e-06,
+      "loss": 1.536,
+      "step": 16590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.644131428571429e-06,
+      "loss": 0.6717,
+      "step": 16591
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.643862857142858e-06,
+      "loss": 1.0166,
+      "step": 16592
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.643594285714287e-06,
+      "loss": 1.0836,
+      "step": 16593
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.643325714285715e-06,
+      "loss": 1.2649,
+      "step": 16594
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.643057142857143e-06,
+      "loss": 0.724,
+      "step": 16595
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.642788571428573e-06,
+      "loss": 1.6497,
+      "step": 16596
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.64252e-06,
+      "loss": 1.3536,
+      "step": 16597
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.64225142857143e-06,
+      "loss": 1.3685,
+      "step": 16598
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.641982857142857e-06,
+      "loss": 1.3141,
+      "step": 16599
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.641714285714286e-06,
+      "loss": 1.5285,
+      "step": 16600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.641445714285714e-06,
+      "loss": 1.1879,
+      "step": 16601
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.641177142857144e-06,
+      "loss": 0.9125,
+      "step": 16602
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.640908571428572e-06,
+      "loss": 1.3507,
+      "step": 16603
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.64064e-06,
+      "loss": 1.2109,
+      "step": 16604
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.64037142857143e-06,
+      "loss": 1.0417,
+      "step": 16605
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.640102857142858e-06,
+      "loss": 1.274,
+      "step": 16606
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.639834285714286e-06,
+      "loss": 1.0093,
+      "step": 16607
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.639565714285714e-06,
+      "loss": 1.0319,
+      "step": 16608
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.639297142857144e-06,
+      "loss": 1.1185,
+      "step": 16609
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.63902857142857e-06,
+      "loss": 1.1715,
+      "step": 16610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.63876e-06,
+      "loss": 1.5617,
+      "step": 16611
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.638491428571429e-06,
+      "loss": 1.4682,
+      "step": 16612
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.638222857142859e-06,
+      "loss": 1.531,
+      "step": 16613
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.637954285714285e-06,
+      "loss": 1.5579,
+      "step": 16614
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.637685714285715e-06,
+      "loss": 1.2704,
+      "step": 16615
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.637417142857143e-06,
+      "loss": 1.1933,
+      "step": 16616
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.637148571428571e-06,
+      "loss": 1.5426,
+      "step": 16617
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.636880000000001e-06,
+      "loss": 1.2747,
+      "step": 16618
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.63661142857143e-06,
+      "loss": 0.6208,
+      "step": 16619
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.636342857142858e-06,
+      "loss": 1.0855,
+      "step": 16620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.636074285714286e-06,
+      "loss": 1.024,
+      "step": 16621
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.635805714285716e-06,
+      "loss": 1.1424,
+      "step": 16622
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.635537142857142e-06,
+      "loss": 1.5014,
+      "step": 16623
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.635268571428572e-06,
+      "loss": 1.3154,
+      "step": 16624
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.635e-06,
+      "loss": 1.3578,
+      "step": 16625
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.63473142857143e-06,
+      "loss": 1.5266,
+      "step": 16626
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.634462857142856e-06,
+      "loss": 1.3997,
+      "step": 16627
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.634194285714286e-06,
+      "loss": 1.2171,
+      "step": 16628
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.633925714285714e-06,
+      "loss": 1.0289,
+      "step": 16629
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.633657142857143e-06,
+      "loss": 1.57,
+      "step": 16630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.633388571428573e-06,
+      "loss": 1.0588,
+      "step": 16631
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.63312e-06,
+      "loss": 1.4839,
+      "step": 16632
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.632851428571429e-06,
+      "loss": 1.3948,
+      "step": 16633
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.632582857142857e-06,
+      "loss": 1.2147,
+      "step": 16634
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.632314285714287e-06,
+      "loss": 1.0643,
+      "step": 16635
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.632045714285715e-06,
+      "loss": 1.0706,
+      "step": 16636
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.631777142857143e-06,
+      "loss": 1.2596,
+      "step": 16637
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.631508571428571e-06,
+      "loss": 1.2351,
+      "step": 16638
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.631240000000001e-06,
+      "loss": 1.5748,
+      "step": 16639
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.630971428571428e-06,
+      "loss": 1.4981,
+      "step": 16640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.630702857142858e-06,
+      "loss": 1.5358,
+      "step": 16641
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.630434285714286e-06,
+      "loss": 1.4887,
+      "step": 16642
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.630165714285714e-06,
+      "loss": 1.2206,
+      "step": 16643
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.629897142857144e-06,
+      "loss": 1.373,
+      "step": 16644
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.629628571428572e-06,
+      "loss": 1.2355,
+      "step": 16645
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.62936e-06,
+      "loss": 1.4567,
+      "step": 16646
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.629091428571428e-06,
+      "loss": 0.9645,
+      "step": 16647
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.628822857142858e-06,
+      "loss": 1.542,
+      "step": 16648
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.628554285714286e-06,
+      "loss": 1.3165,
+      "step": 16649
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.628285714285715e-06,
+      "loss": 1.5318,
+      "step": 16650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.628017142857143e-06,
+      "loss": 1.254,
+      "step": 16651
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.627748571428573e-06,
+      "loss": 1.2852,
+      "step": 16652
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.62748e-06,
+      "loss": 1.3894,
+      "step": 16653
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.627211428571429e-06,
+      "loss": 1.0996,
+      "step": 16654
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.626942857142857e-06,
+      "loss": 1.2951,
+      "step": 16655
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.626674285714285e-06,
+      "loss": 1.5475,
+      "step": 16656
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.626405714285715e-06,
+      "loss": 1.7851,
+      "step": 16657
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.626137142857143e-06,
+      "loss": 1.0899,
+      "step": 16658
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.625868571428573e-06,
+      "loss": 1.5181,
+      "step": 16659
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.6256e-06,
+      "loss": 1.1126,
+      "step": 16660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.62533142857143e-06,
+      "loss": 1.5309,
+      "step": 16661
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.625062857142858e-06,
+      "loss": 1.3771,
+      "step": 16662
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.624794285714286e-06,
+      "loss": 1.54,
+      "step": 16663
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.624525714285714e-06,
+      "loss": 1.3836,
+      "step": 16664
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.624257142857144e-06,
+      "loss": 1.2353,
+      "step": 16665
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.62398857142857e-06,
+      "loss": 1.0856,
+      "step": 16666
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.62372e-06,
+      "loss": 1.0398,
+      "step": 16667
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.623451428571429e-06,
+      "loss": 1.1333,
+      "step": 16668
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.623182857142857e-06,
+      "loss": 1.5016,
+      "step": 16669
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.622914285714287e-06,
+      "loss": 0.9009,
+      "step": 16670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.622645714285715e-06,
+      "loss": 1.1323,
+      "step": 16671
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.622377142857145e-06,
+      "loss": 0.8664,
+      "step": 16672
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.622108571428571e-06,
+      "loss": 0.8872,
+      "step": 16673
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.621840000000001e-06,
+      "loss": 0.7867,
+      "step": 16674
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.62157142857143e-06,
+      "loss": 1.089,
+      "step": 16675
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.621302857142857e-06,
+      "loss": 1.1181,
+      "step": 16676
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.621034285714286e-06,
+      "loss": 1.1457,
+      "step": 16677
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.620765714285715e-06,
+      "loss": 1.1543,
+      "step": 16678
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.620497142857142e-06,
+      "loss": 1.2868,
+      "step": 16679
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.620228571428572e-06,
+      "loss": 1.7502,
+      "step": 16680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.61996e-06,
+      "loss": 1.1736,
+      "step": 16681
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.61969142857143e-06,
+      "loss": 1.2909,
+      "step": 16682
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.619422857142858e-06,
+      "loss": 0.7637,
+      "step": 16683
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.619154285714286e-06,
+      "loss": 1.3273,
+      "step": 16684
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.618885714285716e-06,
+      "loss": 1.4524,
+      "step": 16685
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.618617142857143e-06,
+      "loss": 1.052,
+      "step": 16686
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.618348571428572e-06,
+      "loss": 1.1782,
+      "step": 16687
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.61808e-06,
+      "loss": 0.9536,
+      "step": 16688
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.617811428571429e-06,
+      "loss": 1.1371,
+      "step": 16689
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.617542857142857e-06,
+      "loss": 1.4695,
+      "step": 16690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.617274285714287e-06,
+      "loss": 1.0344,
+      "step": 16691
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.617005714285713e-06,
+      "loss": 1.2749,
+      "step": 16692
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.616737142857143e-06,
+      "loss": 0.8323,
+      "step": 16693
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.616468571428571e-06,
+      "loss": 1.1815,
+      "step": 16694
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.616200000000001e-06,
+      "loss": 1.3103,
+      "step": 16695
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.61593142857143e-06,
+      "loss": 1.0318,
+      "step": 16696
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.615662857142858e-06,
+      "loss": 1.3467,
+      "step": 16697
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.615394285714287e-06,
+      "loss": 1.08,
+      "step": 16698
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.615125714285714e-06,
+      "loss": 1.1895,
+      "step": 16699
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.614857142857144e-06,
+      "loss": 0.9112,
+      "step": 16700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.614588571428572e-06,
+      "loss": 1.3482,
+      "step": 16701
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.61432e-06,
+      "loss": 1.1535,
+      "step": 16702
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.614051428571428e-06,
+      "loss": 1.0042,
+      "step": 16703
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.613782857142858e-06,
+      "loss": 1.3267,
+      "step": 16704
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.613514285714285e-06,
+      "loss": 1.0714,
+      "step": 16705
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.613245714285715e-06,
+      "loss": 1.246,
+      "step": 16706
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.612977142857143e-06,
+      "loss": 1.1758,
+      "step": 16707
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.612708571428573e-06,
+      "loss": 1.6028,
+      "step": 16708
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.61244e-06,
+      "loss": 0.9789,
+      "step": 16709
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.612171428571429e-06,
+      "loss": 1.3362,
+      "step": 16710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.611902857142859e-06,
+      "loss": 1.217,
+      "step": 16711
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.611634285714285e-06,
+      "loss": 0.9401,
+      "step": 16712
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.611365714285715e-06,
+      "loss": 0.8207,
+      "step": 16713
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.611097142857143e-06,
+      "loss": 1.2276,
+      "step": 16714
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.610828571428572e-06,
+      "loss": 1.1552,
+      "step": 16715
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.61056e-06,
+      "loss": 1.2757,
+      "step": 16716
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.61029142857143e-06,
+      "loss": 1.0162,
+      "step": 16717
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.610022857142858e-06,
+      "loss": 1.1909,
+      "step": 16718
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.609754285714286e-06,
+      "loss": 1.0025,
+      "step": 16719
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.609485714285714e-06,
+      "loss": 1.2156,
+      "step": 16720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.609217142857144e-06,
+      "loss": 1.3582,
+      "step": 16721
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.608948571428572e-06,
+      "loss": 1.1662,
+      "step": 16722
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.60868e-06,
+      "loss": 1.0048,
+      "step": 16723
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.60841142857143e-06,
+      "loss": 1.0824,
+      "step": 16724
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.608142857142857e-06,
+      "loss": 1.1173,
+      "step": 16725
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.607874285714287e-06,
+      "loss": 1.1287,
+      "step": 16726
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.607605714285715e-06,
+      "loss": 1.515,
+      "step": 16727
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.607337142857143e-06,
+      "loss": 0.9001,
+      "step": 16728
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.607068571428571e-06,
+      "loss": 1.433,
+      "step": 16729
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.606800000000001e-06,
+      "loss": 1.3866,
+      "step": 16730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.606531428571429e-06,
+      "loss": 1.4385,
+      "step": 16731
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.606262857142857e-06,
+      "loss": 1.5033,
+      "step": 16732
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.605994285714285e-06,
+      "loss": 1.079,
+      "step": 16733
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.605725714285715e-06,
+      "loss": 1.1972,
+      "step": 16734
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.605457142857143e-06,
+      "loss": 1.3057,
+      "step": 16735
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.605188571428572e-06,
+      "loss": 1.064,
+      "step": 16736
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.604920000000002e-06,
+      "loss": 1.2357,
+      "step": 16737
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.604651428571428e-06,
+      "loss": 1.3717,
+      "step": 16738
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.604382857142858e-06,
+      "loss": 1.2126,
+      "step": 16739
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.604114285714286e-06,
+      "loss": 1.7557,
+      "step": 16740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.603845714285716e-06,
+      "loss": 1.0762,
+      "step": 16741
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.603577142857142e-06,
+      "loss": 1.2133,
+      "step": 16742
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.603308571428572e-06,
+      "loss": 1.107,
+      "step": 16743
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.60304e-06,
+      "loss": 1.3538,
+      "step": 16744
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.602771428571429e-06,
+      "loss": 1.1097,
+      "step": 16745
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.602502857142857e-06,
+      "loss": 1.4853,
+      "step": 16746
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.602234285714287e-06,
+      "loss": 0.7194,
+      "step": 16747
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.601965714285713e-06,
+      "loss": 1.2647,
+      "step": 16748
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.601697142857143e-06,
+      "loss": 1.2856,
+      "step": 16749
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.601428571428573e-06,
+      "loss": 1.2174,
+      "step": 16750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.60116e-06,
+      "loss": 0.9558,
+      "step": 16751
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.60089142857143e-06,
+      "loss": 1.3257,
+      "step": 16752
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.600622857142857e-06,
+      "loss": 1.3507,
+      "step": 16753
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.600354285714287e-06,
+      "loss": 1.0116,
+      "step": 16754
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.600085714285714e-06,
+      "loss": 1.0589,
+      "step": 16755
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.599817142857144e-06,
+      "loss": 1.0537,
+      "step": 16756
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.599548571428572e-06,
+      "loss": 1.256,
+      "step": 16757
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.59928e-06,
+      "loss": 1.3892,
+      "step": 16758
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.599011428571428e-06,
+      "loss": 1.3069,
+      "step": 16759
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.598742857142858e-06,
+      "loss": 1.0923,
+      "step": 16760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.598474285714285e-06,
+      "loss": 1.3796,
+      "step": 16761
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.598205714285714e-06,
+      "loss": 1.061,
+      "step": 16762
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.597937142857144e-06,
+      "loss": 1.2199,
+      "step": 16763
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.597668571428572e-06,
+      "loss": 1.1487,
+      "step": 16764
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.5974e-06,
+      "loss": 0.9978,
+      "step": 16765
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.597131428571429e-06,
+      "loss": 0.7687,
+      "step": 16766
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.596862857142859e-06,
+      "loss": 1.1638,
+      "step": 16767
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.596594285714285e-06,
+      "loss": 1.2042,
+      "step": 16768
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.596325714285715e-06,
+      "loss": 1.4258,
+      "step": 16769
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.596057142857143e-06,
+      "loss": 1.239,
+      "step": 16770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.595788571428571e-06,
+      "loss": 1.1223,
+      "step": 16771
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.59552e-06,
+      "loss": 0.9226,
+      "step": 16772
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.59525142857143e-06,
+      "loss": 0.9998,
+      "step": 16773
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.594982857142856e-06,
+      "loss": 1.0209,
+      "step": 16774
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.594714285714286e-06,
+      "loss": 1.2049,
+      "step": 16775
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.594445714285716e-06,
+      "loss": 0.8511,
+      "step": 16776
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.594177142857144e-06,
+      "loss": 0.9793,
+      "step": 16777
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.593908571428572e-06,
+      "loss": 1.4992,
+      "step": 16778
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.59364e-06,
+      "loss": 1.2483,
+      "step": 16779
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.59337142857143e-06,
+      "loss": 1.3909,
+      "step": 16780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.593102857142857e-06,
+      "loss": 0.9404,
+      "step": 16781
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.592834285714286e-06,
+      "loss": 1.4385,
+      "step": 16782
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.592565714285715e-06,
+      "loss": 0.9453,
+      "step": 16783
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.592297142857143e-06,
+      "loss": 1.2602,
+      "step": 16784
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.592028571428571e-06,
+      "loss": 0.9471,
+      "step": 16785
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.59176e-06,
+      "loss": 1.4026,
+      "step": 16786
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.591491428571429e-06,
+      "loss": 1.479,
+      "step": 16787
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.591222857142857e-06,
+      "loss": 0.9931,
+      "step": 16788
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.590954285714287e-06,
+      "loss": 1.2071,
+      "step": 16789
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.590685714285715e-06,
+      "loss": 1.1706,
+      "step": 16790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.590417142857143e-06,
+      "loss": 1.0463,
+      "step": 16791
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.590148571428572e-06,
+      "loss": 1.2252,
+      "step": 16792
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.589880000000001e-06,
+      "loss": 0.9352,
+      "step": 16793
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.589611428571428e-06,
+      "loss": 1.094,
+      "step": 16794
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.589342857142858e-06,
+      "loss": 1.1012,
+      "step": 16795
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.589074285714286e-06,
+      "loss": 1.1456,
+      "step": 16796
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.588805714285714e-06,
+      "loss": 1.1357,
+      "step": 16797
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.588537142857142e-06,
+      "loss": 1.1829,
+      "step": 16798
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.588268571428572e-06,
+      "loss": 1.3991,
+      "step": 16799
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.588e-06,
+      "loss": 1.3238,
+      "step": 16800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.587731428571429e-06,
+      "loss": 1.1106,
+      "step": 16801
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.587462857142858e-06,
+      "loss": 1.2622,
+      "step": 16802
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.587194285714287e-06,
+      "loss": 1.2905,
+      "step": 16803
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.586925714285715e-06,
+      "loss": 1.0818,
+      "step": 16804
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.586657142857143e-06,
+      "loss": 1.3293,
+      "step": 16805
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.586388571428573e-06,
+      "loss": 1.3174,
+      "step": 16806
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58612e-06,
+      "loss": 1.4313,
+      "step": 16807
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58585142857143e-06,
+      "loss": 1.0427,
+      "step": 16808
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.585582857142857e-06,
+      "loss": 1.0638,
+      "step": 16809
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.585314285714287e-06,
+      "loss": 0.972,
+      "step": 16810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.585045714285714e-06,
+      "loss": 0.8156,
+      "step": 16811
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.584777142857144e-06,
+      "loss": 1.6201,
+      "step": 16812
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.584508571428572e-06,
+      "loss": 1.1751,
+      "step": 16813
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58424e-06,
+      "loss": 0.9251,
+      "step": 16814
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58397142857143e-06,
+      "loss": 0.963,
+      "step": 16815
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.583702857142858e-06,
+      "loss": 1.3495,
+      "step": 16816
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.583434285714286e-06,
+      "loss": 0.8479,
+      "step": 16817
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.583165714285714e-06,
+      "loss": 1.1625,
+      "step": 16818
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.582897142857144e-06,
+      "loss": 1.5268,
+      "step": 16819
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58262857142857e-06,
+      "loss": 1.172,
+      "step": 16820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58236e-06,
+      "loss": 1.5034,
+      "step": 16821
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.582091428571429e-06,
+      "loss": 1.3298,
+      "step": 16822
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.581822857142859e-06,
+      "loss": 1.2952,
+      "step": 16823
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.581554285714285e-06,
+      "loss": 1.1628,
+      "step": 16824
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.581285714285715e-06,
+      "loss": 1.3582,
+      "step": 16825
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.581017142857143e-06,
+      "loss": 1.058,
+      "step": 16826
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.580748571428571e-06,
+      "loss": 1.5342,
+      "step": 16827
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.580480000000001e-06,
+      "loss": 1.1542,
+      "step": 16828
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58021142857143e-06,
+      "loss": 1.2508,
+      "step": 16829
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.579942857142858e-06,
+      "loss": 1.3194,
+      "step": 16830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.579674285714286e-06,
+      "loss": 1.356,
+      "step": 16831
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.579405714285716e-06,
+      "loss": 1.572,
+      "step": 16832
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.579137142857144e-06,
+      "loss": 1.4186,
+      "step": 16833
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.578868571428572e-06,
+      "loss": 1.0505,
+      "step": 16834
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.5786e-06,
+      "loss": 1.4358,
+      "step": 16835
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.57833142857143e-06,
+      "loss": 1.3384,
+      "step": 16836
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.578062857142856e-06,
+      "loss": 1.7743,
+      "step": 16837
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.577794285714286e-06,
+      "loss": 0.972,
+      "step": 16838
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.577525714285714e-06,
+      "loss": 1.1718,
+      "step": 16839
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.577257142857143e-06,
+      "loss": 1.317,
+      "step": 16840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.576988571428573e-06,
+      "loss": 1.102,
+      "step": 16841
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.57672e-06,
+      "loss": 1.1805,
+      "step": 16842
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.576451428571429e-06,
+      "loss": 1.1721,
+      "step": 16843
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.576182857142857e-06,
+      "loss": 1.1069,
+      "step": 16844
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.575914285714287e-06,
+      "loss": 1.2161,
+      "step": 16845
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.575645714285715e-06,
+      "loss": 1.3613,
+      "step": 16846
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.575377142857143e-06,
+      "loss": 1.2134,
+      "step": 16847
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.575108571428571e-06,
+      "loss": 1.3577,
+      "step": 16848
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.574840000000001e-06,
+      "loss": 1.4047,
+      "step": 16849
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.574571428571428e-06,
+      "loss": 0.9,
+      "step": 16850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.574302857142858e-06,
+      "loss": 1.0253,
+      "step": 16851
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.574034285714286e-06,
+      "loss": 1.3315,
+      "step": 16852
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.573765714285714e-06,
+      "loss": 1.4868,
+      "step": 16853
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.573497142857142e-06,
+      "loss": 1.255,
+      "step": 16854
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.573228571428572e-06,
+      "loss": 1.4384,
+      "step": 16855
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.572960000000002e-06,
+      "loss": 1.1777,
+      "step": 16856
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.572691428571428e-06,
+      "loss": 1.341,
+      "step": 16857
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.572422857142858e-06,
+      "loss": 1.4307,
+      "step": 16858
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.572154285714286e-06,
+      "loss": 1.1977,
+      "step": 16859
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.571885714285715e-06,
+      "loss": 1.0753,
+      "step": 16860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.571617142857143e-06,
+      "loss": 0.8458,
+      "step": 16861
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.571348571428573e-06,
+      "loss": 2.0416,
+      "step": 16862
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.57108e-06,
+      "loss": 1.1846,
+      "step": 16863
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.570811428571429e-06,
+      "loss": 1.0254,
+      "step": 16864
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.570542857142857e-06,
+      "loss": 1.6734,
+      "step": 16865
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.570274285714285e-06,
+      "loss": 1.034,
+      "step": 16866
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.570005714285714e-06,
+      "loss": 1.3126,
+      "step": 16867
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.569737142857143e-06,
+      "loss": 1.0593,
+      "step": 16868
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.569468571428573e-06,
+      "loss": 1.4016,
+      "step": 16869
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.5692e-06,
+      "loss": 1.2226,
+      "step": 16870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.56893142857143e-06,
+      "loss": 1.0313,
+      "step": 16871
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.568662857142858e-06,
+      "loss": 1.389,
+      "step": 16872
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.568394285714286e-06,
+      "loss": 1.7561,
+      "step": 16873
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.568125714285714e-06,
+      "loss": 0.9641,
+      "step": 16874
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.567857142857144e-06,
+      "loss": 1.4256,
+      "step": 16875
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.56758857142857e-06,
+      "loss": 1.36,
+      "step": 16876
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.56732e-06,
+      "loss": 1.1516,
+      "step": 16877
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.567051428571429e-06,
+      "loss": 0.9697,
+      "step": 16878
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.566782857142857e-06,
+      "loss": 1.3378,
+      "step": 16879
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.566514285714285e-06,
+      "loss": 1.5373,
+      "step": 16880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.566245714285715e-06,
+      "loss": 1.3205,
+      "step": 16881
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.565977142857145e-06,
+      "loss": 1.0496,
+      "step": 16882
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.565708571428571e-06,
+      "loss": 1.0911,
+      "step": 16883
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.565440000000001e-06,
+      "loss": 0.8396,
+      "step": 16884
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.56517142857143e-06,
+      "loss": 1.5042,
+      "step": 16885
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.564902857142857e-06,
+      "loss": 1.1974,
+      "step": 16886
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.564634285714286e-06,
+      "loss": 0.863,
+      "step": 16887
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.564365714285715e-06,
+      "loss": 1.119,
+      "step": 16888
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.564097142857142e-06,
+      "loss": 1.1851,
+      "step": 16889
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.563828571428572e-06,
+      "loss": 1.1582,
+      "step": 16890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.56356e-06,
+      "loss": 1.1527,
+      "step": 16891
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.56329142857143e-06,
+      "loss": 1.2491,
+      "step": 16892
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.563022857142856e-06,
+      "loss": 1.028,
+      "step": 16893
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.562754285714286e-06,
+      "loss": 1.0943,
+      "step": 16894
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.562485714285716e-06,
+      "loss": 1.3561,
+      "step": 16895
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.562217142857143e-06,
+      "loss": 1.3186,
+      "step": 16896
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.561948571428572e-06,
+      "loss": 0.9432,
+      "step": 16897
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.56168e-06,
+      "loss": 1.069,
+      "step": 16898
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.561411428571429e-06,
+      "loss": 1.234,
+      "step": 16899
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.561142857142857e-06,
+      "loss": 1.2908,
+      "step": 16900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.560874285714287e-06,
+      "loss": 0.8854,
+      "step": 16901
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.560605714285713e-06,
+      "loss": 1.3288,
+      "step": 16902
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.560337142857143e-06,
+      "loss": 1.2649,
+      "step": 16903
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.560068571428571e-06,
+      "loss": 1.0697,
+      "step": 16904
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.559800000000001e-06,
+      "loss": 1.5107,
+      "step": 16905
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.559531428571428e-06,
+      "loss": 1.4877,
+      "step": 16906
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.559262857142858e-06,
+      "loss": 1.2368,
+      "step": 16907
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.558994285714287e-06,
+      "loss": 1.4467,
+      "step": 16908
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.558725714285714e-06,
+      "loss": 1.2185,
+      "step": 16909
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.558457142857144e-06,
+      "loss": 1.0155,
+      "step": 16910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.558188571428572e-06,
+      "loss": 1.4446,
+      "step": 16911
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.55792e-06,
+      "loss": 0.7438,
+      "step": 16912
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.557651428571428e-06,
+      "loss": 1.2844,
+      "step": 16913
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.557382857142858e-06,
+      "loss": 0.9595,
+      "step": 16914
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.557114285714286e-06,
+      "loss": 1.582,
+      "step": 16915
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.556845714285715e-06,
+      "loss": 1.2175,
+      "step": 16916
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.556577142857143e-06,
+      "loss": 1.5168,
+      "step": 16917
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.556308571428573e-06,
+      "loss": 1.2121,
+      "step": 16918
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.556039999999999e-06,
+      "loss": 0.8806,
+      "step": 16919
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.555771428571429e-06,
+      "loss": 1.4311,
+      "step": 16920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.555502857142859e-06,
+      "loss": 1.3906,
+      "step": 16921
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.555234285714285e-06,
+      "loss": 1.1622,
+      "step": 16922
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.554965714285715e-06,
+      "loss": 1.2617,
+      "step": 16923
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.554697142857143e-06,
+      "loss": 1.3157,
+      "step": 16924
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.554428571428572e-06,
+      "loss": 1.2848,
+      "step": 16925
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.55416e-06,
+      "loss": 1.1417,
+      "step": 16926
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.55389142857143e-06,
+      "loss": 1.4961,
+      "step": 16927
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.553622857142858e-06,
+      "loss": 1.2906,
+      "step": 16928
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.553354285714286e-06,
+      "loss": 1.8584,
+      "step": 16929
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.553085714285714e-06,
+      "loss": 1.2936,
+      "step": 16930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.552817142857144e-06,
+      "loss": 1.5731,
+      "step": 16931
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.55254857142857e-06,
+      "loss": 1.352,
+      "step": 16932
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.55228e-06,
+      "loss": 1.0162,
+      "step": 16933
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.55201142857143e-06,
+      "loss": 0.9414,
+      "step": 16934
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.551742857142857e-06,
+      "loss": 0.8874,
+      "step": 16935
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.551474285714287e-06,
+      "loss": 1.2459,
+      "step": 16936
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.551205714285715e-06,
+      "loss": 1.0283,
+      "step": 16937
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.550937142857145e-06,
+      "loss": 1.1026,
+      "step": 16938
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.550668571428571e-06,
+      "loss": 1.2846,
+      "step": 16939
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.550400000000001e-06,
+      "loss": 0.8234,
+      "step": 16940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.550131428571429e-06,
+      "loss": 1.7364,
+      "step": 16941
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.549862857142857e-06,
+      "loss": 1.2775,
+      "step": 16942
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.549594285714285e-06,
+      "loss": 1.2881,
+      "step": 16943
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.549325714285715e-06,
+      "loss": 1.3989,
+      "step": 16944
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.549057142857142e-06,
+      "loss": 0.7883,
+      "step": 16945
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.548788571428572e-06,
+      "loss": 1.2399,
+      "step": 16946
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.548520000000002e-06,
+      "loss": 1.2795,
+      "step": 16947
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.548251428571428e-06,
+      "loss": 0.8947,
+      "step": 16948
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.547982857142858e-06,
+      "loss": 1.438,
+      "step": 16949
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.547714285714286e-06,
+      "loss": 1.3738,
+      "step": 16950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.547445714285716e-06,
+      "loss": 1.4965,
+      "step": 16951
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.547177142857142e-06,
+      "loss": 1.0425,
+      "step": 16952
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.546908571428572e-06,
+      "loss": 1.1495,
+      "step": 16953
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.54664e-06,
+      "loss": 1.4471,
+      "step": 16954
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.546371428571429e-06,
+      "loss": 1.0088,
+      "step": 16955
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.546102857142857e-06,
+      "loss": 1.1892,
+      "step": 16956
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.545834285714287e-06,
+      "loss": 1.6038,
+      "step": 16957
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.545565714285713e-06,
+      "loss": 0.9957,
+      "step": 16958
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.545297142857143e-06,
+      "loss": 0.9726,
+      "step": 16959
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.545028571428571e-06,
+      "loss": 0.9602,
+      "step": 16960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.544760000000001e-06,
+      "loss": 1.453,
+      "step": 16961
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.54449142857143e-06,
+      "loss": 1.6943,
+      "step": 16962
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.544222857142857e-06,
+      "loss": 1.549,
+      "step": 16963
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.543954285714287e-06,
+      "loss": 1.2287,
+      "step": 16964
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.543685714285714e-06,
+      "loss": 1.2986,
+      "step": 16965
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.543417142857144e-06,
+      "loss": 1.3679,
+      "step": 16966
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.543148571428572e-06,
+      "loss": 0.7703,
+      "step": 16967
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.54288e-06,
+      "loss": 1.3483,
+      "step": 16968
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.542611428571428e-06,
+      "loss": 1.2081,
+      "step": 16969
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.542342857142858e-06,
+      "loss": 0.8528,
+      "step": 16970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.542074285714285e-06,
+      "loss": 1.36,
+      "step": 16971
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.541805714285714e-06,
+      "loss": 0.6737,
+      "step": 16972
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.541537142857143e-06,
+      "loss": 1.4819,
+      "step": 16973
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.541268571428572e-06,
+      "loss": 1.1205,
+      "step": 16974
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.541e-06,
+      "loss": 1.1991,
+      "step": 16975
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.540731428571429e-06,
+      "loss": 1.2902,
+      "step": 16976
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.540462857142859e-06,
+      "loss": 1.3058,
+      "step": 16977
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.540194285714285e-06,
+      "loss": 1.2645,
+      "step": 16978
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.539925714285715e-06,
+      "loss": 1.5667,
+      "step": 16979
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.539657142857143e-06,
+      "loss": 1.2557,
+      "step": 16980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.539388571428571e-06,
+      "loss": 1.0773,
+      "step": 16981
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.53912e-06,
+      "loss": 1.4615,
+      "step": 16982
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.53885142857143e-06,
+      "loss": 1.351,
+      "step": 16983
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.538582857142858e-06,
+      "loss": 1.3397,
+      "step": 16984
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.538314285714286e-06,
+      "loss": 1.5699,
+      "step": 16985
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.538045714285714e-06,
+      "loss": 1.3453,
+      "step": 16986
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.537777142857144e-06,
+      "loss": 1.4661,
+      "step": 16987
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.537508571428572e-06,
+      "loss": 1.0126,
+      "step": 16988
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.53724e-06,
+      "loss": 1.2139,
+      "step": 16989
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.53697142857143e-06,
+      "loss": 1.2686,
+      "step": 16990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.536702857142857e-06,
+      "loss": 1.2717,
+      "step": 16991
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.536434285714286e-06,
+      "loss": 1.1259,
+      "step": 16992
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.536165714285715e-06,
+      "loss": 1.6942,
+      "step": 16993
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.535897142857143e-06,
+      "loss": 1.2862,
+      "step": 16994
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.535628571428571e-06,
+      "loss": 1.5303,
+      "step": 16995
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.53536e-06,
+      "loss": 1.0234,
+      "step": 16996
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.535091428571429e-06,
+      "loss": 1.0742,
+      "step": 16997
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.534822857142857e-06,
+      "loss": 0.9646,
+      "step": 16998
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.534554285714285e-06,
+      "loss": 1.0926,
+      "step": 16999
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.534285714285715e-06,
+      "loss": 1.0498,
+      "step": 17000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.534017142857143e-06,
+      "loss": 1.0081,
+      "step": 17001
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.533748571428572e-06,
+      "loss": 1.1099,
+      "step": 17002
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.533480000000001e-06,
+      "loss": 1.396,
+      "step": 17003
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.533211428571428e-06,
+      "loss": 1.1424,
+      "step": 17004
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.532942857142858e-06,
+      "loss": 1.0024,
+      "step": 17005
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.532674285714286e-06,
+      "loss": 0.8794,
+      "step": 17006
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.532405714285716e-06,
+      "loss": 1.2822,
+      "step": 17007
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.532137142857142e-06,
+      "loss": 0.9507,
+      "step": 17008
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.531868571428572e-06,
+      "loss": 1.0585,
+      "step": 17009
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.5316e-06,
+      "loss": 0.6151,
+      "step": 17010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.531331428571429e-06,
+      "loss": 1.3505,
+      "step": 17011
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.531062857142857e-06,
+      "loss": 1.4312,
+      "step": 17012
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.530794285714287e-06,
+      "loss": 1.3517,
+      "step": 17013
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.530525714285715e-06,
+      "loss": 1.4987,
+      "step": 17014
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.530257142857143e-06,
+      "loss": 1.2217,
+      "step": 17015
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.529988571428573e-06,
+      "loss": 1.211,
+      "step": 17016
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52972e-06,
+      "loss": 1.1172,
+      "step": 17017
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52945142857143e-06,
+      "loss": 1.2269,
+      "step": 17018
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.529182857142857e-06,
+      "loss": 1.5724,
+      "step": 17019
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.528914285714287e-06,
+      "loss": 1.366,
+      "step": 17020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.528645714285714e-06,
+      "loss": 1.0323,
+      "step": 17021
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.528377142857144e-06,
+      "loss": 1.2883,
+      "step": 17022
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.528108571428572e-06,
+      "loss": 0.994,
+      "step": 17023
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52784e-06,
+      "loss": 0.9402,
+      "step": 17024
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.527571428571428e-06,
+      "loss": 1.1092,
+      "step": 17025
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.527302857142858e-06,
+      "loss": 1.4044,
+      "step": 17026
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.527034285714286e-06,
+      "loss": 1.2737,
+      "step": 17027
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.526765714285714e-06,
+      "loss": 1.5403,
+      "step": 17028
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.526497142857144e-06,
+      "loss": 1.274,
+      "step": 17029
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52622857142857e-06,
+      "loss": 1.3773,
+      "step": 17030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52596e-06,
+      "loss": 1.4535,
+      "step": 17031
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.525691428571429e-06,
+      "loss": 1.0558,
+      "step": 17032
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.525422857142859e-06,
+      "loss": 1.0877,
+      "step": 17033
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.525154285714285e-06,
+      "loss": 1.529,
+      "step": 17034
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.524885714285715e-06,
+      "loss": 0.9276,
+      "step": 17035
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.524617142857143e-06,
+      "loss": 1.1985,
+      "step": 17036
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.524348571428571e-06,
+      "loss": 1.2586,
+      "step": 17037
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52408e-06,
+      "loss": 1.2084,
+      "step": 17038
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52381142857143e-06,
+      "loss": 1.429,
+      "step": 17039
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.523542857142857e-06,
+      "loss": 1.0882,
+      "step": 17040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.523274285714286e-06,
+      "loss": 1.3637,
+      "step": 17041
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.523005714285716e-06,
+      "loss": 1.4262,
+      "step": 17042
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.522737142857144e-06,
+      "loss": 1.2982,
+      "step": 17043
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.522468571428572e-06,
+      "loss": 1.4828,
+      "step": 17044
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.5222e-06,
+      "loss": 1.1851,
+      "step": 17045
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52193142857143e-06,
+      "loss": 1.0745,
+      "step": 17046
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.521662857142856e-06,
+      "loss": 0.9776,
+      "step": 17047
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.521394285714286e-06,
+      "loss": 1.1547,
+      "step": 17048
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.521125714285714e-06,
+      "loss": 1.3917,
+      "step": 17049
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.520857142857143e-06,
+      "loss": 0.9946,
+      "step": 17050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52058857142857e-06,
+      "loss": 1.1168,
+      "step": 17051
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.52032e-06,
+      "loss": 1.0796,
+      "step": 17052
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.520051428571429e-06,
+      "loss": 1.236,
+      "step": 17053
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.519782857142857e-06,
+      "loss": 1.0544,
+      "step": 17054
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.519514285714287e-06,
+      "loss": 1.5977,
+      "step": 17055
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.519245714285715e-06,
+      "loss": 0.9359,
+      "step": 17056
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.518977142857143e-06,
+      "loss": 1.1521,
+      "step": 17057
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.518708571428571e-06,
+      "loss": 1.3187,
+      "step": 17058
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.518440000000001e-06,
+      "loss": 1.2068,
+      "step": 17059
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.518171428571428e-06,
+      "loss": 1.2732,
+      "step": 17060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.517902857142858e-06,
+      "loss": 1.1732,
+      "step": 17061
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.517634285714286e-06,
+      "loss": 1.4362,
+      "step": 17062
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.517365714285714e-06,
+      "loss": 0.8221,
+      "step": 17063
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.517097142857142e-06,
+      "loss": 1.08,
+      "step": 17064
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.516828571428572e-06,
+      "loss": 1.1693,
+      "step": 17065
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.51656e-06,
+      "loss": 1.4051,
+      "step": 17066
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.516291428571428e-06,
+      "loss": 1.1589,
+      "step": 17067
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.516022857142858e-06,
+      "loss": 1.2084,
+      "step": 17068
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.515754285714286e-06,
+      "loss": 1.4998,
+      "step": 17069
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.515485714285715e-06,
+      "loss": 1.5027,
+      "step": 17070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.515217142857143e-06,
+      "loss": 1.0361,
+      "step": 17071
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.514948571428573e-06,
+      "loss": 1.6266,
+      "step": 17072
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.51468e-06,
+      "loss": 1.0849,
+      "step": 17073
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.514411428571429e-06,
+      "loss": 1.0775,
+      "step": 17074
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.514142857142857e-06,
+      "loss": 0.9856,
+      "step": 17075
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.513874285714285e-06,
+      "loss": 1.291,
+      "step": 17076
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.513605714285714e-06,
+      "loss": 1.141,
+      "step": 17077
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.513337142857143e-06,
+      "loss": 1.2936,
+      "step": 17078
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.513068571428572e-06,
+      "loss": 1.1599,
+      "step": 17079
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.5128e-06,
+      "loss": 1.1832,
+      "step": 17080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.51253142857143e-06,
+      "loss": 1.2956,
+      "step": 17081
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.512262857142858e-06,
+      "loss": 1.2136,
+      "step": 17082
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.511994285714286e-06,
+      "loss": 1.0994,
+      "step": 17083
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.511725714285714e-06,
+      "loss": 1.378,
+      "step": 17084
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.511457142857144e-06,
+      "loss": 1.0875,
+      "step": 17085
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.51118857142857e-06,
+      "loss": 1.2859,
+      "step": 17086
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.51092e-06,
+      "loss": 1.2916,
+      "step": 17087
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.510651428571429e-06,
+      "loss": 1.3615,
+      "step": 17088
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.510382857142858e-06,
+      "loss": 0.6638,
+      "step": 17089
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.510114285714285e-06,
+      "loss": 1.2286,
+      "step": 17090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.509845714285715e-06,
+      "loss": 1.6054,
+      "step": 17091
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.509577142857143e-06,
+      "loss": 1.4368,
+      "step": 17092
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.509308571428571e-06,
+      "loss": 1.2222,
+      "step": 17093
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.509040000000001e-06,
+      "loss": 1.08,
+      "step": 17094
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.50877142857143e-06,
+      "loss": 0.9415,
+      "step": 17095
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.508502857142857e-06,
+      "loss": 1.3823,
+      "step": 17096
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.508234285714286e-06,
+      "loss": 1.4567,
+      "step": 17097
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.507965714285715e-06,
+      "loss": 1.3771,
+      "step": 17098
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.507697142857142e-06,
+      "loss": 1.2248,
+      "step": 17099
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.507428571428572e-06,
+      "loss": 1.0011,
+      "step": 17100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.50716e-06,
+      "loss": 1.2921,
+      "step": 17101
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.50689142857143e-06,
+      "loss": 1.2137,
+      "step": 17102
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.506622857142856e-06,
+      "loss": 0.9903,
+      "step": 17103
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.506354285714286e-06,
+      "loss": 0.9267,
+      "step": 17104
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.506085714285714e-06,
+      "loss": 1.5814,
+      "step": 17105
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.505817142857143e-06,
+      "loss": 0.9232,
+      "step": 17106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.505548571428572e-06,
+      "loss": 1.4051,
+      "step": 17107
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.50528e-06,
+      "loss": 1.0855,
+      "step": 17108
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.505011428571429e-06,
+      "loss": 1.1595,
+      "step": 17109
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.504742857142857e-06,
+      "loss": 1.0476,
+      "step": 17110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.504474285714287e-06,
+      "loss": 0.8932,
+      "step": 17111
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.504205714285715e-06,
+      "loss": 1.145,
+      "step": 17112
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.503937142857143e-06,
+      "loss": 1.2066,
+      "step": 17113
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.503668571428571e-06,
+      "loss": 1.1205,
+      "step": 17114
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.503400000000001e-06,
+      "loss": 0.6459,
+      "step": 17115
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.503131428571428e-06,
+      "loss": 1.2251,
+      "step": 17116
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.502862857142858e-06,
+      "loss": 0.7715,
+      "step": 17117
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.502594285714286e-06,
+      "loss": 1.2762,
+      "step": 17118
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.502325714285714e-06,
+      "loss": 0.7718,
+      "step": 17119
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.502057142857144e-06,
+      "loss": 1.0468,
+      "step": 17120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.501788571428572e-06,
+      "loss": 1.2046,
+      "step": 17121
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.50152e-06,
+      "loss": 1.5222,
+      "step": 17122
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.501251428571428e-06,
+      "loss": 0.8369,
+      "step": 17123
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.500982857142858e-06,
+      "loss": 1.2332,
+      "step": 17124
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.500714285714286e-06,
+      "loss": 1.175,
+      "step": 17125
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.500445714285715e-06,
+      "loss": 1.6185,
+      "step": 17126
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.500177142857143e-06,
+      "loss": 1.3143,
+      "step": 17127
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.499908571428573e-06,
+      "loss": 0.9507,
+      "step": 17128
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.499639999999999e-06,
+      "loss": 1.1872,
+      "step": 17129
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.499371428571429e-06,
+      "loss": 1.1994,
+      "step": 17130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.499102857142857e-06,
+      "loss": 1.3473,
+      "step": 17131
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.498834285714285e-06,
+      "loss": 1.2007,
+      "step": 17132
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.498565714285715e-06,
+      "loss": 1.4257,
+      "step": 17133
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.498297142857143e-06,
+      "loss": 1.5223,
+      "step": 17134
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.498028571428573e-06,
+      "loss": 1.233,
+      "step": 17135
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49776e-06,
+      "loss": 1.2192,
+      "step": 17136
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49749142857143e-06,
+      "loss": 1.2057,
+      "step": 17137
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.497222857142858e-06,
+      "loss": 0.8581,
+      "step": 17138
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.496954285714286e-06,
+      "loss": 1.0129,
+      "step": 17139
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.496685714285714e-06,
+      "loss": 1.3438,
+      "step": 17140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.496417142857144e-06,
+      "loss": 1.298,
+      "step": 17141
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49614857142857e-06,
+      "loss": 1.3319,
+      "step": 17142
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49588e-06,
+      "loss": 1.3939,
+      "step": 17143
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.495611428571428e-06,
+      "loss": 1.2851,
+      "step": 17144
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.495342857142857e-06,
+      "loss": 1.0912,
+      "step": 17145
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.495074285714287e-06,
+      "loss": 1.5111,
+      "step": 17146
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.494805714285715e-06,
+      "loss": 0.8405,
+      "step": 17147
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.494537142857145e-06,
+      "loss": 1.4622,
+      "step": 17148
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.494268571428571e-06,
+      "loss": 1.2324,
+      "step": 17149
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.494000000000001e-06,
+      "loss": 1.2912,
+      "step": 17150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.493731428571429e-06,
+      "loss": 1.3735,
+      "step": 17151
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.493462857142857e-06,
+      "loss": 0.7399,
+      "step": 17152
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.493194285714285e-06,
+      "loss": 0.9366,
+      "step": 17153
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.492925714285715e-06,
+      "loss": 1.4939,
+      "step": 17154
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.492657142857142e-06,
+      "loss": 1.1871,
+      "step": 17155
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.492388571428572e-06,
+      "loss": 1.4961,
+      "step": 17156
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49212e-06,
+      "loss": 1.0317,
+      "step": 17157
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49185142857143e-06,
+      "loss": 1.2521,
+      "step": 17158
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.491582857142858e-06,
+      "loss": 1.5855,
+      "step": 17159
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.491314285714286e-06,
+      "loss": 0.8258,
+      "step": 17160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.491045714285716e-06,
+      "loss": 1.1406,
+      "step": 17161
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.490777142857142e-06,
+      "loss": 1.228,
+      "step": 17162
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.490508571428572e-06,
+      "loss": 1.1802,
+      "step": 17163
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49024e-06,
+      "loss": 1.2353,
+      "step": 17164
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.489971428571429e-06,
+      "loss": 1.1925,
+      "step": 17165
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.489702857142857e-06,
+      "loss": 0.8927,
+      "step": 17166
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.489434285714287e-06,
+      "loss": 1.0204,
+      "step": 17167
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.489165714285713e-06,
+      "loss": 1.0182,
+      "step": 17168
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.488897142857143e-06,
+      "loss": 1.3815,
+      "step": 17169
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.488628571428571e-06,
+      "loss": 0.9218,
+      "step": 17170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.488360000000001e-06,
+      "loss": 0.8279,
+      "step": 17171
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.48809142857143e-06,
+      "loss": 0.6761,
+      "step": 17172
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.487822857142857e-06,
+      "loss": 1.4908,
+      "step": 17173
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.487554285714287e-06,
+      "loss": 1.3065,
+      "step": 17174
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.487285714285714e-06,
+      "loss": 1.2523,
+      "step": 17175
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.487017142857144e-06,
+      "loss": 1.5422,
+      "step": 17176
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.486748571428572e-06,
+      "loss": 1.0351,
+      "step": 17177
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.48648e-06,
+      "loss": 1.6124,
+      "step": 17178
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.486211428571428e-06,
+      "loss": 1.1694,
+      "step": 17179
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.485942857142858e-06,
+      "loss": 1.0493,
+      "step": 17180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.485674285714286e-06,
+      "loss": 0.8894,
+      "step": 17181
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.485405714285714e-06,
+      "loss": 1.4282,
+      "step": 17182
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.485137142857143e-06,
+      "loss": 1.2105,
+      "step": 17183
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.484868571428572e-06,
+      "loss": 1.1467,
+      "step": 17184
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.4846e-06,
+      "loss": 1.2686,
+      "step": 17185
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.484331428571429e-06,
+      "loss": 1.7397,
+      "step": 17186
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.484062857142859e-06,
+      "loss": 1.2732,
+      "step": 17187
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.483794285714285e-06,
+      "loss": 1.6139,
+      "step": 17188
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.483525714285715e-06,
+      "loss": 1.6129,
+      "step": 17189
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.483257142857143e-06,
+      "loss": 1.2633,
+      "step": 17190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.482988571428571e-06,
+      "loss": 1.5097,
+      "step": 17191
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.48272e-06,
+      "loss": 1.1352,
+      "step": 17192
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.48245142857143e-06,
+      "loss": 1.3473,
+      "step": 17193
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.482182857142858e-06,
+      "loss": 1.0765,
+      "step": 17194
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.481914285714286e-06,
+      "loss": 1.1961,
+      "step": 17195
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.481645714285714e-06,
+      "loss": 1.0794,
+      "step": 17196
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.481377142857144e-06,
+      "loss": 1.1309,
+      "step": 17197
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.481108571428572e-06,
+      "loss": 1.3308,
+      "step": 17198
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.48084e-06,
+      "loss": 1.4879,
+      "step": 17199
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.48057142857143e-06,
+      "loss": 1.3224,
+      "step": 17200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.480302857142857e-06,
+      "loss": 0.8323,
+      "step": 17201
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.480034285714286e-06,
+      "loss": 1.0531,
+      "step": 17202
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.479765714285715e-06,
+      "loss": 1.1779,
+      "step": 17203
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.479497142857143e-06,
+      "loss": 0.7687,
+      "step": 17204
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.479228571428571e-06,
+      "loss": 1.0847,
+      "step": 17205
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.47896e-06,
+      "loss": 1.067,
+      "step": 17206
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.478691428571429e-06,
+      "loss": 0.9711,
+      "step": 17207
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.478422857142857e-06,
+      "loss": 1.3082,
+      "step": 17208
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.478154285714285e-06,
+      "loss": 1.2073,
+      "step": 17209
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.477885714285715e-06,
+      "loss": 1.1257,
+      "step": 17210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.477617142857143e-06,
+      "loss": 1.619,
+      "step": 17211
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.477348571428572e-06,
+      "loss": 1.4744,
+      "step": 17212
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.477080000000001e-06,
+      "loss": 1.1456,
+      "step": 17213
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.476811428571428e-06,
+      "loss": 1.398,
+      "step": 17214
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.476542857142858e-06,
+      "loss": 1.2294,
+      "step": 17215
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.476274285714286e-06,
+      "loss": 1.3139,
+      "step": 17216
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.476005714285716e-06,
+      "loss": 1.0667,
+      "step": 17217
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.475737142857142e-06,
+      "loss": 1.666,
+      "step": 17218
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.475468571428572e-06,
+      "loss": 1.1146,
+      "step": 17219
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.4752e-06,
+      "loss": 1.2229,
+      "step": 17220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.474931428571429e-06,
+      "loss": 1.2985,
+      "step": 17221
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.474662857142857e-06,
+      "loss": 1.3506,
+      "step": 17222
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.474394285714287e-06,
+      "loss": 1.6074,
+      "step": 17223
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.474125714285715e-06,
+      "loss": 1.33,
+      "step": 17224
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.473857142857143e-06,
+      "loss": 0.8408,
+      "step": 17225
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.473588571428573e-06,
+      "loss": 0.9631,
+      "step": 17226
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.47332e-06,
+      "loss": 1.0074,
+      "step": 17227
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.473051428571429e-06,
+      "loss": 0.9649,
+      "step": 17228
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.472782857142857e-06,
+      "loss": 1.0432,
+      "step": 17229
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.472514285714287e-06,
+      "loss": 1.3004,
+      "step": 17230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.472245714285714e-06,
+      "loss": 1.2392,
+      "step": 17231
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.471977142857144e-06,
+      "loss": 1.3248,
+      "step": 17232
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.471708571428572e-06,
+      "loss": 1.0641,
+      "step": 17233
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.47144e-06,
+      "loss": 0.9136,
+      "step": 17234
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.471171428571428e-06,
+      "loss": 1.7547,
+      "step": 17235
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.470902857142858e-06,
+      "loss": 1.2841,
+      "step": 17236
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.470634285714286e-06,
+      "loss": 1.1643,
+      "step": 17237
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.470365714285714e-06,
+      "loss": 1.4378,
+      "step": 17238
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.470097142857144e-06,
+      "loss": 1.2923,
+      "step": 17239
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.469828571428572e-06,
+      "loss": 1.1142,
+      "step": 17240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46956e-06,
+      "loss": 1.1069,
+      "step": 17241
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.469291428571429e-06,
+      "loss": 1.2859,
+      "step": 17242
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.469022857142859e-06,
+      "loss": 1.18,
+      "step": 17243
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.468754285714285e-06,
+      "loss": 1.3092,
+      "step": 17244
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.468485714285715e-06,
+      "loss": 1.3757,
+      "step": 17245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.468217142857143e-06,
+      "loss": 0.786,
+      "step": 17246
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.467948571428571e-06,
+      "loss": 1.0371,
+      "step": 17247
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46768e-06,
+      "loss": 0.9105,
+      "step": 17248
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46741142857143e-06,
+      "loss": 0.931,
+      "step": 17249
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.467142857142857e-06,
+      "loss": 1.4706,
+      "step": 17250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.466874285714286e-06,
+      "loss": 1.1118,
+      "step": 17251
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.466605714285716e-06,
+      "loss": 1.3022,
+      "step": 17252
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.466337142857144e-06,
+      "loss": 1.0991,
+      "step": 17253
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.466068571428572e-06,
+      "loss": 1.2839,
+      "step": 17254
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.4658e-06,
+      "loss": 0.8232,
+      "step": 17255
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46553142857143e-06,
+      "loss": 1.0123,
+      "step": 17256
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.465262857142856e-06,
+      "loss": 1.0863,
+      "step": 17257
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.464994285714286e-06,
+      "loss": 1.2025,
+      "step": 17258
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.464725714285714e-06,
+      "loss": 1.2567,
+      "step": 17259
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.464457142857143e-06,
+      "loss": 1.1706,
+      "step": 17260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46418857142857e-06,
+      "loss": 1.3014,
+      "step": 17261
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46392e-06,
+      "loss": 1.3386,
+      "step": 17262
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.463651428571429e-06,
+      "loss": 0.9808,
+      "step": 17263
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.463382857142857e-06,
+      "loss": 1.0846,
+      "step": 17264
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.463114285714287e-06,
+      "loss": 1.0672,
+      "step": 17265
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.462845714285715e-06,
+      "loss": 1.2627,
+      "step": 17266
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.462577142857143e-06,
+      "loss": 1.3392,
+      "step": 17267
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.462308571428571e-06,
+      "loss": 1.4357,
+      "step": 17268
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.462040000000001e-06,
+      "loss": 1.3896,
+      "step": 17269
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.461771428571428e-06,
+      "loss": 0.839,
+      "step": 17270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.461502857142858e-06,
+      "loss": 0.9945,
+      "step": 17271
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.461234285714286e-06,
+      "loss": 0.9856,
+      "step": 17272
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.460965714285714e-06,
+      "loss": 1.2092,
+      "step": 17273
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.460697142857142e-06,
+      "loss": 0.8436,
+      "step": 17274
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.460428571428572e-06,
+      "loss": 1.2255,
+      "step": 17275
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46016e-06,
+      "loss": 1.1771,
+      "step": 17276
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.459891428571428e-06,
+      "loss": 1.1746,
+      "step": 17277
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.459622857142858e-06,
+      "loss": 1.4458,
+      "step": 17278
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.459354285714286e-06,
+      "loss": 1.313,
+      "step": 17279
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.459085714285715e-06,
+      "loss": 0.7746,
+      "step": 17280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.458817142857143e-06,
+      "loss": 1.2344,
+      "step": 17281
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.458548571428573e-06,
+      "loss": 1.3499,
+      "step": 17282
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.45828e-06,
+      "loss": 1.6098,
+      "step": 17283
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.458011428571429e-06,
+      "loss": 1.219,
+      "step": 17284
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.457742857142857e-06,
+      "loss": 1.3985,
+      "step": 17285
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.457474285714287e-06,
+      "loss": 1.3734,
+      "step": 17286
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.457205714285714e-06,
+      "loss": 1.0424,
+      "step": 17287
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.456937142857143e-06,
+      "loss": 0.8961,
+      "step": 17288
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.456668571428572e-06,
+      "loss": 1.0918,
+      "step": 17289
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.4564e-06,
+      "loss": 1.205,
+      "step": 17290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.45613142857143e-06,
+      "loss": 1.2644,
+      "step": 17291
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.455862857142858e-06,
+      "loss": 1.1425,
+      "step": 17292
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.455594285714286e-06,
+      "loss": 0.9923,
+      "step": 17293
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.455325714285714e-06,
+      "loss": 0.9836,
+      "step": 17294
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.455057142857144e-06,
+      "loss": 1.1331,
+      "step": 17295
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.45478857142857e-06,
+      "loss": 1.4243,
+      "step": 17296
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.45452e-06,
+      "loss": 1.2075,
+      "step": 17297
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.454251428571429e-06,
+      "loss": 1.3308,
+      "step": 17298
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.453982857142858e-06,
+      "loss": 1.5344,
+      "step": 17299
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.453714285714285e-06,
+      "loss": 0.9285,
+      "step": 17300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.453445714285715e-06,
+      "loss": 1.1551,
+      "step": 17301
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.453177142857143e-06,
+      "loss": 1.2092,
+      "step": 17302
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.452908571428571e-06,
+      "loss": 0.9597,
+      "step": 17303
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.452640000000001e-06,
+      "loss": 0.8438,
+      "step": 17304
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.45237142857143e-06,
+      "loss": 1.2654,
+      "step": 17305
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.452102857142857e-06,
+      "loss": 0.9945,
+      "step": 17306
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.451834285714286e-06,
+      "loss": 1.1756,
+      "step": 17307
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.451565714285715e-06,
+      "loss": 1.0017,
+      "step": 17308
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.451297142857144e-06,
+      "loss": 1.4333,
+      "step": 17309
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.451028571428572e-06,
+      "loss": 1.1274,
+      "step": 17310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.45076e-06,
+      "loss": 1.2723,
+      "step": 17311
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.45049142857143e-06,
+      "loss": 1.1658,
+      "step": 17312
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.450222857142856e-06,
+      "loss": 1.2276,
+      "step": 17313
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.449954285714286e-06,
+      "loss": 1.7046,
+      "step": 17314
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.449685714285714e-06,
+      "loss": 1.1468,
+      "step": 17315
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.449417142857143e-06,
+      "loss": 1.0957,
+      "step": 17316
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.449148571428572e-06,
+      "loss": 1.0272,
+      "step": 17317
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.44888e-06,
+      "loss": 1.149,
+      "step": 17318
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.448611428571429e-06,
+      "loss": 1.5455,
+      "step": 17319
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.448342857142857e-06,
+      "loss": 1.0466,
+      "step": 17320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.448074285714287e-06,
+      "loss": 0.589,
+      "step": 17321
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.447805714285715e-06,
+      "loss": 1.1241,
+      "step": 17322
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.447537142857143e-06,
+      "loss": 0.9163,
+      "step": 17323
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.447268571428571e-06,
+      "loss": 1.3943,
+      "step": 17324
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.447000000000001e-06,
+      "loss": 1.082,
+      "step": 17325
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.446731428571428e-06,
+      "loss": 1.0077,
+      "step": 17326
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.446462857142858e-06,
+      "loss": 1.1658,
+      "step": 17327
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.446194285714286e-06,
+      "loss": 1.1612,
+      "step": 17328
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.445925714285714e-06,
+      "loss": 1.0509,
+      "step": 17329
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.445657142857144e-06,
+      "loss": 1.1069,
+      "step": 17330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.445388571428572e-06,
+      "loss": 1.4135,
+      "step": 17331
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.445120000000002e-06,
+      "loss": 1.1931,
+      "step": 17332
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.444851428571428e-06,
+      "loss": 1.4763,
+      "step": 17333
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.444582857142858e-06,
+      "loss": 1.512,
+      "step": 17334
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.444314285714286e-06,
+      "loss": 1.3403,
+      "step": 17335
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.444045714285715e-06,
+      "loss": 1.5975,
+      "step": 17336
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.443777142857143e-06,
+      "loss": 1.1943,
+      "step": 17337
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.443508571428573e-06,
+      "loss": 1.2047,
+      "step": 17338
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.443239999999999e-06,
+      "loss": 1.23,
+      "step": 17339
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.442971428571429e-06,
+      "loss": 1.364,
+      "step": 17340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.442702857142857e-06,
+      "loss": 1.4151,
+      "step": 17341
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.442434285714285e-06,
+      "loss": 1.4933,
+      "step": 17342
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.442165714285715e-06,
+      "loss": 1.4418,
+      "step": 17343
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.441897142857143e-06,
+      "loss": 1.4557,
+      "step": 17344
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.441628571428573e-06,
+      "loss": 1.1419,
+      "step": 17345
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.44136e-06,
+      "loss": 1.1515,
+      "step": 17346
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.44109142857143e-06,
+      "loss": 1.608,
+      "step": 17347
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.440822857142858e-06,
+      "loss": 1.3886,
+      "step": 17348
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.440554285714286e-06,
+      "loss": 1.3279,
+      "step": 17349
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.440285714285714e-06,
+      "loss": 1.2068,
+      "step": 17350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.440017142857144e-06,
+      "loss": 0.6966,
+      "step": 17351
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.43974857142857e-06,
+      "loss": 1.0966,
+      "step": 17352
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.43948e-06,
+      "loss": 1.2977,
+      "step": 17353
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.439211428571428e-06,
+      "loss": 1.3938,
+      "step": 17354
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.438942857142858e-06,
+      "loss": 1.2959,
+      "step": 17355
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.438674285714287e-06,
+      "loss": 1.4137,
+      "step": 17356
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.438405714285715e-06,
+      "loss": 0.8456,
+      "step": 17357
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.438137142857145e-06,
+      "loss": 0.7078,
+      "step": 17358
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.437868571428571e-06,
+      "loss": 0.9339,
+      "step": 17359
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.437600000000001e-06,
+      "loss": 1.398,
+      "step": 17360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.437331428571429e-06,
+      "loss": 0.9551,
+      "step": 17361
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.437062857142857e-06,
+      "loss": 0.9824,
+      "step": 17362
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.436794285714285e-06,
+      "loss": 1.3212,
+      "step": 17363
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.436525714285715e-06,
+      "loss": 1.1644,
+      "step": 17364
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.436257142857142e-06,
+      "loss": 1.0218,
+      "step": 17365
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.435988571428572e-06,
+      "loss": 1.2768,
+      "step": 17366
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.43572e-06,
+      "loss": 1.3096,
+      "step": 17367
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.43545142857143e-06,
+      "loss": 1.2441,
+      "step": 17368
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.435182857142858e-06,
+      "loss": 1.5032,
+      "step": 17369
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.434914285714286e-06,
+      "loss": 1.2463,
+      "step": 17370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.434645714285716e-06,
+      "loss": 1.2908,
+      "step": 17371
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.434377142857142e-06,
+      "loss": 1.196,
+      "step": 17372
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.434108571428572e-06,
+      "loss": 1.2379,
+      "step": 17373
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.43384e-06,
+      "loss": 1.7583,
+      "step": 17374
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.433571428571429e-06,
+      "loss": 0.8637,
+      "step": 17375
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.433302857142857e-06,
+      "loss": 1.4545,
+      "step": 17376
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.433034285714287e-06,
+      "loss": 1.3464,
+      "step": 17377
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.432765714285713e-06,
+      "loss": 1.5212,
+      "step": 17378
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.432497142857143e-06,
+      "loss": 1.2194,
+      "step": 17379
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.432228571428571e-06,
+      "loss": 1.2931,
+      "step": 17380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.431960000000001e-06,
+      "loss": 1.2215,
+      "step": 17381
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.43169142857143e-06,
+      "loss": 1.5117,
+      "step": 17382
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.431422857142857e-06,
+      "loss": 1.0208,
+      "step": 17383
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.431154285714287e-06,
+      "loss": 1.0568,
+      "step": 17384
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.430885714285714e-06,
+      "loss": 1.1374,
+      "step": 17385
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.430617142857144e-06,
+      "loss": 1.0923,
+      "step": 17386
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.430348571428572e-06,
+      "loss": 1.542,
+      "step": 17387
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.43008e-06,
+      "loss": 1.1395,
+      "step": 17388
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.429811428571428e-06,
+      "loss": 0.9106,
+      "step": 17389
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.429542857142858e-06,
+      "loss": 1.0534,
+      "step": 17390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.429274285714286e-06,
+      "loss": 1.3152,
+      "step": 17391
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.429005714285714e-06,
+      "loss": 1.4997,
+      "step": 17392
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.428737142857143e-06,
+      "loss": 0.8731,
+      "step": 17393
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.428468571428572e-06,
+      "loss": 1.4061,
+      "step": 17394
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.4282e-06,
+      "loss": 1.1501,
+      "step": 17395
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.427931428571429e-06,
+      "loss": 1.0586,
+      "step": 17396
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.427662857142859e-06,
+      "loss": 1.5352,
+      "step": 17397
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.427394285714285e-06,
+      "loss": 1.4053,
+      "step": 17398
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.427125714285715e-06,
+      "loss": 1.1883,
+      "step": 17399
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.426857142857143e-06,
+      "loss": 1.3226,
+      "step": 17400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.426588571428571e-06,
+      "loss": 1.4735,
+      "step": 17401
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.42632e-06,
+      "loss": 0.8545,
+      "step": 17402
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.42605142857143e-06,
+      "loss": 1.1141,
+      "step": 17403
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.425782857142858e-06,
+      "loss": 1.0446,
+      "step": 17404
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.425514285714286e-06,
+      "loss": 1.2941,
+      "step": 17405
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.425245714285714e-06,
+      "loss": 1.0023,
+      "step": 17406
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.424977142857144e-06,
+      "loss": 1.1708,
+      "step": 17407
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.424708571428572e-06,
+      "loss": 1.0929,
+      "step": 17408
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.42444e-06,
+      "loss": 1.0981,
+      "step": 17409
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.42417142857143e-06,
+      "loss": 1.6747,
+      "step": 17410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.423902857142857e-06,
+      "loss": 1.2748,
+      "step": 17411
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.423634285714286e-06,
+      "loss": 1.2104,
+      "step": 17412
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.423365714285715e-06,
+      "loss": 1.6291,
+      "step": 17413
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.423097142857144e-06,
+      "loss": 1.0997,
+      "step": 17414
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.422828571428571e-06,
+      "loss": 1.0955,
+      "step": 17415
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.42256e-06,
+      "loss": 1.2247,
+      "step": 17416
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.422291428571429e-06,
+      "loss": 1.1031,
+      "step": 17417
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.422022857142857e-06,
+      "loss": 1.1326,
+      "step": 17418
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.421754285714285e-06,
+      "loss": 0.9657,
+      "step": 17419
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.421485714285715e-06,
+      "loss": 1.4414,
+      "step": 17420
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.421217142857143e-06,
+      "loss": 1.1089,
+      "step": 17421
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.420948571428572e-06,
+      "loss": 1.2588,
+      "step": 17422
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.420680000000001e-06,
+      "loss": 1.0675,
+      "step": 17423
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.420411428571428e-06,
+      "loss": 1.189,
+      "step": 17424
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.420142857142858e-06,
+      "loss": 1.4615,
+      "step": 17425
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.419874285714286e-06,
+      "loss": 1.028,
+      "step": 17426
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.419605714285716e-06,
+      "loss": 0.8968,
+      "step": 17427
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.419337142857142e-06,
+      "loss": 1.2995,
+      "step": 17428
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.419068571428572e-06,
+      "loss": 1.2297,
+      "step": 17429
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.4188e-06,
+      "loss": 1.0146,
+      "step": 17430
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.418531428571429e-06,
+      "loss": 1.1851,
+      "step": 17431
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.418262857142857e-06,
+      "loss": 1.3659,
+      "step": 17432
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.417994285714287e-06,
+      "loss": 1.0129,
+      "step": 17433
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.417725714285715e-06,
+      "loss": 1.4114,
+      "step": 17434
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.417457142857143e-06,
+      "loss": 0.9441,
+      "step": 17435
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.417188571428573e-06,
+      "loss": 1.1065,
+      "step": 17436
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.416920000000001e-06,
+      "loss": 0.8459,
+      "step": 17437
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.416651428571429e-06,
+      "loss": 1.1614,
+      "step": 17438
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.416382857142857e-06,
+      "loss": 1.2195,
+      "step": 17439
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.416114285714287e-06,
+      "loss": 1.2981,
+      "step": 17440
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.415845714285714e-06,
+      "loss": 1.4955,
+      "step": 17441
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.415577142857144e-06,
+      "loss": 1.3397,
+      "step": 17442
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.415308571428572e-06,
+      "loss": 1.3878,
+      "step": 17443
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.41504e-06,
+      "loss": 1.2838,
+      "step": 17444
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.414771428571428e-06,
+      "loss": 1.1581,
+      "step": 17445
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.414502857142858e-06,
+      "loss": 0.9924,
+      "step": 17446
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.414234285714286e-06,
+      "loss": 1.2782,
+      "step": 17447
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.413965714285714e-06,
+      "loss": 1.3357,
+      "step": 17448
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.413697142857144e-06,
+      "loss": 1.1034,
+      "step": 17449
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.413428571428572e-06,
+      "loss": 0.9984,
+      "step": 17450
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.41316e-06,
+      "loss": 1.067,
+      "step": 17451
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.412891428571429e-06,
+      "loss": 1.4003,
+      "step": 17452
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.412622857142859e-06,
+      "loss": 1.1475,
+      "step": 17453
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.412354285714285e-06,
+      "loss": 1.1136,
+      "step": 17454
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.412085714285715e-06,
+      "loss": 1.0793,
+      "step": 17455
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.411817142857143e-06,
+      "loss": 1.2765,
+      "step": 17456
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.411548571428571e-06,
+      "loss": 0.8638,
+      "step": 17457
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.41128e-06,
+      "loss": 0.8572,
+      "step": 17458
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.41101142857143e-06,
+      "loss": 1.1985,
+      "step": 17459
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.410742857142857e-06,
+      "loss": 1.1673,
+      "step": 17460
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.410474285714286e-06,
+      "loss": 1.0503,
+      "step": 17461
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.410205714285716e-06,
+      "loss": 1.0397,
+      "step": 17462
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.409937142857144e-06,
+      "loss": 1.3466,
+      "step": 17463
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.409668571428572e-06,
+      "loss": 1.2805,
+      "step": 17464
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.4094e-06,
+      "loss": 1.6502,
+      "step": 17465
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.40913142857143e-06,
+      "loss": 1.0548,
+      "step": 17466
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.408862857142856e-06,
+      "loss": 1.1684,
+      "step": 17467
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.408594285714286e-06,
+      "loss": 0.9243,
+      "step": 17468
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.408325714285714e-06,
+      "loss": 1.183,
+      "step": 17469
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.408057142857143e-06,
+      "loss": 1.0301,
+      "step": 17470
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.40778857142857e-06,
+      "loss": 0.9446,
+      "step": 17471
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.40752e-06,
+      "loss": 1.1122,
+      "step": 17472
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.407251428571429e-06,
+      "loss": 0.8949,
+      "step": 17473
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.406982857142857e-06,
+      "loss": 1.1509,
+      "step": 17474
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.406714285714287e-06,
+      "loss": 1.0289,
+      "step": 17475
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.406445714285715e-06,
+      "loss": 0.9996,
+      "step": 17476
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.406177142857143e-06,
+      "loss": 0.8749,
+      "step": 17477
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.405908571428571e-06,
+      "loss": 1.3308,
+      "step": 17478
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.405640000000001e-06,
+      "loss": 0.9617,
+      "step": 17479
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.405371428571428e-06,
+      "loss": 1.2759,
+      "step": 17480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.405102857142858e-06,
+      "loss": 1.2709,
+      "step": 17481
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.404834285714286e-06,
+      "loss": 1.3035,
+      "step": 17482
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.404565714285716e-06,
+      "loss": 1.2986,
+      "step": 17483
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.404297142857142e-06,
+      "loss": 1.2381,
+      "step": 17484
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.404028571428572e-06,
+      "loss": 1.2149,
+      "step": 17485
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.40376e-06,
+      "loss": 1.4632,
+      "step": 17486
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.403491428571428e-06,
+      "loss": 1.2753,
+      "step": 17487
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.403222857142858e-06,
+      "loss": 1.1009,
+      "step": 17488
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.402954285714286e-06,
+      "loss": 1.3718,
+      "step": 17489
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.402685714285715e-06,
+      "loss": 0.7904,
+      "step": 17490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.402417142857143e-06,
+      "loss": 1.3031,
+      "step": 17491
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.402148571428573e-06,
+      "loss": 1.2518,
+      "step": 17492
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.40188e-06,
+      "loss": 1.5145,
+      "step": 17493
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.401611428571429e-06,
+      "loss": 1.1674,
+      "step": 17494
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.401342857142857e-06,
+      "loss": 1.4988,
+      "step": 17495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.401074285714287e-06,
+      "loss": 1.1719,
+      "step": 17496
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.400805714285714e-06,
+      "loss": 0.857,
+      "step": 17497
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.400537142857143e-06,
+      "loss": 0.9471,
+      "step": 17498
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.400268571428572e-06,
+      "loss": 0.8988,
+      "step": 17499
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.4e-06,
+      "loss": 1.2942,
+      "step": 17500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.39973142857143e-06,
+      "loss": 0.9531,
+      "step": 17501
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.399462857142858e-06,
+      "loss": 1.0017,
+      "step": 17502
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.399194285714286e-06,
+      "loss": 1.1582,
+      "step": 17503
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.398925714285714e-06,
+      "loss": 1.1998,
+      "step": 17504
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.398657142857144e-06,
+      "loss": 1.533,
+      "step": 17505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.398388571428572e-06,
+      "loss": 1.3833,
+      "step": 17506
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.39812e-06,
+      "loss": 1.1335,
+      "step": 17507
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.397851428571429e-06,
+      "loss": 0.8725,
+      "step": 17508
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.397582857142858e-06,
+      "loss": 0.7909,
+      "step": 17509
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.397314285714285e-06,
+      "loss": 1.2551,
+      "step": 17510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.397045714285715e-06,
+      "loss": 1.5256,
+      "step": 17511
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.396777142857143e-06,
+      "loss": 0.8143,
+      "step": 17512
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.396508571428571e-06,
+      "loss": 1.3303,
+      "step": 17513
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.396240000000001e-06,
+      "loss": 0.7661,
+      "step": 17514
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.39597142857143e-06,
+      "loss": 1.083,
+      "step": 17515
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.395702857142857e-06,
+      "loss": 1.2407,
+      "step": 17516
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.395434285714286e-06,
+      "loss": 1.6152,
+      "step": 17517
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.395165714285715e-06,
+      "loss": 1.4421,
+      "step": 17518
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.394897142857144e-06,
+      "loss": 1.0388,
+      "step": 17519
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.394628571428572e-06,
+      "loss": 1.2352,
+      "step": 17520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.39436e-06,
+      "loss": 1.3563,
+      "step": 17521
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.39409142857143e-06,
+      "loss": 1.3875,
+      "step": 17522
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.393822857142856e-06,
+      "loss": 1.4637,
+      "step": 17523
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.393554285714286e-06,
+      "loss": 0.7743,
+      "step": 17524
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.393285714285714e-06,
+      "loss": 1.4262,
+      "step": 17525
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.393017142857143e-06,
+      "loss": 0.7876,
+      "step": 17526
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.392748571428572e-06,
+      "loss": 0.9037,
+      "step": 17527
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.39248e-06,
+      "loss": 0.7748,
+      "step": 17528
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.392211428571429e-06,
+      "loss": 1.321,
+      "step": 17529
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.391942857142857e-06,
+      "loss": 1.1727,
+      "step": 17530
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.391674285714287e-06,
+      "loss": 0.889,
+      "step": 17531
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.391405714285715e-06,
+      "loss": 1.7654,
+      "step": 17532
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.391137142857143e-06,
+      "loss": 1.1447,
+      "step": 17533
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.390868571428571e-06,
+      "loss": 1.309,
+      "step": 17534
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.390600000000001e-06,
+      "loss": 0.841,
+      "step": 17535
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.390331428571428e-06,
+      "loss": 1.3914,
+      "step": 17536
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.390062857142858e-06,
+      "loss": 1.1145,
+      "step": 17537
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.389794285714286e-06,
+      "loss": 1.2103,
+      "step": 17538
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.389525714285714e-06,
+      "loss": 0.893,
+      "step": 17539
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.389257142857144e-06,
+      "loss": 1.3942,
+      "step": 17540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.388988571428572e-06,
+      "loss": 1.0689,
+      "step": 17541
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.388720000000002e-06,
+      "loss": 0.8421,
+      "step": 17542
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.388451428571428e-06,
+      "loss": 1.2173,
+      "step": 17543
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.388182857142858e-06,
+      "loss": 1.4377,
+      "step": 17544
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.387914285714286e-06,
+      "loss": 1.3755,
+      "step": 17545
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.387645714285715e-06,
+      "loss": 1.3337,
+      "step": 17546
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.387377142857143e-06,
+      "loss": 1.4992,
+      "step": 17547
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.387108571428573e-06,
+      "loss": 1.0088,
+      "step": 17548
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.386839999999999e-06,
+      "loss": 1.2574,
+      "step": 17549
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.386571428571429e-06,
+      "loss": 1.2748,
+      "step": 17550
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.386302857142857e-06,
+      "loss": 0.9127,
+      "step": 17551
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.386034285714285e-06,
+      "loss": 0.9361,
+      "step": 17552
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.385765714285715e-06,
+      "loss": 1.0649,
+      "step": 17553
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.385497142857143e-06,
+      "loss": 1.2016,
+      "step": 17554
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.385228571428573e-06,
+      "loss": 1.2774,
+      "step": 17555
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.38496e-06,
+      "loss": 1.1839,
+      "step": 17556
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.38469142857143e-06,
+      "loss": 0.8505,
+      "step": 17557
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.384422857142858e-06,
+      "loss": 1.1389,
+      "step": 17558
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.384154285714286e-06,
+      "loss": 1.1143,
+      "step": 17559
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.383885714285714e-06,
+      "loss": 1.1752,
+      "step": 17560
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.383617142857144e-06,
+      "loss": 1.3488,
+      "step": 17561
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.38334857142857e-06,
+      "loss": 1.4465,
+      "step": 17562
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.38308e-06,
+      "loss": 1.1281,
+      "step": 17563
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.382811428571428e-06,
+      "loss": 1.4544,
+      "step": 17564
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.382542857142858e-06,
+      "loss": 1.2472,
+      "step": 17565
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.382274285714287e-06,
+      "loss": 1.3252,
+      "step": 17566
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.382005714285715e-06,
+      "loss": 1.2002,
+      "step": 17567
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.381737142857145e-06,
+      "loss": 1.8185,
+      "step": 17568
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.381468571428571e-06,
+      "loss": 1.2757,
+      "step": 17569
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.381200000000001e-06,
+      "loss": 0.9015,
+      "step": 17570
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.380931428571429e-06,
+      "loss": 0.7835,
+      "step": 17571
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.380662857142857e-06,
+      "loss": 1.2479,
+      "step": 17572
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.380394285714285e-06,
+      "loss": 0.9067,
+      "step": 17573
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.380125714285715e-06,
+      "loss": 1.4235,
+      "step": 17574
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.379857142857142e-06,
+      "loss": 1.293,
+      "step": 17575
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.379588571428572e-06,
+      "loss": 0.8889,
+      "step": 17576
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.37932e-06,
+      "loss": 1.1093,
+      "step": 17577
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.37905142857143e-06,
+      "loss": 1.2218,
+      "step": 17578
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.378782857142858e-06,
+      "loss": 1.0218,
+      "step": 17579
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.378514285714286e-06,
+      "loss": 1.0363,
+      "step": 17580
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.378245714285716e-06,
+      "loss": 1.2714,
+      "step": 17581
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.377977142857142e-06,
+      "loss": 1.3971,
+      "step": 17582
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.377708571428572e-06,
+      "loss": 1.1824,
+      "step": 17583
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.37744e-06,
+      "loss": 1.4122,
+      "step": 17584
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.377171428571429e-06,
+      "loss": 1.6562,
+      "step": 17585
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.376902857142857e-06,
+      "loss": 1.4682,
+      "step": 17586
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.376634285714287e-06,
+      "loss": 0.7708,
+      "step": 17587
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.376365714285715e-06,
+      "loss": 0.9325,
+      "step": 17588
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.376097142857143e-06,
+      "loss": 1.2014,
+      "step": 17589
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.375828571428571e-06,
+      "loss": 1.1586,
+      "step": 17590
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.375560000000001e-06,
+      "loss": 1.1412,
+      "step": 17591
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.37529142857143e-06,
+      "loss": 0.9461,
+      "step": 17592
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.375022857142857e-06,
+      "loss": 1.5847,
+      "step": 17593
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.374754285714287e-06,
+      "loss": 1.1233,
+      "step": 17594
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.374485714285714e-06,
+      "loss": 1.2736,
+      "step": 17595
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.374217142857144e-06,
+      "loss": 1.5081,
+      "step": 17596
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.373948571428572e-06,
+      "loss": 1.4419,
+      "step": 17597
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.37368e-06,
+      "loss": 1.1308,
+      "step": 17598
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.373411428571428e-06,
+      "loss": 1.173,
+      "step": 17599
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.373142857142858e-06,
+      "loss": 1.5396,
+      "step": 17600
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.372874285714286e-06,
+      "loss": 1.0626,
+      "step": 17601
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.372605714285714e-06,
+      "loss": 1.0066,
+      "step": 17602
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.372337142857143e-06,
+      "loss": 1.3071,
+      "step": 17603
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.372068571428572e-06,
+      "loss": 0.9319,
+      "step": 17604
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.3718e-06,
+      "loss": 1.3491,
+      "step": 17605
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.371531428571429e-06,
+      "loss": 1.1433,
+      "step": 17606
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.371262857142859e-06,
+      "loss": 1.2974,
+      "step": 17607
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.370994285714285e-06,
+      "loss": 0.8207,
+      "step": 17608
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.370725714285715e-06,
+      "loss": 1.2005,
+      "step": 17609
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.370457142857143e-06,
+      "loss": 1.6609,
+      "step": 17610
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.370188571428573e-06,
+      "loss": 1.1939,
+      "step": 17611
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.36992e-06,
+      "loss": 1.2513,
+      "step": 17612
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.36965142857143e-06,
+      "loss": 1.4755,
+      "step": 17613
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.369382857142858e-06,
+      "loss": 1.1962,
+      "step": 17614
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.369114285714286e-06,
+      "loss": 1.1107,
+      "step": 17615
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.368845714285714e-06,
+      "loss": 1.0832,
+      "step": 17616
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.368577142857144e-06,
+      "loss": 0.9728,
+      "step": 17617
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.368308571428572e-06,
+      "loss": 0.7931,
+      "step": 17618
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.36804e-06,
+      "loss": 1.629,
+      "step": 17619
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.36777142857143e-06,
+      "loss": 1.0994,
+      "step": 17620
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.367502857142857e-06,
+      "loss": 0.9831,
+      "step": 17621
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.367234285714286e-06,
+      "loss": 1.1732,
+      "step": 17622
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.366965714285715e-06,
+      "loss": 1.4596,
+      "step": 17623
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.366697142857144e-06,
+      "loss": 1.2104,
+      "step": 17624
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.366428571428571e-06,
+      "loss": 1.0085,
+      "step": 17625
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.36616e-06,
+      "loss": 1.4047,
+      "step": 17626
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.365891428571429e-06,
+      "loss": 1.0394,
+      "step": 17627
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.365622857142857e-06,
+      "loss": 1.2174,
+      "step": 17628
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.365354285714285e-06,
+      "loss": 0.9,
+      "step": 17629
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.365085714285715e-06,
+      "loss": 1.1832,
+      "step": 17630
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.364817142857143e-06,
+      "loss": 1.5137,
+      "step": 17631
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.364548571428572e-06,
+      "loss": 1.1609,
+      "step": 17632
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.364280000000001e-06,
+      "loss": 0.8293,
+      "step": 17633
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.36401142857143e-06,
+      "loss": 1.5332,
+      "step": 17634
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.363742857142858e-06,
+      "loss": 1.294,
+      "step": 17635
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.363474285714286e-06,
+      "loss": 1.3826,
+      "step": 17636
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.363205714285716e-06,
+      "loss": 1.2233,
+      "step": 17637
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.362937142857142e-06,
+      "loss": 1.0983,
+      "step": 17638
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.362668571428572e-06,
+      "loss": 1.1375,
+      "step": 17639
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.3624e-06,
+      "loss": 1.109,
+      "step": 17640
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.362131428571429e-06,
+      "loss": 1.529,
+      "step": 17641
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.361862857142857e-06,
+      "loss": 1.0794,
+      "step": 17642
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.361594285714287e-06,
+      "loss": 1.3093,
+      "step": 17643
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.361325714285713e-06,
+      "loss": 1.5584,
+      "step": 17644
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.361057142857143e-06,
+      "loss": 0.9773,
+      "step": 17645
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.360788571428573e-06,
+      "loss": 1.4543,
+      "step": 17646
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.360520000000001e-06,
+      "loss": 1.1888,
+      "step": 17647
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.360251428571429e-06,
+      "loss": 1.3674,
+      "step": 17648
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.359982857142857e-06,
+      "loss": 1.646,
+      "step": 17649
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.359714285714287e-06,
+      "loss": 1.1684,
+      "step": 17650
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.359445714285714e-06,
+      "loss": 0.9524,
+      "step": 17651
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.359177142857144e-06,
+      "loss": 1.4632,
+      "step": 17652
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.358908571428572e-06,
+      "loss": 1.2262,
+      "step": 17653
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.35864e-06,
+      "loss": 1.7738,
+      "step": 17654
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.358371428571428e-06,
+      "loss": 0.7159,
+      "step": 17655
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.358102857142858e-06,
+      "loss": 1.0629,
+      "step": 17656
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.357834285714286e-06,
+      "loss": 1.0609,
+      "step": 17657
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.357565714285714e-06,
+      "loss": 1.1092,
+      "step": 17658
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.357297142857144e-06,
+      "loss": 1.3035,
+      "step": 17659
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.357028571428572e-06,
+      "loss": 1.0813,
+      "step": 17660
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.35676e-06,
+      "loss": 1.3351,
+      "step": 17661
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.356491428571429e-06,
+      "loss": 1.3206,
+      "step": 17662
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.356222857142859e-06,
+      "loss": 1.0287,
+      "step": 17663
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.355954285714285e-06,
+      "loss": 1.1196,
+      "step": 17664
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.355685714285715e-06,
+      "loss": 1.1387,
+      "step": 17665
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.355417142857143e-06,
+      "loss": 1.239,
+      "step": 17666
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.355148571428571e-06,
+      "loss": 1.4629,
+      "step": 17667
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.35488e-06,
+      "loss": 1.0163,
+      "step": 17668
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.35461142857143e-06,
+      "loss": 1.2328,
+      "step": 17669
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.354342857142857e-06,
+      "loss": 1.1558,
+      "step": 17670
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.354074285714286e-06,
+      "loss": 1.5628,
+      "step": 17671
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.353805714285716e-06,
+      "loss": 1.5345,
+      "step": 17672
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.353537142857144e-06,
+      "loss": 1.0836,
+      "step": 17673
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.353268571428572e-06,
+      "loss": 0.7737,
+      "step": 17674
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.353e-06,
+      "loss": 1.0512,
+      "step": 17675
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.35273142857143e-06,
+      "loss": 1.4384,
+      "step": 17676
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.352462857142856e-06,
+      "loss": 1.051,
+      "step": 17677
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.352194285714286e-06,
+      "loss": 0.9497,
+      "step": 17678
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.351925714285714e-06,
+      "loss": 0.951,
+      "step": 17679
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.351657142857144e-06,
+      "loss": 1.2331,
+      "step": 17680
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.35138857142857e-06,
+      "loss": 1.2172,
+      "step": 17681
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.35112e-06,
+      "loss": 0.9901,
+      "step": 17682
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.350851428571429e-06,
+      "loss": 1.3183,
+      "step": 17683
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.350582857142857e-06,
+      "loss": 1.0465,
+      "step": 17684
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.350314285714287e-06,
+      "loss": 0.9425,
+      "step": 17685
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.350045714285715e-06,
+      "loss": 1.7082,
+      "step": 17686
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.349777142857143e-06,
+      "loss": 1.5401,
+      "step": 17687
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.349508571428571e-06,
+      "loss": 0.8424,
+      "step": 17688
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.349240000000001e-06,
+      "loss": 1.3089,
+      "step": 17689
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.348971428571428e-06,
+      "loss": 1.0456,
+      "step": 17690
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.348702857142858e-06,
+      "loss": 1.1005,
+      "step": 17691
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.348434285714286e-06,
+      "loss": 1.0618,
+      "step": 17692
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.348165714285716e-06,
+      "loss": 1.2259,
+      "step": 17693
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.347897142857142e-06,
+      "loss": 0.9787,
+      "step": 17694
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.347628571428572e-06,
+      "loss": 1.171,
+      "step": 17695
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.34736e-06,
+      "loss": 1.5831,
+      "step": 17696
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.347091428571428e-06,
+      "loss": 1.577,
+      "step": 17697
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.346822857142858e-06,
+      "loss": 1.6325,
+      "step": 17698
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.346554285714286e-06,
+      "loss": 1.4931,
+      "step": 17699
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.346285714285715e-06,
+      "loss": 1.2609,
+      "step": 17700
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.346017142857143e-06,
+      "loss": 1.2522,
+      "step": 17701
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.345748571428573e-06,
+      "loss": 1.4046,
+      "step": 17702
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.34548e-06,
+      "loss": 0.8824,
+      "step": 17703
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.345211428571429e-06,
+      "loss": 1.4224,
+      "step": 17704
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.344942857142857e-06,
+      "loss": 0.9683,
+      "step": 17705
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.344674285714287e-06,
+      "loss": 1.627,
+      "step": 17706
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.344405714285714e-06,
+      "loss": 1.0916,
+      "step": 17707
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.344137142857143e-06,
+      "loss": 1.3057,
+      "step": 17708
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.343868571428572e-06,
+      "loss": 0.9028,
+      "step": 17709
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.3436e-06,
+      "loss": 0.9975,
+      "step": 17710
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.34333142857143e-06,
+      "loss": 0.9962,
+      "step": 17711
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.343062857142858e-06,
+      "loss": 1.3986,
+      "step": 17712
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.342794285714286e-06,
+      "loss": 1.0611,
+      "step": 17713
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.342525714285714e-06,
+      "loss": 1.1928,
+      "step": 17714
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.342257142857144e-06,
+      "loss": 0.999,
+      "step": 17715
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.341988571428572e-06,
+      "loss": 1.3071,
+      "step": 17716
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.34172e-06,
+      "loss": 0.8778,
+      "step": 17717
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.341451428571429e-06,
+      "loss": 1.0121,
+      "step": 17718
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.341182857142858e-06,
+      "loss": 1.4576,
+      "step": 17719
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.340914285714285e-06,
+      "loss": 0.7637,
+      "step": 17720
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.340645714285715e-06,
+      "loss": 1.0025,
+      "step": 17721
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.340377142857143e-06,
+      "loss": 1.2405,
+      "step": 17722
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.340108571428571e-06,
+      "loss": 1.5427,
+      "step": 17723
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.339840000000001e-06,
+      "loss": 1.0327,
+      "step": 17724
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.33957142857143e-06,
+      "loss": 1.1244,
+      "step": 17725
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.339302857142857e-06,
+      "loss": 1.1742,
+      "step": 17726
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.339034285714286e-06,
+      "loss": 0.9845,
+      "step": 17727
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.338765714285715e-06,
+      "loss": 1.1298,
+      "step": 17728
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.338497142857144e-06,
+      "loss": 0.975,
+      "step": 17729
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.338228571428572e-06,
+      "loss": 1.0568,
+      "step": 17730
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.33796e-06,
+      "loss": 1.0919,
+      "step": 17731
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.33769142857143e-06,
+      "loss": 1.0418,
+      "step": 17732
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.337422857142856e-06,
+      "loss": 1.0815,
+      "step": 17733
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.337154285714286e-06,
+      "loss": 1.1214,
+      "step": 17734
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.336885714285714e-06,
+      "loss": 1.413,
+      "step": 17735
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.336617142857143e-06,
+      "loss": 1.3665,
+      "step": 17736
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.336348571428572e-06,
+      "loss": 0.8485,
+      "step": 17737
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.33608e-06,
+      "loss": 1.3557,
+      "step": 17738
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.33581142857143e-06,
+      "loss": 1.4823,
+      "step": 17739
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.335542857142857e-06,
+      "loss": 1.1583,
+      "step": 17740
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.335274285714287e-06,
+      "loss": 0.7601,
+      "step": 17741
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.335005714285715e-06,
+      "loss": 1.2298,
+      "step": 17742
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.334737142857143e-06,
+      "loss": 0.8525,
+      "step": 17743
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.334468571428571e-06,
+      "loss": 0.9357,
+      "step": 17744
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.334200000000001e-06,
+      "loss": 0.9348,
+      "step": 17745
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.333931428571428e-06,
+      "loss": 1.7009,
+      "step": 17746
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.333662857142858e-06,
+      "loss": 1.0285,
+      "step": 17747
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.333394285714286e-06,
+      "loss": 1.4887,
+      "step": 17748
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.333125714285714e-06,
+      "loss": 1.7047,
+      "step": 17749
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.332857142857142e-06,
+      "loss": 1.1091,
+      "step": 17750
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.332588571428572e-06,
+      "loss": 0.8973,
+      "step": 17751
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.332320000000002e-06,
+      "loss": 1.0833,
+      "step": 17752
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.332051428571428e-06,
+      "loss": 0.9877,
+      "step": 17753
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.331782857142858e-06,
+      "loss": 1.1834,
+      "step": 17754
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.331514285714286e-06,
+      "loss": 0.952,
+      "step": 17755
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.331245714285715e-06,
+      "loss": 0.7639,
+      "step": 17756
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.330977142857143e-06,
+      "loss": 1.4605,
+      "step": 17757
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.330708571428573e-06,
+      "loss": 1.374,
+      "step": 17758
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.330439999999999e-06,
+      "loss": 1.1942,
+      "step": 17759
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.330171428571429e-06,
+      "loss": 1.1573,
+      "step": 17760
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.329902857142857e-06,
+      "loss": 0.8475,
+      "step": 17761
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.329634285714287e-06,
+      "loss": 1.4707,
+      "step": 17762
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.329365714285713e-06,
+      "loss": 0.9743,
+      "step": 17763
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.329097142857143e-06,
+      "loss": 0.8535,
+      "step": 17764
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.328828571428573e-06,
+      "loss": 1.0549,
+      "step": 17765
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.32856e-06,
+      "loss": 1.1976,
+      "step": 17766
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.32829142857143e-06,
+      "loss": 1.0263,
+      "step": 17767
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.328022857142858e-06,
+      "loss": 1.0334,
+      "step": 17768
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.327754285714286e-06,
+      "loss": 1.5989,
+      "step": 17769
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.327485714285714e-06,
+      "loss": 1.2921,
+      "step": 17770
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.327217142857144e-06,
+      "loss": 1.2238,
+      "step": 17771
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.32694857142857e-06,
+      "loss": 1.2889,
+      "step": 17772
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.32668e-06,
+      "loss": 1.0795,
+      "step": 17773
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.326411428571428e-06,
+      "loss": 1.1545,
+      "step": 17774
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.326142857142858e-06,
+      "loss": 1.0345,
+      "step": 17775
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.325874285714285e-06,
+      "loss": 0.9642,
+      "step": 17776
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.325605714285715e-06,
+      "loss": 1.1166,
+      "step": 17777
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.325337142857145e-06,
+      "loss": 1.152,
+      "step": 17778
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.325068571428571e-06,
+      "loss": 1.235,
+      "step": 17779
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.324800000000001e-06,
+      "loss": 1.1961,
+      "step": 17780
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.324531428571429e-06,
+      "loss": 1.0473,
+      "step": 17781
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.324262857142857e-06,
+      "loss": 1.3741,
+      "step": 17782
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.323994285714285e-06,
+      "loss": 1.4148,
+      "step": 17783
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.323725714285715e-06,
+      "loss": 0.9628,
+      "step": 17784
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.323457142857143e-06,
+      "loss": 1.3863,
+      "step": 17785
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.323188571428572e-06,
+      "loss": 1.1436,
+      "step": 17786
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.32292e-06,
+      "loss": 1.0577,
+      "step": 17787
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.32265142857143e-06,
+      "loss": 1.0458,
+      "step": 17788
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.322382857142856e-06,
+      "loss": 1.3098,
+      "step": 17789
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.322114285714286e-06,
+      "loss": 1.4084,
+      "step": 17790
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.321845714285716e-06,
+      "loss": 1.529,
+      "step": 17791
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.321577142857142e-06,
+      "loss": 1.0745,
+      "step": 17792
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.321308571428572e-06,
+      "loss": 1.3227,
+      "step": 17793
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.32104e-06,
+      "loss": 1.4055,
+      "step": 17794
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.320771428571429e-06,
+      "loss": 1.1973,
+      "step": 17795
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.320502857142857e-06,
+      "loss": 0.8982,
+      "step": 17796
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.320234285714287e-06,
+      "loss": 0.8664,
+      "step": 17797
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.319965714285715e-06,
+      "loss": 1.2011,
+      "step": 17798
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.319697142857143e-06,
+      "loss": 1.0682,
+      "step": 17799
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.319428571428571e-06,
+      "loss": 0.8533,
+      "step": 17800
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.319160000000001e-06,
+      "loss": 1.1971,
+      "step": 17801
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.318891428571428e-06,
+      "loss": 1.2686,
+      "step": 17802
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.318622857142857e-06,
+      "loss": 1.2471,
+      "step": 17803
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.318354285714287e-06,
+      "loss": 1.077,
+      "step": 17804
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.318085714285714e-06,
+      "loss": 1.1338,
+      "step": 17805
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.317817142857144e-06,
+      "loss": 1.3504,
+      "step": 17806
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.317548571428572e-06,
+      "loss": 1.3019,
+      "step": 17807
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.317280000000002e-06,
+      "loss": 0.9025,
+      "step": 17808
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.317011428571428e-06,
+      "loss": 0.9493,
+      "step": 17809
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.316742857142858e-06,
+      "loss": 1.2348,
+      "step": 17810
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.316474285714286e-06,
+      "loss": 1.3192,
+      "step": 17811
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.316205714285714e-06,
+      "loss": 0.9661,
+      "step": 17812
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.315937142857143e-06,
+      "loss": 0.8351,
+      "step": 17813
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.315668571428572e-06,
+      "loss": 1.2237,
+      "step": 17814
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.315399999999999e-06,
+      "loss": 1.0462,
+      "step": 17815
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.315131428571429e-06,
+      "loss": 1.0571,
+      "step": 17816
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.314862857142859e-06,
+      "loss": 0.9661,
+      "step": 17817
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.314594285714285e-06,
+      "loss": 1.7254,
+      "step": 17818
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.314325714285715e-06,
+      "loss": 1.5244,
+      "step": 17819
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.314057142857143e-06,
+      "loss": 1.2144,
+      "step": 17820
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.313788571428573e-06,
+      "loss": 1.2137,
+      "step": 17821
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.31352e-06,
+      "loss": 1.3294,
+      "step": 17822
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.31325142857143e-06,
+      "loss": 0.7501,
+      "step": 17823
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.312982857142858e-06,
+      "loss": 1.0921,
+      "step": 17824
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.312714285714286e-06,
+      "loss": 1.0438,
+      "step": 17825
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.312445714285714e-06,
+      "loss": 1.4009,
+      "step": 17826
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.312177142857144e-06,
+      "loss": 0.9231,
+      "step": 17827
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.31190857142857e-06,
+      "loss": 1.0633,
+      "step": 17828
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.31164e-06,
+      "loss": 1.2616,
+      "step": 17829
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.31137142857143e-06,
+      "loss": 0.6882,
+      "step": 17830
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.311102857142858e-06,
+      "loss": 1.1016,
+      "step": 17831
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.310834285714286e-06,
+      "loss": 0.8567,
+      "step": 17832
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.310565714285715e-06,
+      "loss": 1.031,
+      "step": 17833
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.310297142857144e-06,
+      "loss": 1.3884,
+      "step": 17834
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.310028571428571e-06,
+      "loss": 1.3311,
+      "step": 17835
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.30976e-06,
+      "loss": 1.516,
+      "step": 17836
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.309491428571429e-06,
+      "loss": 1.1168,
+      "step": 17837
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.309222857142857e-06,
+      "loss": 0.9481,
+      "step": 17838
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.308954285714285e-06,
+      "loss": 0.9772,
+      "step": 17839
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.308685714285715e-06,
+      "loss": 1.6558,
+      "step": 17840
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.308417142857142e-06,
+      "loss": 1.0659,
+      "step": 17841
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.308148571428572e-06,
+      "loss": 1.2618,
+      "step": 17842
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.307880000000001e-06,
+      "loss": 1.0739,
+      "step": 17843
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.30761142857143e-06,
+      "loss": 1.0248,
+      "step": 17844
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.307342857142858e-06,
+      "loss": 1.3216,
+      "step": 17845
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.307074285714286e-06,
+      "loss": 1.1885,
+      "step": 17846
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.306805714285716e-06,
+      "loss": 1.0625,
+      "step": 17847
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.306537142857142e-06,
+      "loss": 1.4584,
+      "step": 17848
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.306268571428572e-06,
+      "loss": 1.0335,
+      "step": 17849
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.306e-06,
+      "loss": 1.1714,
+      "step": 17850
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.305731428571429e-06,
+      "loss": 0.9668,
+      "step": 17851
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.305462857142857e-06,
+      "loss": 1.0266,
+      "step": 17852
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.305194285714287e-06,
+      "loss": 1.4287,
+      "step": 17853
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.304925714285715e-06,
+      "loss": 1.388,
+      "step": 17854
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.304657142857143e-06,
+      "loss": 1.5479,
+      "step": 17855
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.304388571428571e-06,
+      "loss": 1.133,
+      "step": 17856
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.304120000000001e-06,
+      "loss": 1.4215,
+      "step": 17857
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.303851428571429e-06,
+      "loss": 0.8602,
+      "step": 17858
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.303582857142857e-06,
+      "loss": 1.113,
+      "step": 17859
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.303314285714287e-06,
+      "loss": 1.0398,
+      "step": 17860
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.303045714285714e-06,
+      "loss": 1.538,
+      "step": 17861
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.302777142857144e-06,
+      "loss": 1.1629,
+      "step": 17862
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.302508571428572e-06,
+      "loss": 1.0929,
+      "step": 17863
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.30224e-06,
+      "loss": 1.1761,
+      "step": 17864
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.301971428571428e-06,
+      "loss": 1.1653,
+      "step": 17865
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.301702857142858e-06,
+      "loss": 0.852,
+      "step": 17866
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.301434285714286e-06,
+      "loss": 1.2454,
+      "step": 17867
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.301165714285714e-06,
+      "loss": 1.2899,
+      "step": 17868
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.300897142857142e-06,
+      "loss": 1.3253,
+      "step": 17869
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.300628571428572e-06,
+      "loss": 1.3374,
+      "step": 17870
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.30036e-06,
+      "loss": 1.0492,
+      "step": 17871
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.300091428571429e-06,
+      "loss": 1.4617,
+      "step": 17872
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.299822857142859e-06,
+      "loss": 1.6727,
+      "step": 17873
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.299554285714285e-06,
+      "loss": 1.2928,
+      "step": 17874
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.299285714285715e-06,
+      "loss": 1.2794,
+      "step": 17875
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.299017142857143e-06,
+      "loss": 0.9178,
+      "step": 17876
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.298748571428571e-06,
+      "loss": 1.4865,
+      "step": 17877
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.29848e-06,
+      "loss": 1.3721,
+      "step": 17878
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.29821142857143e-06,
+      "loss": 1.499,
+      "step": 17879
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.297942857142857e-06,
+      "loss": 1.2874,
+      "step": 17880
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.297674285714286e-06,
+      "loss": 1.1226,
+      "step": 17881
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.297405714285714e-06,
+      "loss": 0.9853,
+      "step": 17882
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.297137142857144e-06,
+      "loss": 1.2869,
+      "step": 17883
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.296868571428572e-06,
+      "loss": 1.4419,
+      "step": 17884
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.2966e-06,
+      "loss": 1.5177,
+      "step": 17885
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.29633142857143e-06,
+      "loss": 1.0541,
+      "step": 17886
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.296062857142856e-06,
+      "loss": 1.2595,
+      "step": 17887
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.295794285714286e-06,
+      "loss": 1.1221,
+      "step": 17888
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.295525714285714e-06,
+      "loss": 0.843,
+      "step": 17889
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.295257142857144e-06,
+      "loss": 0.8946,
+      "step": 17890
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.29498857142857e-06,
+      "loss": 1.0644,
+      "step": 17891
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.29472e-06,
+      "loss": 0.8319,
+      "step": 17892
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.294451428571429e-06,
+      "loss": 0.9889,
+      "step": 17893
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.294182857142857e-06,
+      "loss": 1.5918,
+      "step": 17894
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.293914285714285e-06,
+      "loss": 1.0776,
+      "step": 17895
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.293645714285715e-06,
+      "loss": 1.0633,
+      "step": 17896
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.293377142857143e-06,
+      "loss": 1.3401,
+      "step": 17897
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.293108571428571e-06,
+      "loss": 1.5043,
+      "step": 17898
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.292840000000001e-06,
+      "loss": 1.0269,
+      "step": 17899
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.292571428571428e-06,
+      "loss": 0.8897,
+      "step": 17900
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.292302857142858e-06,
+      "loss": 1.026,
+      "step": 17901
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.292034285714286e-06,
+      "loss": 1.1809,
+      "step": 17902
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.291765714285716e-06,
+      "loss": 1.6858,
+      "step": 17903
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.291497142857142e-06,
+      "loss": 1.1568,
+      "step": 17904
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.291228571428572e-06,
+      "loss": 0.9623,
+      "step": 17905
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.29096e-06,
+      "loss": 1.3219,
+      "step": 17906
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.290691428571428e-06,
+      "loss": 0.9246,
+      "step": 17907
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.290422857142857e-06,
+      "loss": 0.9532,
+      "step": 17908
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.290154285714286e-06,
+      "loss": 1.238,
+      "step": 17909
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.289885714285715e-06,
+      "loss": 1.4929,
+      "step": 17910
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.289617142857143e-06,
+      "loss": 1.0045,
+      "step": 17911
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.289348571428573e-06,
+      "loss": 1.283,
+      "step": 17912
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.289080000000001e-06,
+      "loss": 1.1591,
+      "step": 17913
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.288811428571429e-06,
+      "loss": 1.1974,
+      "step": 17914
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.288542857142857e-06,
+      "loss": 1.4676,
+      "step": 17915
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.288274285714287e-06,
+      "loss": 1.4168,
+      "step": 17916
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.288005714285714e-06,
+      "loss": 1.9172,
+      "step": 17917
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.287737142857143e-06,
+      "loss": 1.371,
+      "step": 17918
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.287468571428572e-06,
+      "loss": 1.0901,
+      "step": 17919
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.2872e-06,
+      "loss": 1.2805,
+      "step": 17920
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.286931428571428e-06,
+      "loss": 1.5537,
+      "step": 17921
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.286662857142858e-06,
+      "loss": 1.1156,
+      "step": 17922
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.286394285714286e-06,
+      "loss": 1.4947,
+      "step": 17923
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.286125714285714e-06,
+      "loss": 0.9606,
+      "step": 17924
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.285857142857144e-06,
+      "loss": 1.4018,
+      "step": 17925
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.285588571428572e-06,
+      "loss": 1.117,
+      "step": 17926
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.28532e-06,
+      "loss": 1.1292,
+      "step": 17927
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.285051428571429e-06,
+      "loss": 1.1928,
+      "step": 17928
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.284782857142858e-06,
+      "loss": 1.3311,
+      "step": 17929
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.284514285714285e-06,
+      "loss": 1.093,
+      "step": 17930
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.284245714285715e-06,
+      "loss": 0.8918,
+      "step": 17931
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.283977142857143e-06,
+      "loss": 1.4328,
+      "step": 17932
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.283708571428571e-06,
+      "loss": 1.5415,
+      "step": 17933
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.28344e-06,
+      "loss": 1.4077,
+      "step": 17934
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.28317142857143e-06,
+      "loss": 0.7454,
+      "step": 17935
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.282902857142859e-06,
+      "loss": 1.1304,
+      "step": 17936
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.282634285714286e-06,
+      "loss": 0.986,
+      "step": 17937
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.282365714285715e-06,
+      "loss": 1.2721,
+      "step": 17938
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.282097142857144e-06,
+      "loss": 0.9913,
+      "step": 17939
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.281828571428572e-06,
+      "loss": 0.9066,
+      "step": 17940
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.28156e-06,
+      "loss": 1.2423,
+      "step": 17941
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.28129142857143e-06,
+      "loss": 0.7262,
+      "step": 17942
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.281022857142856e-06,
+      "loss": 1.3143,
+      "step": 17943
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.280754285714286e-06,
+      "loss": 1.2929,
+      "step": 17944
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.280485714285714e-06,
+      "loss": 1.3879,
+      "step": 17945
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.280217142857143e-06,
+      "loss": 0.8871,
+      "step": 17946
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.27994857142857e-06,
+      "loss": 1.1861,
+      "step": 17947
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.27968e-06,
+      "loss": 1.476,
+      "step": 17948
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.27941142857143e-06,
+      "loss": 0.9353,
+      "step": 17949
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.279142857142857e-06,
+      "loss": 1.4478,
+      "step": 17950
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.278874285714287e-06,
+      "loss": 1.329,
+      "step": 17951
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.278605714285715e-06,
+      "loss": 1.207,
+      "step": 17952
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.278337142857143e-06,
+      "loss": 1.0992,
+      "step": 17953
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.278068571428571e-06,
+      "loss": 1.5932,
+      "step": 17954
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.277800000000001e-06,
+      "loss": 1.1674,
+      "step": 17955
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.277531428571428e-06,
+      "loss": 1.026,
+      "step": 17956
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.277262857142858e-06,
+      "loss": 1.3823,
+      "step": 17957
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.276994285714286e-06,
+      "loss": 1.0894,
+      "step": 17958
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.276725714285716e-06,
+      "loss": 1.0381,
+      "step": 17959
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.276457142857142e-06,
+      "loss": 0.734,
+      "step": 17960
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.276188571428572e-06,
+      "loss": 1.2907,
+      "step": 17961
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.27592e-06,
+      "loss": 1.2317,
+      "step": 17962
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.275651428571428e-06,
+      "loss": 0.7468,
+      "step": 17963
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.275382857142858e-06,
+      "loss": 0.6615,
+      "step": 17964
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.275114285714286e-06,
+      "loss": 1.0419,
+      "step": 17965
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.274845714285715e-06,
+      "loss": 1.1973,
+      "step": 17966
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.274577142857143e-06,
+      "loss": 1.0941,
+      "step": 17967
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.274308571428573e-06,
+      "loss": 1.2779,
+      "step": 17968
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.274039999999999e-06,
+      "loss": 1.6899,
+      "step": 17969
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.273771428571429e-06,
+      "loss": 0.8943,
+      "step": 17970
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.273502857142857e-06,
+      "loss": 1.3158,
+      "step": 17971
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.273234285714287e-06,
+      "loss": 1.3365,
+      "step": 17972
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.272965714285713e-06,
+      "loss": 0.9363,
+      "step": 17973
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.272697142857143e-06,
+      "loss": 0.979,
+      "step": 17974
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.272428571428571e-06,
+      "loss": 1.5828,
+      "step": 17975
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.27216e-06,
+      "loss": 1.6733,
+      "step": 17976
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.27189142857143e-06,
+      "loss": 1.2864,
+      "step": 17977
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.271622857142858e-06,
+      "loss": 1.7735,
+      "step": 17978
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.271354285714286e-06,
+      "loss": 1.3614,
+      "step": 17979
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.271085714285714e-06,
+      "loss": 0.9066,
+      "step": 17980
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.270817142857144e-06,
+      "loss": 0.9501,
+      "step": 17981
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.270548571428572e-06,
+      "loss": 1.2847,
+      "step": 17982
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.27028e-06,
+      "loss": 1.0507,
+      "step": 17983
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.270011428571428e-06,
+      "loss": 1.2668,
+      "step": 17984
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.269742857142858e-06,
+      "loss": 0.7635,
+      "step": 17985
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.269474285714285e-06,
+      "loss": 1.3302,
+      "step": 17986
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.269205714285715e-06,
+      "loss": 1.0352,
+      "step": 17987
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.268937142857143e-06,
+      "loss": 1.0624,
+      "step": 17988
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.268668571428571e-06,
+      "loss": 1.0683,
+      "step": 17989
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.268400000000001e-06,
+      "loss": 0.8832,
+      "step": 17990
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.268131428571429e-06,
+      "loss": 0.8883,
+      "step": 17991
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.267862857142857e-06,
+      "loss": 1.3713,
+      "step": 17992
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.267594285714285e-06,
+      "loss": 0.8756,
+      "step": 17993
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.267325714285715e-06,
+      "loss": 1.0639,
+      "step": 17994
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.267057142857143e-06,
+      "loss": 1.1703,
+      "step": 17995
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.266788571428572e-06,
+      "loss": 1.1857,
+      "step": 17996
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.26652e-06,
+      "loss": 1.0175,
+      "step": 17997
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.26625142857143e-06,
+      "loss": 0.8704,
+      "step": 17998
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.265982857142856e-06,
+      "loss": 1.1549,
+      "step": 17999
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.265714285714286e-06,
+      "loss": 1.2912,
+      "step": 18000
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.265445714285714e-06,
+      "loss": 1.532,
+      "step": 18001
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.265177142857142e-06,
+      "loss": 0.8658,
+      "step": 18002
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.264908571428572e-06,
+      "loss": 0.8711,
+      "step": 18003
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.26464e-06,
+      "loss": 0.9506,
+      "step": 18004
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.26437142857143e-06,
+      "loss": 1.2128,
+      "step": 18005
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.264102857142857e-06,
+      "loss": 1.6138,
+      "step": 18006
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.263834285714287e-06,
+      "loss": 1.142,
+      "step": 18007
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.263565714285715e-06,
+      "loss": 1.3935,
+      "step": 18008
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.263297142857143e-06,
+      "loss": 1.2055,
+      "step": 18009
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.263028571428571e-06,
+      "loss": 0.8639,
+      "step": 18010
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.262760000000001e-06,
+      "loss": 1.271,
+      "step": 18011
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.262491428571428e-06,
+      "loss": 1.3596,
+      "step": 18012
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.262222857142857e-06,
+      "loss": 1.3728,
+      "step": 18013
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.261954285714286e-06,
+      "loss": 1.109,
+      "step": 18014
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.261685714285714e-06,
+      "loss": 1.2993,
+      "step": 18015
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.261417142857144e-06,
+      "loss": 0.85,
+      "step": 18016
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.261148571428572e-06,
+      "loss": 1.053,
+      "step": 18017
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.260880000000002e-06,
+      "loss": 1.4389,
+      "step": 18018
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.260611428571428e-06,
+      "loss": 1.2584,
+      "step": 18019
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.260342857142858e-06,
+      "loss": 1.0958,
+      "step": 18020
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.260074285714286e-06,
+      "loss": 1.1147,
+      "step": 18021
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.259805714285714e-06,
+      "loss": 0.9691,
+      "step": 18022
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.259537142857143e-06,
+      "loss": 1.2164,
+      "step": 18023
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.259268571428572e-06,
+      "loss": 1.0229,
+      "step": 18024
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.258999999999999e-06,
+      "loss": 1.0587,
+      "step": 18025
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.258731428571429e-06,
+      "loss": 1.2152,
+      "step": 18026
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.258462857142857e-06,
+      "loss": 1.0868,
+      "step": 18027
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.258194285714285e-06,
+      "loss": 1.2096,
+      "step": 18028
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.257925714285715e-06,
+      "loss": 1.1676,
+      "step": 18029
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.257657142857143e-06,
+      "loss": 1.3109,
+      "step": 18030
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.257388571428573e-06,
+      "loss": 1.1251,
+      "step": 18031
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.25712e-06,
+      "loss": 1.4191,
+      "step": 18032
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.25685142857143e-06,
+      "loss": 1.4805,
+      "step": 18033
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.256582857142858e-06,
+      "loss": 1.368,
+      "step": 18034
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.256314285714286e-06,
+      "loss": 1.0719,
+      "step": 18035
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.256045714285714e-06,
+      "loss": 1.271,
+      "step": 18036
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.255777142857144e-06,
+      "loss": 0.9222,
+      "step": 18037
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.25550857142857e-06,
+      "loss": 1.016,
+      "step": 18038
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.25524e-06,
+      "loss": 0.8679,
+      "step": 18039
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.254971428571428e-06,
+      "loss": 0.8192,
+      "step": 18040
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.254702857142858e-06,
+      "loss": 1.1497,
+      "step": 18041
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.254434285714286e-06,
+      "loss": 1.1208,
+      "step": 18042
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.254165714285715e-06,
+      "loss": 1.2621,
+      "step": 18043
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.253897142857144e-06,
+      "loss": 0.9652,
+      "step": 18044
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.253628571428571e-06,
+      "loss": 1.4609,
+      "step": 18045
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.25336e-06,
+      "loss": 1.5848,
+      "step": 18046
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.253091428571429e-06,
+      "loss": 1.1324,
+      "step": 18047
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.252822857142857e-06,
+      "loss": 0.7594,
+      "step": 18048
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.252554285714285e-06,
+      "loss": 1.1192,
+      "step": 18049
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.252285714285715e-06,
+      "loss": 1.0857,
+      "step": 18050
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.252017142857142e-06,
+      "loss": 1.1941,
+      "step": 18051
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.251748571428572e-06,
+      "loss": 1.2896,
+      "step": 18052
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.25148e-06,
+      "loss": 1.0221,
+      "step": 18053
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.25121142857143e-06,
+      "loss": 1.2153,
+      "step": 18054
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.250942857142858e-06,
+      "loss": 1.1748,
+      "step": 18055
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.250674285714286e-06,
+      "loss": 1.4453,
+      "step": 18056
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.250405714285716e-06,
+      "loss": 1.1968,
+      "step": 18057
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.250137142857142e-06,
+      "loss": 0.9462,
+      "step": 18058
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.249868571428572e-06,
+      "loss": 1.2691,
+      "step": 18059
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.2496e-06,
+      "loss": 1.604,
+      "step": 18060
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.249331428571429e-06,
+      "loss": 0.9754,
+      "step": 18061
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.249062857142857e-06,
+      "loss": 1.3329,
+      "step": 18062
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.248794285714287e-06,
+      "loss": 1.3742,
+      "step": 18063
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.248525714285715e-06,
+      "loss": 1.3971,
+      "step": 18064
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.248257142857143e-06,
+      "loss": 1.1475,
+      "step": 18065
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.247988571428571e-06,
+      "loss": 1.2694,
+      "step": 18066
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.247720000000001e-06,
+      "loss": 1.315,
+      "step": 18067
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.247451428571429e-06,
+      "loss": 1.3305,
+      "step": 18068
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.247182857142857e-06,
+      "loss": 1.0034,
+      "step": 18069
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.246914285714287e-06,
+      "loss": 1.0936,
+      "step": 18070
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.246645714285714e-06,
+      "loss": 1.2967,
+      "step": 18071
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.246377142857144e-06,
+      "loss": 1.0728,
+      "step": 18072
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.246108571428572e-06,
+      "loss": 1.47,
+      "step": 18073
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.24584e-06,
+      "loss": 1.0593,
+      "step": 18074
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.245571428571428e-06,
+      "loss": 1.3481,
+      "step": 18075
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.245302857142858e-06,
+      "loss": 1.2289,
+      "step": 18076
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.245034285714286e-06,
+      "loss": 1.1278,
+      "step": 18077
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.244765714285714e-06,
+      "loss": 1.0622,
+      "step": 18078
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.244497142857142e-06,
+      "loss": 1.6152,
+      "step": 18079
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.244228571428572e-06,
+      "loss": 1.1802,
+      "step": 18080
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.24396e-06,
+      "loss": 1.3167,
+      "step": 18081
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.243691428571429e-06,
+      "loss": 1.344,
+      "step": 18082
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.243422857142859e-06,
+      "loss": 0.8756,
+      "step": 18083
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.243154285714285e-06,
+      "loss": 1.4085,
+      "step": 18084
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.242885714285715e-06,
+      "loss": 1.1433,
+      "step": 18085
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.242617142857143e-06,
+      "loss": 0.7424,
+      "step": 18086
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.242348571428573e-06,
+      "loss": 1.3681,
+      "step": 18087
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.24208e-06,
+      "loss": 1.5687,
+      "step": 18088
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.24181142857143e-06,
+      "loss": 1.0828,
+      "step": 18089
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.241542857142857e-06,
+      "loss": 1.462,
+      "step": 18090
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.241274285714286e-06,
+      "loss": 1.1025,
+      "step": 18091
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.241005714285714e-06,
+      "loss": 1.3181,
+      "step": 18092
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.240737142857144e-06,
+      "loss": 0.8995,
+      "step": 18093
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.240468571428572e-06,
+      "loss": 1.1585,
+      "step": 18094
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.2402e-06,
+      "loss": 1.506,
+      "step": 18095
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.23993142857143e-06,
+      "loss": 1.2028,
+      "step": 18096
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.239662857142856e-06,
+      "loss": 1.3469,
+      "step": 18097
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.239394285714286e-06,
+      "loss": 0.8964,
+      "step": 18098
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.239125714285714e-06,
+      "loss": 1.3255,
+      "step": 18099
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.238857142857144e-06,
+      "loss": 0.9096,
+      "step": 18100
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.23858857142857e-06,
+      "loss": 1.3882,
+      "step": 18101
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.23832e-06,
+      "loss": 0.9104,
+      "step": 18102
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.238051428571429e-06,
+      "loss": 1.0962,
+      "step": 18103
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.237782857142857e-06,
+      "loss": 0.995,
+      "step": 18104
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.237514285714285e-06,
+      "loss": 1.0558,
+      "step": 18105
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.237245714285715e-06,
+      "loss": 1.0777,
+      "step": 18106
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.236977142857143e-06,
+      "loss": 1.0809,
+      "step": 18107
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.236708571428571e-06,
+      "loss": 1.2391,
+      "step": 18108
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.236440000000001e-06,
+      "loss": 1.0764,
+      "step": 18109
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.23617142857143e-06,
+      "loss": 1.4208,
+      "step": 18110
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.235902857142858e-06,
+      "loss": 0.7748,
+      "step": 18111
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.235634285714286e-06,
+      "loss": 1.161,
+      "step": 18112
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.235365714285716e-06,
+      "loss": 1.5139,
+      "step": 18113
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.235097142857142e-06,
+      "loss": 1.1011,
+      "step": 18114
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.234828571428572e-06,
+      "loss": 1.2567,
+      "step": 18115
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.23456e-06,
+      "loss": 1.7358,
+      "step": 18116
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.234291428571428e-06,
+      "loss": 1.2731,
+      "step": 18117
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.234022857142857e-06,
+      "loss": 1.3036,
+      "step": 18118
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.233754285714286e-06,
+      "loss": 1.181,
+      "step": 18119
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.233485714285715e-06,
+      "loss": 1.3944,
+      "step": 18120
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.233217142857143e-06,
+      "loss": 1.3519,
+      "step": 18121
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.232948571428573e-06,
+      "loss": 1.4578,
+      "step": 18122
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.232680000000001e-06,
+      "loss": 1.0384,
+      "step": 18123
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.232411428571429e-06,
+      "loss": 0.8812,
+      "step": 18124
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.232142857142857e-06,
+      "loss": 1.0158,
+      "step": 18125
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.231874285714287e-06,
+      "loss": 1.1289,
+      "step": 18126
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.231605714285714e-06,
+      "loss": 1.2508,
+      "step": 18127
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.231337142857143e-06,
+      "loss": 1.3462,
+      "step": 18128
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.231068571428572e-06,
+      "loss": 0.7854,
+      "step": 18129
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.2308e-06,
+      "loss": 1.7401,
+      "step": 18130
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.230531428571428e-06,
+      "loss": 1.1209,
+      "step": 18131
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.230262857142858e-06,
+      "loss": 1.2707,
+      "step": 18132
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.229994285714286e-06,
+      "loss": 1.4781,
+      "step": 18133
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.229725714285714e-06,
+      "loss": 1.2069,
+      "step": 18134
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.229457142857144e-06,
+      "loss": 1.0907,
+      "step": 18135
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.229188571428572e-06,
+      "loss": 1.1129,
+      "step": 18136
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.22892e-06,
+      "loss": 0.8709,
+      "step": 18137
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.228651428571429e-06,
+      "loss": 1.2309,
+      "step": 18138
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.228382857142858e-06,
+      "loss": 1.4477,
+      "step": 18139
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.228114285714285e-06,
+      "loss": 1.3743,
+      "step": 18140
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.227845714285715e-06,
+      "loss": 1.2171,
+      "step": 18141
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.227577142857143e-06,
+      "loss": 1.5092,
+      "step": 18142
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.227308571428571e-06,
+      "loss": 1.2604,
+      "step": 18143
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.22704e-06,
+      "loss": 1.4939,
+      "step": 18144
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.22677142857143e-06,
+      "loss": 1.3573,
+      "step": 18145
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.226502857142857e-06,
+      "loss": 1.2129,
+      "step": 18146
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.226234285714286e-06,
+      "loss": 0.6769,
+      "step": 18147
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.225965714285715e-06,
+      "loss": 1.0749,
+      "step": 18148
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.225697142857144e-06,
+      "loss": 0.9551,
+      "step": 18149
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.225428571428572e-06,
+      "loss": 1.1954,
+      "step": 18150
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.22516e-06,
+      "loss": 1.2454,
+      "step": 18151
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.22489142857143e-06,
+      "loss": 1.4911,
+      "step": 18152
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.224622857142856e-06,
+      "loss": 1.0312,
+      "step": 18153
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.224354285714286e-06,
+      "loss": 0.9264,
+      "step": 18154
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.224085714285714e-06,
+      "loss": 1.0407,
+      "step": 18155
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.223817142857144e-06,
+      "loss": 1.5979,
+      "step": 18156
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.22354857142857e-06,
+      "loss": 1.2987,
+      "step": 18157
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.22328e-06,
+      "loss": 1.328,
+      "step": 18158
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.223011428571429e-06,
+      "loss": 1.1196,
+      "step": 18159
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.222742857142857e-06,
+      "loss": 1.2404,
+      "step": 18160
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.222474285714287e-06,
+      "loss": 1.1861,
+      "step": 18161
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.222205714285715e-06,
+      "loss": 1.2411,
+      "step": 18162
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.221937142857143e-06,
+      "loss": 1.4826,
+      "step": 18163
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.221668571428571e-06,
+      "loss": 1.3428,
+      "step": 18164
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.221400000000001e-06,
+      "loss": 1.4151,
+      "step": 18165
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.221131428571428e-06,
+      "loss": 0.9936,
+      "step": 18166
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.220862857142858e-06,
+      "loss": 0.9872,
+      "step": 18167
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.220594285714286e-06,
+      "loss": 1.3507,
+      "step": 18168
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.220325714285716e-06,
+      "loss": 1.2461,
+      "step": 18169
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.220057142857142e-06,
+      "loss": 1.579,
+      "step": 18170
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.219788571428572e-06,
+      "loss": 0.9432,
+      "step": 18171
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.21952e-06,
+      "loss": 1.0457,
+      "step": 18172
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.219251428571428e-06,
+      "loss": 1.53,
+      "step": 18173
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.218982857142858e-06,
+      "loss": 1.4023,
+      "step": 18174
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.218714285714286e-06,
+      "loss": 1.209,
+      "step": 18175
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.218445714285715e-06,
+      "loss": 1.3819,
+      "step": 18176
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.218177142857143e-06,
+      "loss": 1.1996,
+      "step": 18177
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.217908571428573e-06,
+      "loss": 0.9813,
+      "step": 18178
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.21764e-06,
+      "loss": 1.5007,
+      "step": 18179
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.217371428571429e-06,
+      "loss": 1.1232,
+      "step": 18180
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.217102857142857e-06,
+      "loss": 1.4301,
+      "step": 18181
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.216834285714287e-06,
+      "loss": 1.6501,
+      "step": 18182
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.216565714285713e-06,
+      "loss": 1.3575,
+      "step": 18183
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.216297142857143e-06,
+      "loss": 1.4176,
+      "step": 18184
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.216028571428571e-06,
+      "loss": 1.3047,
+      "step": 18185
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.21576e-06,
+      "loss": 0.6952,
+      "step": 18186
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.21549142857143e-06,
+      "loss": 0.9597,
+      "step": 18187
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.215222857142858e-06,
+      "loss": 1.2911,
+      "step": 18188
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.214954285714286e-06,
+      "loss": 1.0232,
+      "step": 18189
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.214685714285714e-06,
+      "loss": 1.1517,
+      "step": 18190
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.214417142857144e-06,
+      "loss": 1.3773,
+      "step": 18191
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.214148571428572e-06,
+      "loss": 1.1082,
+      "step": 18192
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.21388e-06,
+      "loss": 1.1213,
+      "step": 18193
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.213611428571428e-06,
+      "loss": 1.1542,
+      "step": 18194
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.213342857142858e-06,
+      "loss": 1.4208,
+      "step": 18195
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.213074285714285e-06,
+      "loss": 0.7721,
+      "step": 18196
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.212805714285715e-06,
+      "loss": 1.1442,
+      "step": 18197
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.212537142857143e-06,
+      "loss": 1.2576,
+      "step": 18198
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.212268571428571e-06,
+      "loss": 1.0418,
+      "step": 18199
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.212000000000001e-06,
+      "loss": 1.0952,
+      "step": 18200
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.211731428571429e-06,
+      "loss": 0.9199,
+      "step": 18201
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.211462857142857e-06,
+      "loss": 1.1746,
+      "step": 18202
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.211194285714285e-06,
+      "loss": 0.9501,
+      "step": 18203
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.210925714285715e-06,
+      "loss": 1.2191,
+      "step": 18204
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.210657142857143e-06,
+      "loss": 0.9133,
+      "step": 18205
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.210388571428572e-06,
+      "loss": 1.2962,
+      "step": 18206
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.21012e-06,
+      "loss": 1.1815,
+      "step": 18207
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.20985142857143e-06,
+      "loss": 0.9976,
+      "step": 18208
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.209582857142856e-06,
+      "loss": 1.2326,
+      "step": 18209
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.209314285714286e-06,
+      "loss": 0.9179,
+      "step": 18210
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.209045714285714e-06,
+      "loss": 1.1595,
+      "step": 18211
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.208777142857142e-06,
+      "loss": 0.5653,
+      "step": 18212
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.208508571428572e-06,
+      "loss": 0.7858,
+      "step": 18213
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.20824e-06,
+      "loss": 0.9952,
+      "step": 18214
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.20797142857143e-06,
+      "loss": 1.5768,
+      "step": 18215
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.207702857142857e-06,
+      "loss": 1.164,
+      "step": 18216
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.207434285714287e-06,
+      "loss": 1.0806,
+      "step": 18217
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.207165714285715e-06,
+      "loss": 1.0322,
+      "step": 18218
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.206897142857143e-06,
+      "loss": 1.392,
+      "step": 18219
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.206628571428571e-06,
+      "loss": 1.1102,
+      "step": 18220
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.206360000000001e-06,
+      "loss": 1.0764,
+      "step": 18221
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.206091428571428e-06,
+      "loss": 0.9276,
+      "step": 18222
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.205822857142857e-06,
+      "loss": 1.7103,
+      "step": 18223
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.205554285714286e-06,
+      "loss": 1.527,
+      "step": 18224
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.205285714285714e-06,
+      "loss": 0.9245,
+      "step": 18225
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.205017142857144e-06,
+      "loss": 1.598,
+      "step": 18226
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.204748571428572e-06,
+      "loss": 1.2944,
+      "step": 18227
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.204480000000002e-06,
+      "loss": 1.4176,
+      "step": 18228
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.204211428571428e-06,
+      "loss": 1.2758,
+      "step": 18229
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.203942857142858e-06,
+      "loss": 1.2405,
+      "step": 18230
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.203674285714286e-06,
+      "loss": 0.9509,
+      "step": 18231
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.203405714285714e-06,
+      "loss": 0.9805,
+      "step": 18232
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.203137142857143e-06,
+      "loss": 1.2428,
+      "step": 18233
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.202868571428572e-06,
+      "loss": 1.5974,
+      "step": 18234
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.202599999999999e-06,
+      "loss": 1.285,
+      "step": 18235
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.202331428571429e-06,
+      "loss": 1.3571,
+      "step": 18236
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.202062857142857e-06,
+      "loss": 0.8278,
+      "step": 18237
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.201794285714287e-06,
+      "loss": 1.0871,
+      "step": 18238
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.201525714285715e-06,
+      "loss": 0.9455,
+      "step": 18239
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.201257142857143e-06,
+      "loss": 1.2437,
+      "step": 18240
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.200988571428573e-06,
+      "loss": 1.288,
+      "step": 18241
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.20072e-06,
+      "loss": 1.3567,
+      "step": 18242
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.20045142857143e-06,
+      "loss": 1.2295,
+      "step": 18243
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.200182857142858e-06,
+      "loss": 1.3033,
+      "step": 18244
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.199914285714286e-06,
+      "loss": 1.2262,
+      "step": 18245
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.199645714285714e-06,
+      "loss": 1.2699,
+      "step": 18246
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.199377142857144e-06,
+      "loss": 1.3374,
+      "step": 18247
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.19910857142857e-06,
+      "loss": 0.9182,
+      "step": 18248
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.19884e-06,
+      "loss": 1.2507,
+      "step": 18249
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.198571428571428e-06,
+      "loss": 1.0254,
+      "step": 18250
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.198302857142858e-06,
+      "loss": 0.9017,
+      "step": 18251
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.198034285714286e-06,
+      "loss": 1.5631,
+      "step": 18252
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.197765714285715e-06,
+      "loss": 1.3421,
+      "step": 18253
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.197497142857144e-06,
+      "loss": 1.0341,
+      "step": 18254
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.197228571428571e-06,
+      "loss": 0.9692,
+      "step": 18255
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.19696e-06,
+      "loss": 0.8676,
+      "step": 18256
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.196691428571429e-06,
+      "loss": 1.0563,
+      "step": 18257
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.196422857142857e-06,
+      "loss": 1.0378,
+      "step": 18258
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.196154285714285e-06,
+      "loss": 1.4805,
+      "step": 18259
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.195885714285715e-06,
+      "loss": 0.7793,
+      "step": 18260
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.195617142857143e-06,
+      "loss": 1.4254,
+      "step": 18261
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.195348571428572e-06,
+      "loss": 1.3085,
+      "step": 18262
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.19508e-06,
+      "loss": 0.9271,
+      "step": 18263
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.19481142857143e-06,
+      "loss": 1.1874,
+      "step": 18264
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.194542857142858e-06,
+      "loss": 1.0718,
+      "step": 18265
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.194274285714286e-06,
+      "loss": 1.0855,
+      "step": 18266
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.194005714285716e-06,
+      "loss": 0.991,
+      "step": 18267
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.193737142857142e-06,
+      "loss": 1.3098,
+      "step": 18268
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.193468571428572e-06,
+      "loss": 0.9124,
+      "step": 18269
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.1932e-06,
+      "loss": 1.647,
+      "step": 18270
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.192931428571429e-06,
+      "loss": 1.1939,
+      "step": 18271
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.192662857142857e-06,
+      "loss": 1.2216,
+      "step": 18272
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.192394285714287e-06,
+      "loss": 1.1616,
+      "step": 18273
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.192125714285715e-06,
+      "loss": 1.4807,
+      "step": 18274
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.191857142857143e-06,
+      "loss": 1.201,
+      "step": 18275
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.191588571428571e-06,
+      "loss": 1.2005,
+      "step": 18276
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.191320000000001e-06,
+      "loss": 1.1734,
+      "step": 18277
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.191051428571429e-06,
+      "loss": 1.0926,
+      "step": 18278
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.190782857142857e-06,
+      "loss": 1.295,
+      "step": 18279
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.190514285714287e-06,
+      "loss": 0.9326,
+      "step": 18280
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.190245714285714e-06,
+      "loss": 1.2304,
+      "step": 18281
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.189977142857144e-06,
+      "loss": 1.7546,
+      "step": 18282
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.189708571428572e-06,
+      "loss": 1.088,
+      "step": 18283
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.189440000000002e-06,
+      "loss": 1.2642,
+      "step": 18284
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.189171428571428e-06,
+      "loss": 1.4828,
+      "step": 18285
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.188902857142858e-06,
+      "loss": 1.2799,
+      "step": 18286
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.188634285714286e-06,
+      "loss": 1.2869,
+      "step": 18287
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.188365714285714e-06,
+      "loss": 1.4238,
+      "step": 18288
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.188097142857142e-06,
+      "loss": 0.8151,
+      "step": 18289
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.187828571428572e-06,
+      "loss": 1.1607,
+      "step": 18290
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.18756e-06,
+      "loss": 1.0475,
+      "step": 18291
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.187291428571429e-06,
+      "loss": 1.4333,
+      "step": 18292
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.187022857142859e-06,
+      "loss": 1.6807,
+      "step": 18293
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.186754285714285e-06,
+      "loss": 1.0874,
+      "step": 18294
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.186485714285715e-06,
+      "loss": 1.3178,
+      "step": 18295
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.186217142857143e-06,
+      "loss": 1.2952,
+      "step": 18296
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.185948571428573e-06,
+      "loss": 1.6318,
+      "step": 18297
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.18568e-06,
+      "loss": 0.8782,
+      "step": 18298
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.18541142857143e-06,
+      "loss": 0.8615,
+      "step": 18299
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.185142857142857e-06,
+      "loss": 1.0304,
+      "step": 18300
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.184874285714286e-06,
+      "loss": 1.332,
+      "step": 18301
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.184605714285714e-06,
+      "loss": 1.3237,
+      "step": 18302
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.184337142857144e-06,
+      "loss": 0.9689,
+      "step": 18303
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.184068571428572e-06,
+      "loss": 1.1978,
+      "step": 18304
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.1838e-06,
+      "loss": 1.2155,
+      "step": 18305
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.18353142857143e-06,
+      "loss": 1.0266,
+      "step": 18306
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.183262857142858e-06,
+      "loss": 1.2905,
+      "step": 18307
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.182994285714286e-06,
+      "loss": 1.1426,
+      "step": 18308
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.182725714285714e-06,
+      "loss": 1.0718,
+      "step": 18309
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.182457142857144e-06,
+      "loss": 1.0164,
+      "step": 18310
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.18218857142857e-06,
+      "loss": 1.1174,
+      "step": 18311
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.18192e-06,
+      "loss": 1.217,
+      "step": 18312
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.181651428571429e-06,
+      "loss": 0.9425,
+      "step": 18313
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.181382857142857e-06,
+      "loss": 1.381,
+      "step": 18314
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.181114285714285e-06,
+      "loss": 1.4069,
+      "step": 18315
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.180845714285715e-06,
+      "loss": 1.0243,
+      "step": 18316
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.180577142857143e-06,
+      "loss": 1.4335,
+      "step": 18317
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.180308571428571e-06,
+      "loss": 1.4455,
+      "step": 18318
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.180040000000001e-06,
+      "loss": 0.9981,
+      "step": 18319
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.17977142857143e-06,
+      "loss": 1.201,
+      "step": 18320
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.179502857142858e-06,
+      "loss": 1.3252,
+      "step": 18321
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.179234285714286e-06,
+      "loss": 0.9183,
+      "step": 18322
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.178965714285716e-06,
+      "loss": 1.301,
+      "step": 18323
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.178697142857142e-06,
+      "loss": 1.125,
+      "step": 18324
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.178428571428572e-06,
+      "loss": 1.2771,
+      "step": 18325
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.17816e-06,
+      "loss": 1.0332,
+      "step": 18326
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.177891428571428e-06,
+      "loss": 1.361,
+      "step": 18327
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.177622857142857e-06,
+      "loss": 1.3151,
+      "step": 18328
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.177354285714286e-06,
+      "loss": 1.4116,
+      "step": 18329
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.177085714285715e-06,
+      "loss": 1.005,
+      "step": 18330
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.176817142857143e-06,
+      "loss": 1.0982,
+      "step": 18331
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.176548571428573e-06,
+      "loss": 1.0457,
+      "step": 18332
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.176280000000001e-06,
+      "loss": 1.1619,
+      "step": 18333
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.176011428571429e-06,
+      "loss": 1.0186,
+      "step": 18334
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.175742857142857e-06,
+      "loss": 1.1658,
+      "step": 18335
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.175474285714287e-06,
+      "loss": 1.1407,
+      "step": 18336
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.175205714285714e-06,
+      "loss": 1.3305,
+      "step": 18337
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.174937142857143e-06,
+      "loss": 1.4422,
+      "step": 18338
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.174668571428572e-06,
+      "loss": 0.9494,
+      "step": 18339
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.1744e-06,
+      "loss": 1.3172,
+      "step": 18340
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.174131428571428e-06,
+      "loss": 1.0955,
+      "step": 18341
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.173862857142858e-06,
+      "loss": 1.3985,
+      "step": 18342
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.173594285714286e-06,
+      "loss": 1.6049,
+      "step": 18343
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.173325714285714e-06,
+      "loss": 0.9859,
+      "step": 18344
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.173057142857144e-06,
+      "loss": 1.2749,
+      "step": 18345
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.172788571428572e-06,
+      "loss": 0.8264,
+      "step": 18346
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.17252e-06,
+      "loss": 1.179,
+      "step": 18347
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.172251428571429e-06,
+      "loss": 0.8146,
+      "step": 18348
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.171982857142858e-06,
+      "loss": 1.5421,
+      "step": 18349
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.171714285714285e-06,
+      "loss": 1.3544,
+      "step": 18350
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.171445714285715e-06,
+      "loss": 1.3417,
+      "step": 18351
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.171177142857143e-06,
+      "loss": 1.5423,
+      "step": 18352
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.170908571428573e-06,
+      "loss": 1.3648,
+      "step": 18353
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.17064e-06,
+      "loss": 1.3552,
+      "step": 18354
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.17037142857143e-06,
+      "loss": 1.7231,
+      "step": 18355
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.170102857142857e-06,
+      "loss": 1.2673,
+      "step": 18356
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.169834285714286e-06,
+      "loss": 1.2709,
+      "step": 18357
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.169565714285715e-06,
+      "loss": 1.3457,
+      "step": 18358
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.169297142857144e-06,
+      "loss": 0.9952,
+      "step": 18359
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.169028571428572e-06,
+      "loss": 1.2073,
+      "step": 18360
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.16876e-06,
+      "loss": 0.8153,
+      "step": 18361
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.16849142857143e-06,
+      "loss": 0.8797,
+      "step": 18362
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.168222857142856e-06,
+      "loss": 0.9036,
+      "step": 18363
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.167954285714286e-06,
+      "loss": 1.0906,
+      "step": 18364
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.167685714285714e-06,
+      "loss": 0.9483,
+      "step": 18365
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.167417142857144e-06,
+      "loss": 1.0935,
+      "step": 18366
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.16714857142857e-06,
+      "loss": 1.1538,
+      "step": 18367
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.16688e-06,
+      "loss": 1.0366,
+      "step": 18368
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.166611428571429e-06,
+      "loss": 1.0904,
+      "step": 18369
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.166342857142857e-06,
+      "loss": 1.473,
+      "step": 18370
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.166074285714287e-06,
+      "loss": 0.821,
+      "step": 18371
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.165805714285715e-06,
+      "loss": 1.3653,
+      "step": 18372
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.165537142857143e-06,
+      "loss": 1.2806,
+      "step": 18373
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.165268571428571e-06,
+      "loss": 1.0286,
+      "step": 18374
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.165000000000001e-06,
+      "loss": 1.5693,
+      "step": 18375
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.164731428571428e-06,
+      "loss": 0.9937,
+      "step": 18376
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.164462857142858e-06,
+      "loss": 1.1274,
+      "step": 18377
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.164194285714286e-06,
+      "loss": 0.9855,
+      "step": 18378
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.163925714285716e-06,
+      "loss": 1.0235,
+      "step": 18379
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.163657142857142e-06,
+      "loss": 1.0453,
+      "step": 18380
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.163388571428572e-06,
+      "loss": 1.3311,
+      "step": 18381
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.16312e-06,
+      "loss": 1.0697,
+      "step": 18382
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.162851428571428e-06,
+      "loss": 1.5447,
+      "step": 18383
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.162582857142858e-06,
+      "loss": 1.0644,
+      "step": 18384
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.162314285714286e-06,
+      "loss": 1.1564,
+      "step": 18385
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.162045714285715e-06,
+      "loss": 1.3385,
+      "step": 18386
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.161777142857143e-06,
+      "loss": 0.8739,
+      "step": 18387
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.161508571428573e-06,
+      "loss": 1.2327,
+      "step": 18388
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.16124e-06,
+      "loss": 0.9946,
+      "step": 18389
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.160971428571429e-06,
+      "loss": 1.1651,
+      "step": 18390
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.160702857142857e-06,
+      "loss": 1.0831,
+      "step": 18391
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.160434285714287e-06,
+      "loss": 1.4471,
+      "step": 18392
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.160165714285713e-06,
+      "loss": 0.9151,
+      "step": 18393
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.159897142857143e-06,
+      "loss": 0.9345,
+      "step": 18394
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.159628571428571e-06,
+      "loss": 1.4516,
+      "step": 18395
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.15936e-06,
+      "loss": 1.1488,
+      "step": 18396
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.15909142857143e-06,
+      "loss": 1.0136,
+      "step": 18397
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.158822857142858e-06,
+      "loss": 1.6719,
+      "step": 18398
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.158554285714286e-06,
+      "loss": 1.3568,
+      "step": 18399
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.158285714285714e-06,
+      "loss": 1.3059,
+      "step": 18400
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.158017142857144e-06,
+      "loss": 1.1464,
+      "step": 18401
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.157748571428572e-06,
+      "loss": 1.3551,
+      "step": 18402
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.15748e-06,
+      "loss": 1.414,
+      "step": 18403
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.157211428571428e-06,
+      "loss": 1.0285,
+      "step": 18404
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.156942857142858e-06,
+      "loss": 1.3004,
+      "step": 18405
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.156674285714285e-06,
+      "loss": 1.0819,
+      "step": 18406
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.156405714285715e-06,
+      "loss": 1.1322,
+      "step": 18407
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.156137142857143e-06,
+      "loss": 1.0188,
+      "step": 18408
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.155868571428571e-06,
+      "loss": 1.3002,
+      "step": 18409
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.155600000000001e-06,
+      "loss": 0.7202,
+      "step": 18410
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.155331428571429e-06,
+      "loss": 1.0242,
+      "step": 18411
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.155062857142859e-06,
+      "loss": 0.8348,
+      "step": 18412
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.154794285714285e-06,
+      "loss": 1.1839,
+      "step": 18413
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.154525714285715e-06,
+      "loss": 1.1911,
+      "step": 18414
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.154257142857143e-06,
+      "loss": 1.1965,
+      "step": 18415
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.153988571428572e-06,
+      "loss": 1.6591,
+      "step": 18416
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.15372e-06,
+      "loss": 0.8326,
+      "step": 18417
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.15345142857143e-06,
+      "loss": 1.3134,
+      "step": 18418
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.153182857142856e-06,
+      "loss": 0.9078,
+      "step": 18419
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.152914285714286e-06,
+      "loss": 1.0474,
+      "step": 18420
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.152645714285714e-06,
+      "loss": 0.8752,
+      "step": 18421
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.152377142857142e-06,
+      "loss": 1.2492,
+      "step": 18422
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.152108571428572e-06,
+      "loss": 1.7221,
+      "step": 18423
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.15184e-06,
+      "loss": 1.0772,
+      "step": 18424
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.15157142857143e-06,
+      "loss": 1.332,
+      "step": 18425
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.151302857142857e-06,
+      "loss": 1.1384,
+      "step": 18426
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.151034285714287e-06,
+      "loss": 1.0454,
+      "step": 18427
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.150765714285715e-06,
+      "loss": 1.1541,
+      "step": 18428
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.150497142857143e-06,
+      "loss": 0.7839,
+      "step": 18429
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.150228571428571e-06,
+      "loss": 1.006,
+      "step": 18430
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.149960000000001e-06,
+      "loss": 1.1489,
+      "step": 18431
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.149691428571428e-06,
+      "loss": 1.2466,
+      "step": 18432
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.149422857142857e-06,
+      "loss": 1.3458,
+      "step": 18433
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.149154285714286e-06,
+      "loss": 1.1481,
+      "step": 18434
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.148885714285715e-06,
+      "loss": 1.1732,
+      "step": 18435
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.148617142857144e-06,
+      "loss": 1.2531,
+      "step": 18436
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.148348571428572e-06,
+      "loss": 0.9427,
+      "step": 18437
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.148080000000002e-06,
+      "loss": 1.2455,
+      "step": 18438
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.147811428571428e-06,
+      "loss": 1.4132,
+      "step": 18439
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.147542857142858e-06,
+      "loss": 1.2458,
+      "step": 18440
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.147274285714286e-06,
+      "loss": 0.9515,
+      "step": 18441
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.147005714285714e-06,
+      "loss": 1.1261,
+      "step": 18442
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.146737142857143e-06,
+      "loss": 1.2654,
+      "step": 18443
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.146468571428572e-06,
+      "loss": 1.168,
+      "step": 18444
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.146199999999999e-06,
+      "loss": 1.2526,
+      "step": 18445
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.145931428571429e-06,
+      "loss": 1.0944,
+      "step": 18446
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.145662857142857e-06,
+      "loss": 1.4304,
+      "step": 18447
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.145394285714287e-06,
+      "loss": 1.3271,
+      "step": 18448
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.145125714285715e-06,
+      "loss": 1.3958,
+      "step": 18449
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.144857142857143e-06,
+      "loss": 1.0206,
+      "step": 18450
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.144588571428573e-06,
+      "loss": 0.9709,
+      "step": 18451
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.14432e-06,
+      "loss": 1.5197,
+      "step": 18452
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.14405142857143e-06,
+      "loss": 1.2183,
+      "step": 18453
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.143782857142858e-06,
+      "loss": 1.0092,
+      "step": 18454
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.143514285714286e-06,
+      "loss": 0.9506,
+      "step": 18455
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.143245714285714e-06,
+      "loss": 1.2621,
+      "step": 18456
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.142977142857144e-06,
+      "loss": 1.4713,
+      "step": 18457
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.142708571428572e-06,
+      "loss": 1.0833,
+      "step": 18458
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.14244e-06,
+      "loss": 1.2121,
+      "step": 18459
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.142171428571428e-06,
+      "loss": 1.1769,
+      "step": 18460
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.141902857142858e-06,
+      "loss": 1.2137,
+      "step": 18461
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.141634285714286e-06,
+      "loss": 0.9813,
+      "step": 18462
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.141365714285715e-06,
+      "loss": 1.2344,
+      "step": 18463
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.141097142857144e-06,
+      "loss": 1.1514,
+      "step": 18464
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.140828571428571e-06,
+      "loss": 0.7275,
+      "step": 18465
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.14056e-06,
+      "loss": 0.944,
+      "step": 18466
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.140291428571429e-06,
+      "loss": 1.472,
+      "step": 18467
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.140022857142857e-06,
+      "loss": 1.4456,
+      "step": 18468
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.139754285714285e-06,
+      "loss": 1.2428,
+      "step": 18469
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.139485714285715e-06,
+      "loss": 1.2793,
+      "step": 18470
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.139217142857143e-06,
+      "loss": 1.1342,
+      "step": 18471
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.138948571428572e-06,
+      "loss": 1.4156,
+      "step": 18472
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.13868e-06,
+      "loss": 1.3192,
+      "step": 18473
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.13841142857143e-06,
+      "loss": 1.2177,
+      "step": 18474
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.138142857142858e-06,
+      "loss": 1.0995,
+      "step": 18475
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.137874285714286e-06,
+      "loss": 1.0547,
+      "step": 18476
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.137605714285716e-06,
+      "loss": 1.1232,
+      "step": 18477
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.137337142857142e-06,
+      "loss": 1.0118,
+      "step": 18478
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.137068571428572e-06,
+      "loss": 1.2615,
+      "step": 18479
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.1368e-06,
+      "loss": 1.0926,
+      "step": 18480
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.13653142857143e-06,
+      "loss": 0.852,
+      "step": 18481
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.136262857142857e-06,
+      "loss": 1.1767,
+      "step": 18482
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.135994285714287e-06,
+      "loss": 1.3474,
+      "step": 18483
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.135725714285715e-06,
+      "loss": 1.508,
+      "step": 18484
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.135457142857143e-06,
+      "loss": 1.4427,
+      "step": 18485
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.135188571428571e-06,
+      "loss": 1.618,
+      "step": 18486
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.134920000000001e-06,
+      "loss": 1.3714,
+      "step": 18487
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.134651428571429e-06,
+      "loss": 1.0204,
+      "step": 18488
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.134382857142857e-06,
+      "loss": 1.2192,
+      "step": 18489
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.134114285714287e-06,
+      "loss": 1.4169,
+      "step": 18490
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.133845714285714e-06,
+      "loss": 1.4333,
+      "step": 18491
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.133577142857144e-06,
+      "loss": 1.2324,
+      "step": 18492
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.133308571428572e-06,
+      "loss": 0.9643,
+      "step": 18493
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.133040000000002e-06,
+      "loss": 0.9955,
+      "step": 18494
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.132771428571428e-06,
+      "loss": 1.6009,
+      "step": 18495
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.132502857142858e-06,
+      "loss": 1.0956,
+      "step": 18496
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.132234285714286e-06,
+      "loss": 0.906,
+      "step": 18497
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.131965714285714e-06,
+      "loss": 1.0136,
+      "step": 18498
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.131697142857142e-06,
+      "loss": 1.5391,
+      "step": 18499
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.131428571428572e-06,
+      "loss": 1.6627,
+      "step": 18500
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.13116e-06,
+      "loss": 0.8936,
+      "step": 18501
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.130891428571429e-06,
+      "loss": 0.882,
+      "step": 18502
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.130622857142859e-06,
+      "loss": 1.3404,
+      "step": 18503
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.130354285714287e-06,
+      "loss": 1.3588,
+      "step": 18504
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.130085714285715e-06,
+      "loss": 1.0248,
+      "step": 18505
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.129817142857143e-06,
+      "loss": 1.191,
+      "step": 18506
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.129548571428573e-06,
+      "loss": 1.0155,
+      "step": 18507
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.12928e-06,
+      "loss": 1.2957,
+      "step": 18508
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.12901142857143e-06,
+      "loss": 1.1311,
+      "step": 18509
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.128742857142857e-06,
+      "loss": 0.8517,
+      "step": 18510
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.128474285714286e-06,
+      "loss": 1.1234,
+      "step": 18511
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.128205714285714e-06,
+      "loss": 1.5501,
+      "step": 18512
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.127937142857144e-06,
+      "loss": 1.0707,
+      "step": 18513
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.127668571428572e-06,
+      "loss": 0.9958,
+      "step": 18514
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.1274e-06,
+      "loss": 1.2294,
+      "step": 18515
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.12713142857143e-06,
+      "loss": 1.1665,
+      "step": 18516
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.126862857142858e-06,
+      "loss": 1.1331,
+      "step": 18517
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.126594285714286e-06,
+      "loss": 0.9763,
+      "step": 18518
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.126325714285714e-06,
+      "loss": 1.2737,
+      "step": 18519
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.126057142857144e-06,
+      "loss": 1.4391,
+      "step": 18520
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.12578857142857e-06,
+      "loss": 1.327,
+      "step": 18521
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.12552e-06,
+      "loss": 1.5298,
+      "step": 18522
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.125251428571429e-06,
+      "loss": 1.4681,
+      "step": 18523
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.124982857142857e-06,
+      "loss": 1.3782,
+      "step": 18524
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.124714285714285e-06,
+      "loss": 1.1317,
+      "step": 18525
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.124445714285715e-06,
+      "loss": 0.8639,
+      "step": 18526
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.124177142857143e-06,
+      "loss": 0.807,
+      "step": 18527
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.123908571428571e-06,
+      "loss": 1.5235,
+      "step": 18528
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.123640000000001e-06,
+      "loss": 1.5215,
+      "step": 18529
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.12337142857143e-06,
+      "loss": 0.9546,
+      "step": 18530
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.123102857142858e-06,
+      "loss": 1.1681,
+      "step": 18531
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.122834285714286e-06,
+      "loss": 1.2994,
+      "step": 18532
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.122565714285716e-06,
+      "loss": 0.7852,
+      "step": 18533
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.122297142857142e-06,
+      "loss": 1.4301,
+      "step": 18534
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.122028571428572e-06,
+      "loss": 1.0574,
+      "step": 18535
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.12176e-06,
+      "loss": 1.1146,
+      "step": 18536
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.121491428571428e-06,
+      "loss": 1.0763,
+      "step": 18537
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.121222857142857e-06,
+      "loss": 1.6425,
+      "step": 18538
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.120954285714286e-06,
+      "loss": 1.2382,
+      "step": 18539
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.120685714285715e-06,
+      "loss": 1.2954,
+      "step": 18540
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.120417142857143e-06,
+      "loss": 1.187,
+      "step": 18541
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.120148571428573e-06,
+      "loss": 1.3082,
+      "step": 18542
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.119880000000001e-06,
+      "loss": 0.8139,
+      "step": 18543
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.119611428571429e-06,
+      "loss": 1.3284,
+      "step": 18544
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.119342857142857e-06,
+      "loss": 1.1029,
+      "step": 18545
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.119074285714287e-06,
+      "loss": 1.2641,
+      "step": 18546
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.118805714285714e-06,
+      "loss": 1.2782,
+      "step": 18547
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.118537142857143e-06,
+      "loss": 1.1325,
+      "step": 18548
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.118268571428572e-06,
+      "loss": 1.2463,
+      "step": 18549
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.118e-06,
+      "loss": 1.4697,
+      "step": 18550
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.117731428571428e-06,
+      "loss": 1.4785,
+      "step": 18551
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.117462857142858e-06,
+      "loss": 0.8045,
+      "step": 18552
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.117194285714286e-06,
+      "loss": 1.3623,
+      "step": 18553
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.116925714285714e-06,
+      "loss": 1.09,
+      "step": 18554
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.116657142857144e-06,
+      "loss": 0.9731,
+      "step": 18555
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.116388571428572e-06,
+      "loss": 1.4581,
+      "step": 18556
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.11612e-06,
+      "loss": 0.8459,
+      "step": 18557
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.115851428571429e-06,
+      "loss": 0.8089,
+      "step": 18558
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.115582857142858e-06,
+      "loss": 1.3463,
+      "step": 18559
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.115314285714285e-06,
+      "loss": 1.1083,
+      "step": 18560
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.115045714285715e-06,
+      "loss": 1.299,
+      "step": 18561
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.114777142857143e-06,
+      "loss": 1.211,
+      "step": 18562
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.114508571428573e-06,
+      "loss": 1.4899,
+      "step": 18563
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.11424e-06,
+      "loss": 0.9624,
+      "step": 18564
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.11397142857143e-06,
+      "loss": 1.3035,
+      "step": 18565
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.113702857142857e-06,
+      "loss": 1.3853,
+      "step": 18566
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.113434285714286e-06,
+      "loss": 0.8839,
+      "step": 18567
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.113165714285715e-06,
+      "loss": 1.2796,
+      "step": 18568
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.112897142857144e-06,
+      "loss": 1.3946,
+      "step": 18569
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.112628571428572e-06,
+      "loss": 1.2919,
+      "step": 18570
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.11236e-06,
+      "loss": 1.2398,
+      "step": 18571
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.11209142857143e-06,
+      "loss": 0.8939,
+      "step": 18572
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.111822857142856e-06,
+      "loss": 1.0883,
+      "step": 18573
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.111554285714286e-06,
+      "loss": 0.7882,
+      "step": 18574
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.111285714285714e-06,
+      "loss": 0.7469,
+      "step": 18575
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.111017142857144e-06,
+      "loss": 1.2594,
+      "step": 18576
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.11074857142857e-06,
+      "loss": 1.1103,
+      "step": 18577
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.11048e-06,
+      "loss": 1.1176,
+      "step": 18578
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.110211428571429e-06,
+      "loss": 1.4547,
+      "step": 18579
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.109942857142857e-06,
+      "loss": 0.9132,
+      "step": 18580
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.109674285714287e-06,
+      "loss": 1.0824,
+      "step": 18581
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.109405714285715e-06,
+      "loss": 1.1009,
+      "step": 18582
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.109137142857143e-06,
+      "loss": 1.5897,
+      "step": 18583
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.108868571428571e-06,
+      "loss": 1.0647,
+      "step": 18584
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.108600000000001e-06,
+      "loss": 1.599,
+      "step": 18585
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.10833142857143e-06,
+      "loss": 1.1644,
+      "step": 18586
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.108062857142858e-06,
+      "loss": 0.8117,
+      "step": 18587
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.107794285714286e-06,
+      "loss": 1.3737,
+      "step": 18588
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.107525714285716e-06,
+      "loss": 1.3997,
+      "step": 18589
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.107257142857142e-06,
+      "loss": 1.0261,
+      "step": 18590
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.106988571428572e-06,
+      "loss": 1.2342,
+      "step": 18591
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.10672e-06,
+      "loss": 0.7943,
+      "step": 18592
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.106451428571428e-06,
+      "loss": 1.1924,
+      "step": 18593
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.106182857142858e-06,
+      "loss": 1.5941,
+      "step": 18594
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.105914285714286e-06,
+      "loss": 1.384,
+      "step": 18595
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.105645714285715e-06,
+      "loss": 1.3066,
+      "step": 18596
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.105377142857143e-06,
+      "loss": 1.2349,
+      "step": 18597
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.105108571428573e-06,
+      "loss": 1.2675,
+      "step": 18598
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.10484e-06,
+      "loss": 1.0974,
+      "step": 18599
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.104571428571429e-06,
+      "loss": 1.0898,
+      "step": 18600
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.104302857142857e-06,
+      "loss": 1.1187,
+      "step": 18601
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.104034285714287e-06,
+      "loss": 1.2313,
+      "step": 18602
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.103765714285713e-06,
+      "loss": 1.1595,
+      "step": 18603
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.103497142857143e-06,
+      "loss": 0.88,
+      "step": 18604
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.103228571428571e-06,
+      "loss": 1.024,
+      "step": 18605
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.10296e-06,
+      "loss": 1.3314,
+      "step": 18606
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.10269142857143e-06,
+      "loss": 1.1775,
+      "step": 18607
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.102422857142858e-06,
+      "loss": 1.3327,
+      "step": 18608
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.102154285714288e-06,
+      "loss": 1.5813,
+      "step": 18609
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.101885714285714e-06,
+      "loss": 1.0929,
+      "step": 18610
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.101617142857144e-06,
+      "loss": 1.0808,
+      "step": 18611
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.101348571428572e-06,
+      "loss": 0.9521,
+      "step": 18612
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.10108e-06,
+      "loss": 0.8539,
+      "step": 18613
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.100811428571428e-06,
+      "loss": 1.1798,
+      "step": 18614
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.100542857142858e-06,
+      "loss": 1.3468,
+      "step": 18615
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.100274285714285e-06,
+      "loss": 0.8499,
+      "step": 18616
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.100005714285715e-06,
+      "loss": 1.3575,
+      "step": 18617
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.099737142857143e-06,
+      "loss": 1.3034,
+      "step": 18618
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.099468571428571e-06,
+      "loss": 0.715,
+      "step": 18619
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.099200000000001e-06,
+      "loss": 1.0368,
+      "step": 18620
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.098931428571429e-06,
+      "loss": 1.11,
+      "step": 18621
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.098662857142859e-06,
+      "loss": 0.946,
+      "step": 18622
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.098394285714285e-06,
+      "loss": 0.992,
+      "step": 18623
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.098125714285715e-06,
+      "loss": 1.5636,
+      "step": 18624
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.097857142857143e-06,
+      "loss": 1.2106,
+      "step": 18625
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.097588571428572e-06,
+      "loss": 1.2869,
+      "step": 18626
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.09732e-06,
+      "loss": 1.2957,
+      "step": 18627
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.09705142857143e-06,
+      "loss": 1.1688,
+      "step": 18628
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.096782857142856e-06,
+      "loss": 1.1658,
+      "step": 18629
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.096514285714286e-06,
+      "loss": 1.2495,
+      "step": 18630
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.096245714285714e-06,
+      "loss": 1.2162,
+      "step": 18631
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.095977142857144e-06,
+      "loss": 1.0217,
+      "step": 18632
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.095708571428572e-06,
+      "loss": 0.974,
+      "step": 18633
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.09544e-06,
+      "loss": 1.3299,
+      "step": 18634
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.09517142857143e-06,
+      "loss": 1.2632,
+      "step": 18635
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.094902857142857e-06,
+      "loss": 1.6056,
+      "step": 18636
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.094634285714287e-06,
+      "loss": 1.3571,
+      "step": 18637
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.094365714285715e-06,
+      "loss": 1.2441,
+      "step": 18638
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.094097142857143e-06,
+      "loss": 1.4651,
+      "step": 18639
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.093828571428571e-06,
+      "loss": 1.2285,
+      "step": 18640
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.093560000000001e-06,
+      "loss": 1.1991,
+      "step": 18641
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.093291428571428e-06,
+      "loss": 1.2084,
+      "step": 18642
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.093022857142857e-06,
+      "loss": 1.0287,
+      "step": 18643
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.092754285714286e-06,
+      "loss": 0.8697,
+      "step": 18644
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.092485714285715e-06,
+      "loss": 1.3745,
+      "step": 18645
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.092217142857144e-06,
+      "loss": 1.2758,
+      "step": 18646
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.091948571428572e-06,
+      "loss": 0.9661,
+      "step": 18647
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.091680000000002e-06,
+      "loss": 1.1758,
+      "step": 18648
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.091411428571428e-06,
+      "loss": 1.0865,
+      "step": 18649
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.091142857142858e-06,
+      "loss": 1.173,
+      "step": 18650
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.090874285714286e-06,
+      "loss": 1.1453,
+      "step": 18651
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.090605714285714e-06,
+      "loss": 1.247,
+      "step": 18652
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.090337142857143e-06,
+      "loss": 1.3232,
+      "step": 18653
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.090068571428572e-06,
+      "loss": 1.2417,
+      "step": 18654
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.0898e-06,
+      "loss": 0.917,
+      "step": 18655
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.089531428571429e-06,
+      "loss": 0.9836,
+      "step": 18656
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.089262857142857e-06,
+      "loss": 1.436,
+      "step": 18657
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.088994285714287e-06,
+      "loss": 1.784,
+      "step": 18658
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.088725714285713e-06,
+      "loss": 0.7697,
+      "step": 18659
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.088457142857143e-06,
+      "loss": 1.4879,
+      "step": 18660
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.088188571428573e-06,
+      "loss": 1.2704,
+      "step": 18661
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.08792e-06,
+      "loss": 0.9966,
+      "step": 18662
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.08765142857143e-06,
+      "loss": 1.4211,
+      "step": 18663
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.087382857142858e-06,
+      "loss": 0.8287,
+      "step": 18664
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.087114285714286e-06,
+      "loss": 1.3873,
+      "step": 18665
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.086845714285714e-06,
+      "loss": 1.0419,
+      "step": 18666
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.086577142857144e-06,
+      "loss": 1.0356,
+      "step": 18667
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.086308571428572e-06,
+      "loss": 1.145,
+      "step": 18668
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.08604e-06,
+      "loss": 0.896,
+      "step": 18669
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.085771428571428e-06,
+      "loss": 1.0077,
+      "step": 18670
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.085502857142858e-06,
+      "loss": 0.9901,
+      "step": 18671
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.085234285714285e-06,
+      "loss": 0.9741,
+      "step": 18672
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.084965714285715e-06,
+      "loss": 1.566,
+      "step": 18673
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.084697142857144e-06,
+      "loss": 1.3967,
+      "step": 18674
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.084428571428571e-06,
+      "loss": 1.0902,
+      "step": 18675
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.08416e-06,
+      "loss": 1.2105,
+      "step": 18676
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.083891428571429e-06,
+      "loss": 0.9511,
+      "step": 18677
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.083622857142859e-06,
+      "loss": 1.0215,
+      "step": 18678
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.083354285714285e-06,
+      "loss": 1.7989,
+      "step": 18679
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.083085714285715e-06,
+      "loss": 1.2225,
+      "step": 18680
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.082817142857143e-06,
+      "loss": 0.8694,
+      "step": 18681
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.082548571428572e-06,
+      "loss": 1.0614,
+      "step": 18682
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.08228e-06,
+      "loss": 1.0876,
+      "step": 18683
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.08201142857143e-06,
+      "loss": 0.9651,
+      "step": 18684
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.081742857142856e-06,
+      "loss": 1.0319,
+      "step": 18685
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.081474285714286e-06,
+      "loss": 0.8718,
+      "step": 18686
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.081205714285716e-06,
+      "loss": 1.1122,
+      "step": 18687
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.080937142857142e-06,
+      "loss": 1.6745,
+      "step": 18688
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.080668571428572e-06,
+      "loss": 1.0967,
+      "step": 18689
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.0804e-06,
+      "loss": 1.1138,
+      "step": 18690
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.08013142857143e-06,
+      "loss": 1.29,
+      "step": 18691
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.079862857142857e-06,
+      "loss": 1.0604,
+      "step": 18692
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.079594285714287e-06,
+      "loss": 1.5419,
+      "step": 18693
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.079325714285715e-06,
+      "loss": 1.355,
+      "step": 18694
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.079057142857143e-06,
+      "loss": 1.5669,
+      "step": 18695
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.078788571428571e-06,
+      "loss": 1.2451,
+      "step": 18696
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.078520000000001e-06,
+      "loss": 0.927,
+      "step": 18697
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.078251428571427e-06,
+      "loss": 1.1476,
+      "step": 18698
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.077982857142857e-06,
+      "loss": 1.303,
+      "step": 18699
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.077714285714287e-06,
+      "loss": 1.3933,
+      "step": 18700
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.077445714285714e-06,
+      "loss": 1.4504,
+      "step": 18701
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.077177142857144e-06,
+      "loss": 1.0779,
+      "step": 18702
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.076908571428572e-06,
+      "loss": 1.213,
+      "step": 18703
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.076640000000002e-06,
+      "loss": 1.4016,
+      "step": 18704
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.076371428571428e-06,
+      "loss": 1.3491,
+      "step": 18705
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.076102857142858e-06,
+      "loss": 1.1121,
+      "step": 18706
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.075834285714286e-06,
+      "loss": 1.4392,
+      "step": 18707
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.075565714285714e-06,
+      "loss": 1.4397,
+      "step": 18708
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.075297142857142e-06,
+      "loss": 1.0552,
+      "step": 18709
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.075028571428572e-06,
+      "loss": 1.043,
+      "step": 18710
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.074759999999999e-06,
+      "loss": 1.1216,
+      "step": 18711
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.074491428571429e-06,
+      "loss": 0.954,
+      "step": 18712
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.074222857142859e-06,
+      "loss": 1.3876,
+      "step": 18713
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.073954285714287e-06,
+      "loss": 0.8726,
+      "step": 18714
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.073685714285715e-06,
+      "loss": 1.3342,
+      "step": 18715
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.073417142857143e-06,
+      "loss": 1.0397,
+      "step": 18716
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.073148571428573e-06,
+      "loss": 1.2309,
+      "step": 18717
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.07288e-06,
+      "loss": 1.1509,
+      "step": 18718
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.07261142857143e-06,
+      "loss": 1.3983,
+      "step": 18719
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.072342857142857e-06,
+      "loss": 1.2845,
+      "step": 18720
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.072074285714286e-06,
+      "loss": 1.2113,
+      "step": 18721
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.071805714285714e-06,
+      "loss": 1.0131,
+      "step": 18722
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.071537142857144e-06,
+      "loss": 1.064,
+      "step": 18723
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.07126857142857e-06,
+      "loss": 1.5317,
+      "step": 18724
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.071e-06,
+      "loss": 1.1137,
+      "step": 18725
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.07073142857143e-06,
+      "loss": 1.1974,
+      "step": 18726
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.070462857142858e-06,
+      "loss": 1.0476,
+      "step": 18727
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.070194285714286e-06,
+      "loss": 1.1801,
+      "step": 18728
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.069925714285714e-06,
+      "loss": 0.7809,
+      "step": 18729
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.069657142857144e-06,
+      "loss": 1.1744,
+      "step": 18730
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.06938857142857e-06,
+      "loss": 0.8775,
+      "step": 18731
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.06912e-06,
+      "loss": 1.3158,
+      "step": 18732
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.068851428571429e-06,
+      "loss": 1.2004,
+      "step": 18733
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.068582857142857e-06,
+      "loss": 1.371,
+      "step": 18734
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.068314285714285e-06,
+      "loss": 1.2006,
+      "step": 18735
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.068045714285715e-06,
+      "loss": 1.4059,
+      "step": 18736
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.067777142857143e-06,
+      "loss": 1.156,
+      "step": 18737
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.067508571428571e-06,
+      "loss": 1.4547,
+      "step": 18738
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.067240000000001e-06,
+      "loss": 1.3952,
+      "step": 18739
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.06697142857143e-06,
+      "loss": 1.2363,
+      "step": 18740
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.066702857142858e-06,
+      "loss": 0.974,
+      "step": 18741
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.066434285714286e-06,
+      "loss": 1.2303,
+      "step": 18742
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.066165714285716e-06,
+      "loss": 0.991,
+      "step": 18743
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.065897142857142e-06,
+      "loss": 0.7549,
+      "step": 18744
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.065628571428572e-06,
+      "loss": 1.0862,
+      "step": 18745
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.06536e-06,
+      "loss": 1.3787,
+      "step": 18746
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.065091428571428e-06,
+      "loss": 0.9668,
+      "step": 18747
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.064822857142857e-06,
+      "loss": 0.8728,
+      "step": 18748
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.064554285714286e-06,
+      "loss": 1.0406,
+      "step": 18749
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.064285714285715e-06,
+      "loss": 1.1755,
+      "step": 18750
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.064017142857143e-06,
+      "loss": 1.2685,
+      "step": 18751
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.063748571428573e-06,
+      "loss": 1.1417,
+      "step": 18752
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.06348e-06,
+      "loss": 1.6422,
+      "step": 18753
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.063211428571429e-06,
+      "loss": 1.0017,
+      "step": 18754
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.062942857142857e-06,
+      "loss": 1.2172,
+      "step": 18755
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.062674285714287e-06,
+      "loss": 1.1913,
+      "step": 18756
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.062405714285714e-06,
+      "loss": 1.2721,
+      "step": 18757
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.062137142857143e-06,
+      "loss": 1.3174,
+      "step": 18758
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.061868571428572e-06,
+      "loss": 1.0024,
+      "step": 18759
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.061600000000001e-06,
+      "loss": 0.9732,
+      "step": 18760
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.061331428571428e-06,
+      "loss": 1.165,
+      "step": 18761
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.061062857142858e-06,
+      "loss": 1.091,
+      "step": 18762
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.060794285714286e-06,
+      "loss": 1.4565,
+      "step": 18763
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.060525714285714e-06,
+      "loss": 0.8146,
+      "step": 18764
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.060257142857142e-06,
+      "loss": 1.0258,
+      "step": 18765
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.059988571428572e-06,
+      "loss": 1.2231,
+      "step": 18766
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05972e-06,
+      "loss": 1.1102,
+      "step": 18767
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.059451428571429e-06,
+      "loss": 0.8255,
+      "step": 18768
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.059182857142858e-06,
+      "loss": 0.7649,
+      "step": 18769
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.058914285714285e-06,
+      "loss": 1.2149,
+      "step": 18770
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.058645714285715e-06,
+      "loss": 1.6265,
+      "step": 18771
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.058377142857143e-06,
+      "loss": 1.3192,
+      "step": 18772
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.058108571428573e-06,
+      "loss": 1.0878,
+      "step": 18773
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05784e-06,
+      "loss": 1.0043,
+      "step": 18774
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05757142857143e-06,
+      "loss": 1.1862,
+      "step": 18775
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.057302857142857e-06,
+      "loss": 1.2404,
+      "step": 18776
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.057034285714286e-06,
+      "loss": 1.2932,
+      "step": 18777
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.056765714285714e-06,
+      "loss": 0.9169,
+      "step": 18778
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.056497142857144e-06,
+      "loss": 1.0586,
+      "step": 18779
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.056228571428572e-06,
+      "loss": 1.5166,
+      "step": 18780
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05596e-06,
+      "loss": 1.3156,
+      "step": 18781
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05569142857143e-06,
+      "loss": 1.2448,
+      "step": 18782
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.055422857142858e-06,
+      "loss": 1.1187,
+      "step": 18783
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.055154285714286e-06,
+      "loss": 1.4549,
+      "step": 18784
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.054885714285714e-06,
+      "loss": 1.0476,
+      "step": 18785
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.054617142857144e-06,
+      "loss": 1.4902,
+      "step": 18786
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05434857142857e-06,
+      "loss": 1.3961,
+      "step": 18787
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05408e-06,
+      "loss": 1.1625,
+      "step": 18788
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.053811428571429e-06,
+      "loss": 1.0648,
+      "step": 18789
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.053542857142857e-06,
+      "loss": 1.1338,
+      "step": 18790
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.053274285714285e-06,
+      "loss": 1.1024,
+      "step": 18791
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.053005714285715e-06,
+      "loss": 0.7616,
+      "step": 18792
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.052737142857143e-06,
+      "loss": 1.3526,
+      "step": 18793
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.052468571428571e-06,
+      "loss": 0.9014,
+      "step": 18794
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.052200000000001e-06,
+      "loss": 1.344,
+      "step": 18795
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05193142857143e-06,
+      "loss": 1.1723,
+      "step": 18796
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.051662857142858e-06,
+      "loss": 1.0625,
+      "step": 18797
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.051394285714286e-06,
+      "loss": 1.2358,
+      "step": 18798
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.051125714285716e-06,
+      "loss": 1.1281,
+      "step": 18799
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.050857142857142e-06,
+      "loss": 1.0718,
+      "step": 18800
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.050588571428572e-06,
+      "loss": 0.9774,
+      "step": 18801
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.05032e-06,
+      "loss": 1.4372,
+      "step": 18802
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.050051428571428e-06,
+      "loss": 1.0543,
+      "step": 18803
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.049782857142856e-06,
+      "loss": 1.2041,
+      "step": 18804
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.049514285714286e-06,
+      "loss": 1.1314,
+      "step": 18805
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.049245714285716e-06,
+      "loss": 1.3477,
+      "step": 18806
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.048977142857143e-06,
+      "loss": 1.0437,
+      "step": 18807
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.048708571428573e-06,
+      "loss": 1.0759,
+      "step": 18808
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.04844e-06,
+      "loss": 0.8486,
+      "step": 18809
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.048171428571429e-06,
+      "loss": 1.3737,
+      "step": 18810
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.047902857142857e-06,
+      "loss": 1.3521,
+      "step": 18811
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.047634285714287e-06,
+      "loss": 1.6707,
+      "step": 18812
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.047365714285713e-06,
+      "loss": 1.1102,
+      "step": 18813
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.047097142857143e-06,
+      "loss": 1.0212,
+      "step": 18814
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.046828571428571e-06,
+      "loss": 1.6402,
+      "step": 18815
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.04656e-06,
+      "loss": 1.2924,
+      "step": 18816
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.046291428571428e-06,
+      "loss": 1.1412,
+      "step": 18817
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.046022857142858e-06,
+      "loss": 1.1381,
+      "step": 18818
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.045754285714288e-06,
+      "loss": 1.3212,
+      "step": 18819
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.045485714285714e-06,
+      "loss": 1.0137,
+      "step": 18820
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.045217142857144e-06,
+      "loss": 1.1223,
+      "step": 18821
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.044948571428572e-06,
+      "loss": 1.1964,
+      "step": 18822
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.04468e-06,
+      "loss": 1.1739,
+      "step": 18823
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.044411428571428e-06,
+      "loss": 1.5142,
+      "step": 18824
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.044142857142858e-06,
+      "loss": 0.9946,
+      "step": 18825
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.043874285714285e-06,
+      "loss": 1.2939,
+      "step": 18826
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.043605714285715e-06,
+      "loss": 0.8258,
+      "step": 18827
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.043337142857143e-06,
+      "loss": 1.2915,
+      "step": 18828
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.043068571428573e-06,
+      "loss": 1.1133,
+      "step": 18829
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.0428e-06,
+      "loss": 1.3954,
+      "step": 18830
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.042531428571429e-06,
+      "loss": 1.4371,
+      "step": 18831
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.042262857142859e-06,
+      "loss": 1.0256,
+      "step": 18832
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.041994285714285e-06,
+      "loss": 1.2965,
+      "step": 18833
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.041725714285715e-06,
+      "loss": 0.9934,
+      "step": 18834
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.041457142857143e-06,
+      "loss": 1.2267,
+      "step": 18835
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.041188571428572e-06,
+      "loss": 1.0902,
+      "step": 18836
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.04092e-06,
+      "loss": 0.8998,
+      "step": 18837
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.04065142857143e-06,
+      "loss": 0.9963,
+      "step": 18838
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.040382857142856e-06,
+      "loss": 1.1141,
+      "step": 18839
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.040114285714286e-06,
+      "loss": 1.0038,
+      "step": 18840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.039845714285714e-06,
+      "loss": 1.2295,
+      "step": 18841
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.039577142857144e-06,
+      "loss": 1.1888,
+      "step": 18842
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.03930857142857e-06,
+      "loss": 1.6708,
+      "step": 18843
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.03904e-06,
+      "loss": 1.3206,
+      "step": 18844
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.03877142857143e-06,
+      "loss": 0.9031,
+      "step": 18845
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.038502857142857e-06,
+      "loss": 1.2608,
+      "step": 18846
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.038234285714287e-06,
+      "loss": 1.2599,
+      "step": 18847
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.037965714285715e-06,
+      "loss": 1.4038,
+      "step": 18848
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.037697142857143e-06,
+      "loss": 0.9375,
+      "step": 18849
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.037428571428571e-06,
+      "loss": 1.1857,
+      "step": 18850
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.037160000000001e-06,
+      "loss": 1.0645,
+      "step": 18851
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.03689142857143e-06,
+      "loss": 1.236,
+      "step": 18852
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.036622857142857e-06,
+      "loss": 1.2467,
+      "step": 18853
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.036354285714286e-06,
+      "loss": 1.329,
+      "step": 18854
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.036085714285715e-06,
+      "loss": 1.3033,
+      "step": 18855
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.035817142857142e-06,
+      "loss": 1.1307,
+      "step": 18856
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.035548571428572e-06,
+      "loss": 1.0892,
+      "step": 18857
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.035280000000002e-06,
+      "loss": 1.3828,
+      "step": 18858
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.035011428571428e-06,
+      "loss": 1.76,
+      "step": 18859
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.034742857142858e-06,
+      "loss": 1.2009,
+      "step": 18860
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.034474285714286e-06,
+      "loss": 0.8924,
+      "step": 18861
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.034205714285714e-06,
+      "loss": 1.2371,
+      "step": 18862
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.033937142857143e-06,
+      "loss": 1.407,
+      "step": 18863
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.033668571428572e-06,
+      "loss": 1.6148,
+      "step": 18864
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.0334e-06,
+      "loss": 1.2549,
+      "step": 18865
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.033131428571429e-06,
+      "loss": 1.0123,
+      "step": 18866
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.032862857142857e-06,
+      "loss": 1.0826,
+      "step": 18867
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.032594285714287e-06,
+      "loss": 1.1723,
+      "step": 18868
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.032325714285713e-06,
+      "loss": 1.3236,
+      "step": 18869
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.032057142857143e-06,
+      "loss": 1.2054,
+      "step": 18870
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.031788571428571e-06,
+      "loss": 0.9224,
+      "step": 18871
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.03152e-06,
+      "loss": 1.0686,
+      "step": 18872
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.03125142857143e-06,
+      "loss": 1.2088,
+      "step": 18873
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.030982857142858e-06,
+      "loss": 0.8274,
+      "step": 18874
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.030714285714286e-06,
+      "loss": 1.2546,
+      "step": 18875
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.030445714285714e-06,
+      "loss": 1.0486,
+      "step": 18876
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.030177142857144e-06,
+      "loss": 1.2007,
+      "step": 18877
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.029908571428572e-06,
+      "loss": 1.334,
+      "step": 18878
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.02964e-06,
+      "loss": 0.9063,
+      "step": 18879
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.029371428571428e-06,
+      "loss": 1.2426,
+      "step": 18880
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.029102857142858e-06,
+      "loss": 1.1835,
+      "step": 18881
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.028834285714285e-06,
+      "loss": 0.9255,
+      "step": 18882
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.028565714285715e-06,
+      "loss": 1.8187,
+      "step": 18883
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.028297142857143e-06,
+      "loss": 0.7074,
+      "step": 18884
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.028028571428571e-06,
+      "loss": 0.9469,
+      "step": 18885
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.02776e-06,
+      "loss": 1.4982,
+      "step": 18886
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.027491428571429e-06,
+      "loss": 1.0082,
+      "step": 18887
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.027222857142859e-06,
+      "loss": 1.5842,
+      "step": 18888
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.026954285714285e-06,
+      "loss": 0.6736,
+      "step": 18889
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.026685714285715e-06,
+      "loss": 1.1656,
+      "step": 18890
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.026417142857143e-06,
+      "loss": 0.9714,
+      "step": 18891
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.026148571428572e-06,
+      "loss": 1.5275,
+      "step": 18892
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.02588e-06,
+      "loss": 1.9237,
+      "step": 18893
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.02561142857143e-06,
+      "loss": 1.1153,
+      "step": 18894
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.025342857142856e-06,
+      "loss": 0.7441,
+      "step": 18895
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.025074285714286e-06,
+      "loss": 1.0024,
+      "step": 18896
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.024805714285714e-06,
+      "loss": 0.8995,
+      "step": 18897
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.024537142857142e-06,
+      "loss": 1.4183,
+      "step": 18898
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.024268571428572e-06,
+      "loss": 0.7988,
+      "step": 18899
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.024e-06,
+      "loss": 1.2703,
+      "step": 18900
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.02373142857143e-06,
+      "loss": 0.9756,
+      "step": 18901
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.023462857142857e-06,
+      "loss": 1.5615,
+      "step": 18902
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.023194285714287e-06,
+      "loss": 1.3641,
+      "step": 18903
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.022925714285715e-06,
+      "loss": 1.0553,
+      "step": 18904
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.022657142857143e-06,
+      "loss": 1.0335,
+      "step": 18905
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.022388571428571e-06,
+      "loss": 1.1201,
+      "step": 18906
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.022120000000001e-06,
+      "loss": 1.1442,
+      "step": 18907
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.021851428571427e-06,
+      "loss": 0.8325,
+      "step": 18908
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.021582857142857e-06,
+      "loss": 1.7498,
+      "step": 18909
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.021314285714285e-06,
+      "loss": 1.1618,
+      "step": 18910
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.021045714285715e-06,
+      "loss": 0.9374,
+      "step": 18911
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.020777142857144e-06,
+      "loss": 0.97,
+      "step": 18912
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.020508571428572e-06,
+      "loss": 1.1733,
+      "step": 18913
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.020240000000002e-06,
+      "loss": 1.8072,
+      "step": 18914
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.019971428571428e-06,
+      "loss": 0.8485,
+      "step": 18915
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.019702857142858e-06,
+      "loss": 0.9783,
+      "step": 18916
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.019434285714286e-06,
+      "loss": 1.6251,
+      "step": 18917
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.019165714285714e-06,
+      "loss": 1.1697,
+      "step": 18918
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.018897142857142e-06,
+      "loss": 0.9248,
+      "step": 18919
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.018628571428572e-06,
+      "loss": 0.7794,
+      "step": 18920
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.018359999999999e-06,
+      "loss": 1.0329,
+      "step": 18921
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.018091428571429e-06,
+      "loss": 0.8001,
+      "step": 18922
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.017822857142857e-06,
+      "loss": 1.5274,
+      "step": 18923
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.017554285714287e-06,
+      "loss": 1.4585,
+      "step": 18924
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.017285714285715e-06,
+      "loss": 1.0562,
+      "step": 18925
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.017017142857143e-06,
+      "loss": 1.6629,
+      "step": 18926
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.016748571428573e-06,
+      "loss": 1.1434,
+      "step": 18927
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.01648e-06,
+      "loss": 1.6989,
+      "step": 18928
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.01621142857143e-06,
+      "loss": 1.0242,
+      "step": 18929
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.015942857142857e-06,
+      "loss": 1.1642,
+      "step": 18930
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.015674285714286e-06,
+      "loss": 0.7252,
+      "step": 18931
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.015405714285714e-06,
+      "loss": 1.2859,
+      "step": 18932
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.015137142857144e-06,
+      "loss": 1.0103,
+      "step": 18933
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.014868571428572e-06,
+      "loss": 0.8524,
+      "step": 18934
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.0146e-06,
+      "loss": 1.3688,
+      "step": 18935
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.014331428571428e-06,
+      "loss": 1.1759,
+      "step": 18936
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.014062857142858e-06,
+      "loss": 1.0953,
+      "step": 18937
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.013794285714286e-06,
+      "loss": 1.5969,
+      "step": 18938
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.013525714285714e-06,
+      "loss": 1.1253,
+      "step": 18939
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.013257142857144e-06,
+      "loss": 0.9756,
+      "step": 18940
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.01298857142857e-06,
+      "loss": 1.2505,
+      "step": 18941
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.01272e-06,
+      "loss": 1.2861,
+      "step": 18942
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.012451428571429e-06,
+      "loss": 1.1305,
+      "step": 18943
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.012182857142857e-06,
+      "loss": 1.0409,
+      "step": 18944
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.011914285714285e-06,
+      "loss": 1.2857,
+      "step": 18945
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.011645714285715e-06,
+      "loss": 1.2819,
+      "step": 18946
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.011377142857143e-06,
+      "loss": 1.29,
+      "step": 18947
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.011108571428571e-06,
+      "loss": 1.0751,
+      "step": 18948
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.01084e-06,
+      "loss": 1.5245,
+      "step": 18949
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.01057142857143e-06,
+      "loss": 1.1105,
+      "step": 18950
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.010302857142858e-06,
+      "loss": 1.502,
+      "step": 18951
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.010034285714286e-06,
+      "loss": 1.211,
+      "step": 18952
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.009765714285716e-06,
+      "loss": 1.1235,
+      "step": 18953
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.009497142857142e-06,
+      "loss": 1.6383,
+      "step": 18954
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.009228571428572e-06,
+      "loss": 1.3526,
+      "step": 18955
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.00896e-06,
+      "loss": 1.3664,
+      "step": 18956
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.00869142857143e-06,
+      "loss": 1.2603,
+      "step": 18957
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.008422857142857e-06,
+      "loss": 0.8575,
+      "step": 18958
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.008154285714286e-06,
+      "loss": 1.2844,
+      "step": 18959
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.007885714285715e-06,
+      "loss": 1.1789,
+      "step": 18960
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.007617142857143e-06,
+      "loss": 1.0743,
+      "step": 18961
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.007348571428571e-06,
+      "loss": 1.1348,
+      "step": 18962
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.00708e-06,
+      "loss": 1.3133,
+      "step": 18963
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.006811428571429e-06,
+      "loss": 1.1241,
+      "step": 18964
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.006542857142857e-06,
+      "loss": 1.6868,
+      "step": 18965
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.006274285714287e-06,
+      "loss": 1.366,
+      "step": 18966
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.006005714285714e-06,
+      "loss": 1.1476,
+      "step": 18967
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.005737142857143e-06,
+      "loss": 1.3722,
+      "step": 18968
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.005468571428572e-06,
+      "loss": 1.2441,
+      "step": 18969
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.005200000000001e-06,
+      "loss": 1.4104,
+      "step": 18970
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.004931428571428e-06,
+      "loss": 1.3672,
+      "step": 18971
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.004662857142858e-06,
+      "loss": 1.0649,
+      "step": 18972
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.004394285714286e-06,
+      "loss": 0.9851,
+      "step": 18973
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.004125714285714e-06,
+      "loss": 1.1106,
+      "step": 18974
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.003857142857142e-06,
+      "loss": 1.1947,
+      "step": 18975
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.003588571428572e-06,
+      "loss": 1.125,
+      "step": 18976
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.00332e-06,
+      "loss": 0.8078,
+      "step": 18977
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.003051428571429e-06,
+      "loss": 1.0812,
+      "step": 18978
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.002782857142858e-06,
+      "loss": 0.8158,
+      "step": 18979
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.002514285714287e-06,
+      "loss": 1.0988,
+      "step": 18980
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.002245714285715e-06,
+      "loss": 1.2341,
+      "step": 18981
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.001977142857143e-06,
+      "loss": 1.1083,
+      "step": 18982
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.001708571428573e-06,
+      "loss": 1.2575,
+      "step": 18983
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.00144e-06,
+      "loss": 1.0369,
+      "step": 18984
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.00117142857143e-06,
+      "loss": 1.3307,
+      "step": 18985
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.000902857142857e-06,
+      "loss": 1.0256,
+      "step": 18986
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.000634285714286e-06,
+      "loss": 1.1045,
+      "step": 18987
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.000365714285714e-06,
+      "loss": 1.3038,
+      "step": 18988
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.000097142857144e-06,
+      "loss": 1.6295,
+      "step": 18989
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.999828571428572e-06,
+      "loss": 1.3938,
+      "step": 18990
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99956e-06,
+      "loss": 0.9406,
+      "step": 18991
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99929142857143e-06,
+      "loss": 0.8615,
+      "step": 18992
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.999022857142858e-06,
+      "loss": 0.9885,
+      "step": 18993
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.998754285714286e-06,
+      "loss": 1.2279,
+      "step": 18994
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.998485714285714e-06,
+      "loss": 1.3351,
+      "step": 18995
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.998217142857144e-06,
+      "loss": 1.0459,
+      "step": 18996
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99794857142857e-06,
+      "loss": 1.1029,
+      "step": 18997
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99768e-06,
+      "loss": 1.2733,
+      "step": 18998
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.997411428571429e-06,
+      "loss": 1.0547,
+      "step": 18999
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.997142857142857e-06,
+      "loss": 1.2429,
+      "step": 19000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.996874285714285e-06,
+      "loss": 1.1382,
+      "step": 19001
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.996605714285715e-06,
+      "loss": 0.9979,
+      "step": 19002
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.996337142857143e-06,
+      "loss": 1.0471,
+      "step": 19003
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.996068571428571e-06,
+      "loss": 1.08,
+      "step": 19004
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.995800000000001e-06,
+      "loss": 1.1745,
+      "step": 19005
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99553142857143e-06,
+      "loss": 0.9776,
+      "step": 19006
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.995262857142858e-06,
+      "loss": 1.517,
+      "step": 19007
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.994994285714286e-06,
+      "loss": 1.0342,
+      "step": 19008
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.994725714285716e-06,
+      "loss": 1.3263,
+      "step": 19009
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.994457142857142e-06,
+      "loss": 1.3831,
+      "step": 19010
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.994188571428572e-06,
+      "loss": 1.0189,
+      "step": 19011
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99392e-06,
+      "loss": 0.9715,
+      "step": 19012
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.993651428571428e-06,
+      "loss": 0.7737,
+      "step": 19013
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.993382857142856e-06,
+      "loss": 0.8799,
+      "step": 19014
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.993114285714286e-06,
+      "loss": 1.1033,
+      "step": 19015
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.992845714285715e-06,
+      "loss": 0.934,
+      "step": 19016
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.992577142857143e-06,
+      "loss": 1.0785,
+      "step": 19017
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.992308571428573e-06,
+      "loss": 1.221,
+      "step": 19018
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99204e-06,
+      "loss": 1.0185,
+      "step": 19019
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.991771428571429e-06,
+      "loss": 1.3605,
+      "step": 19020
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.991502857142857e-06,
+      "loss": 0.8837,
+      "step": 19021
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.991234285714287e-06,
+      "loss": 1.0178,
+      "step": 19022
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.990965714285713e-06,
+      "loss": 0.7261,
+      "step": 19023
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.990697142857143e-06,
+      "loss": 1.3777,
+      "step": 19024
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.990428571428571e-06,
+      "loss": 0.9148,
+      "step": 19025
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.99016e-06,
+      "loss": 1.4272,
+      "step": 19026
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.989891428571428e-06,
+      "loss": 0.9805,
+      "step": 19027
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.989622857142858e-06,
+      "loss": 1.2575,
+      "step": 19028
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.989354285714286e-06,
+      "loss": 1.1296,
+      "step": 19029
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.989085714285714e-06,
+      "loss": 1.2138,
+      "step": 19030
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.988817142857144e-06,
+      "loss": 0.7941,
+      "step": 19031
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.988548571428572e-06,
+      "loss": 1.2455,
+      "step": 19032
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.98828e-06,
+      "loss": 0.9705,
+      "step": 19033
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.988011428571428e-06,
+      "loss": 1.3588,
+      "step": 19034
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.987742857142858e-06,
+      "loss": 1.0753,
+      "step": 19035
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.987474285714285e-06,
+      "loss": 1.2815,
+      "step": 19036
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.987205714285715e-06,
+      "loss": 1.1157,
+      "step": 19037
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.986937142857143e-06,
+      "loss": 1.5643,
+      "step": 19038
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.986668571428573e-06,
+      "loss": 1.1852,
+      "step": 19039
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.9864e-06,
+      "loss": 1.5031,
+      "step": 19040
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.986131428571429e-06,
+      "loss": 1.1683,
+      "step": 19041
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.985862857142857e-06,
+      "loss": 1.1439,
+      "step": 19042
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.985594285714285e-06,
+      "loss": 1.1022,
+      "step": 19043
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.985325714285715e-06,
+      "loss": 0.9104,
+      "step": 19044
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.985057142857143e-06,
+      "loss": 1.2614,
+      "step": 19045
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.984788571428572e-06,
+      "loss": 1.0773,
+      "step": 19046
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.98452e-06,
+      "loss": 1.0049,
+      "step": 19047
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.98425142857143e-06,
+      "loss": 1.6861,
+      "step": 19048
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.983982857142856e-06,
+      "loss": 1.0376,
+      "step": 19049
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.983714285714286e-06,
+      "loss": 1.3605,
+      "step": 19050
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.983445714285714e-06,
+      "loss": 1.4436,
+      "step": 19051
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.983177142857144e-06,
+      "loss": 1.5213,
+      "step": 19052
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.98290857142857e-06,
+      "loss": 1.1897,
+      "step": 19053
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.98264e-06,
+      "loss": 0.9854,
+      "step": 19054
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.982371428571429e-06,
+      "loss": 1.0431,
+      "step": 19055
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.982102857142857e-06,
+      "loss": 1.1634,
+      "step": 19056
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.981834285714287e-06,
+      "loss": 1.2225,
+      "step": 19057
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.981565714285715e-06,
+      "loss": 1.5814,
+      "step": 19058
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.981297142857143e-06,
+      "loss": 0.9374,
+      "step": 19059
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.981028571428571e-06,
+      "loss": 1.1812,
+      "step": 19060
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.980760000000001e-06,
+      "loss": 1.1605,
+      "step": 19061
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.98049142857143e-06,
+      "loss": 0.8282,
+      "step": 19062
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.980222857142857e-06,
+      "loss": 1.414,
+      "step": 19063
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.979954285714286e-06,
+      "loss": 0.9544,
+      "step": 19064
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.979685714285715e-06,
+      "loss": 0.9306,
+      "step": 19065
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.979417142857142e-06,
+      "loss": 1.3338,
+      "step": 19066
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.979148571428572e-06,
+      "loss": 0.851,
+      "step": 19067
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.97888e-06,
+      "loss": 0.968,
+      "step": 19068
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.978611428571428e-06,
+      "loss": 1.1304,
+      "step": 19069
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.978342857142858e-06,
+      "loss": 1.361,
+      "step": 19070
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.978074285714286e-06,
+      "loss": 1.5667,
+      "step": 19071
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.977805714285714e-06,
+      "loss": 0.7068,
+      "step": 19072
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.977537142857143e-06,
+      "loss": 1.1904,
+      "step": 19073
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.977268571428572e-06,
+      "loss": 1.2077,
+      "step": 19074
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.977e-06,
+      "loss": 1.2162,
+      "step": 19075
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.976731428571429e-06,
+      "loss": 1.1176,
+      "step": 19076
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.976462857142857e-06,
+      "loss": 1.0083,
+      "step": 19077
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.976194285714287e-06,
+      "loss": 1.3023,
+      "step": 19078
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.975925714285713e-06,
+      "loss": 1.5029,
+      "step": 19079
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.975657142857143e-06,
+      "loss": 1.3099,
+      "step": 19080
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.975388571428571e-06,
+      "loss": 0.9009,
+      "step": 19081
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.97512e-06,
+      "loss": 1.0287,
+      "step": 19082
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.97485142857143e-06,
+      "loss": 0.9311,
+      "step": 19083
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.974582857142858e-06,
+      "loss": 1.4125,
+      "step": 19084
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.974314285714287e-06,
+      "loss": 1.4029,
+      "step": 19085
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.974045714285714e-06,
+      "loss": 1.3414,
+      "step": 19086
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.973777142857144e-06,
+      "loss": 1.3407,
+      "step": 19087
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.973508571428572e-06,
+      "loss": 0.9771,
+      "step": 19088
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.97324e-06,
+      "loss": 0.9976,
+      "step": 19089
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.972971428571428e-06,
+      "loss": 1.611,
+      "step": 19090
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.972702857142858e-06,
+      "loss": 1.7758,
+      "step": 19091
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.972434285714285e-06,
+      "loss": 1.3316,
+      "step": 19092
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.972165714285715e-06,
+      "loss": 1.2094,
+      "step": 19093
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.971897142857143e-06,
+      "loss": 1.2629,
+      "step": 19094
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.971628571428571e-06,
+      "loss": 1.039,
+      "step": 19095
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.97136e-06,
+      "loss": 1.3037,
+      "step": 19096
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.971091428571429e-06,
+      "loss": 1.3675,
+      "step": 19097
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.970822857142859e-06,
+      "loss": 0.9245,
+      "step": 19098
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.970554285714285e-06,
+      "loss": 1.0358,
+      "step": 19099
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.970285714285715e-06,
+      "loss": 1.1394,
+      "step": 19100
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.970017142857143e-06,
+      "loss": 1.3246,
+      "step": 19101
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.969748571428572e-06,
+      "loss": 1.733,
+      "step": 19102
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.96948e-06,
+      "loss": 1.5065,
+      "step": 19103
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.96921142857143e-06,
+      "loss": 1.303,
+      "step": 19104
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.968942857142856e-06,
+      "loss": 1.1702,
+      "step": 19105
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.968674285714286e-06,
+      "loss": 1.1323,
+      "step": 19106
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.968405714285714e-06,
+      "loss": 0.9324,
+      "step": 19107
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.968137142857144e-06,
+      "loss": 1.3081,
+      "step": 19108
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.967868571428572e-06,
+      "loss": 1.2739,
+      "step": 19109
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.9676e-06,
+      "loss": 0.9973,
+      "step": 19110
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.96733142857143e-06,
+      "loss": 1.0733,
+      "step": 19111
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.967062857142857e-06,
+      "loss": 1.3097,
+      "step": 19112
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.966794285714287e-06,
+      "loss": 1.5389,
+      "step": 19113
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.966525714285715e-06,
+      "loss": 1.0717,
+      "step": 19114
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.966257142857143e-06,
+      "loss": 1.4371,
+      "step": 19115
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.965988571428571e-06,
+      "loss": 1.3227,
+      "step": 19116
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.965720000000001e-06,
+      "loss": 1.2085,
+      "step": 19117
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.965451428571427e-06,
+      "loss": 0.7104,
+      "step": 19118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.965182857142857e-06,
+      "loss": 1.6005,
+      "step": 19119
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.964914285714285e-06,
+      "loss": 1.3199,
+      "step": 19120
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.964645714285715e-06,
+      "loss": 0.8732,
+      "step": 19121
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.964377142857144e-06,
+      "loss": 1.0133,
+      "step": 19122
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.964108571428572e-06,
+      "loss": 1.1378,
+      "step": 19123
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.963840000000002e-06,
+      "loss": 1.1287,
+      "step": 19124
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.963571428571428e-06,
+      "loss": 1.4194,
+      "step": 19125
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.963302857142858e-06,
+      "loss": 0.8249,
+      "step": 19126
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.963034285714286e-06,
+      "loss": 1.1967,
+      "step": 19127
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.962765714285714e-06,
+      "loss": 1.5645,
+      "step": 19128
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.962497142857142e-06,
+      "loss": 1.5268,
+      "step": 19129
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.962228571428572e-06,
+      "loss": 0.8813,
+      "step": 19130
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.96196e-06,
+      "loss": 1.4408,
+      "step": 19131
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.961691428571429e-06,
+      "loss": 1.1412,
+      "step": 19132
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.961422857142857e-06,
+      "loss": 0.8543,
+      "step": 19133
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.961154285714287e-06,
+      "loss": 0.8973,
+      "step": 19134
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.960885714285715e-06,
+      "loss": 0.9711,
+      "step": 19135
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.960617142857143e-06,
+      "loss": 1.3902,
+      "step": 19136
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.960348571428573e-06,
+      "loss": 0.7902,
+      "step": 19137
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.96008e-06,
+      "loss": 0.8396,
+      "step": 19138
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.95981142857143e-06,
+      "loss": 0.9679,
+      "step": 19139
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.959542857142857e-06,
+      "loss": 0.9602,
+      "step": 19140
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.959274285714286e-06,
+      "loss": 1.3042,
+      "step": 19141
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.959005714285714e-06,
+      "loss": 0.8535,
+      "step": 19142
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.958737142857144e-06,
+      "loss": 1.1105,
+      "step": 19143
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.958468571428572e-06,
+      "loss": 0.9802,
+      "step": 19144
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.9582e-06,
+      "loss": 1.0541,
+      "step": 19145
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.957931428571428e-06,
+      "loss": 0.821,
+      "step": 19146
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.957662857142858e-06,
+      "loss": 1.3227,
+      "step": 19147
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.957394285714286e-06,
+      "loss": 1.1938,
+      "step": 19148
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.957125714285714e-06,
+      "loss": 1.0974,
+      "step": 19149
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.956857142857144e-06,
+      "loss": 1.0241,
+      "step": 19150
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.95658857142857e-06,
+      "loss": 1.0773,
+      "step": 19151
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.95632e-06,
+      "loss": 1.0351,
+      "step": 19152
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.956051428571429e-06,
+      "loss": 1.3293,
+      "step": 19153
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.955782857142859e-06,
+      "loss": 0.9536,
+      "step": 19154
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.955514285714285e-06,
+      "loss": 1.06,
+      "step": 19155
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.955245714285715e-06,
+      "loss": 1.1718,
+      "step": 19156
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.954977142857143e-06,
+      "loss": 1.2687,
+      "step": 19157
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.954708571428571e-06,
+      "loss": 0.9866,
+      "step": 19158
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.95444e-06,
+      "loss": 1.6322,
+      "step": 19159
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.95417142857143e-06,
+      "loss": 1.3713,
+      "step": 19160
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.953902857142858e-06,
+      "loss": 1.1254,
+      "step": 19161
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.953634285714286e-06,
+      "loss": 1.1374,
+      "step": 19162
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.953365714285716e-06,
+      "loss": 0.9443,
+      "step": 19163
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.953097142857142e-06,
+      "loss": 1.2778,
+      "step": 19164
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.952828571428572e-06,
+      "loss": 1.0347,
+      "step": 19165
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.95256e-06,
+      "loss": 1.1531,
+      "step": 19166
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.95229142857143e-06,
+      "loss": 1.0823,
+      "step": 19167
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.952022857142857e-06,
+      "loss": 1.3656,
+      "step": 19168
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.951754285714286e-06,
+      "loss": 1.0553,
+      "step": 19169
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.951485714285715e-06,
+      "loss": 1.1491,
+      "step": 19170
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.951217142857143e-06,
+      "loss": 1.7806,
+      "step": 19171
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.950948571428571e-06,
+      "loss": 1.1655,
+      "step": 19172
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.95068e-06,
+      "loss": 0.9569,
+      "step": 19173
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.950411428571429e-06,
+      "loss": 1.2821,
+      "step": 19174
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.950142857142857e-06,
+      "loss": 1.0582,
+      "step": 19175
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.949874285714287e-06,
+      "loss": 0.8389,
+      "step": 19176
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.949605714285715e-06,
+      "loss": 1.0237,
+      "step": 19177
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.949337142857143e-06,
+      "loss": 1.1897,
+      "step": 19178
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.949068571428572e-06,
+      "loss": 0.9474,
+      "step": 19179
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.948800000000001e-06,
+      "loss": 0.9201,
+      "step": 19180
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.948531428571428e-06,
+      "loss": 0.8649,
+      "step": 19181
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.948262857142858e-06,
+      "loss": 0.8541,
+      "step": 19182
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.947994285714286e-06,
+      "loss": 0.9345,
+      "step": 19183
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.947725714285714e-06,
+      "loss": 1.3174,
+      "step": 19184
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.947457142857142e-06,
+      "loss": 1.129,
+      "step": 19185
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.947188571428572e-06,
+      "loss": 1.2328,
+      "step": 19186
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.94692e-06,
+      "loss": 1.5032,
+      "step": 19187
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.946651428571429e-06,
+      "loss": 1.3006,
+      "step": 19188
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.946382857142858e-06,
+      "loss": 0.7112,
+      "step": 19189
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.946114285714287e-06,
+      "loss": 0.9294,
+      "step": 19190
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.945845714285715e-06,
+      "loss": 1.5171,
+      "step": 19191
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.945577142857143e-06,
+      "loss": 1.3285,
+      "step": 19192
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.945308571428573e-06,
+      "loss": 1.2767,
+      "step": 19193
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.94504e-06,
+      "loss": 1.2937,
+      "step": 19194
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.94477142857143e-06,
+      "loss": 1.2268,
+      "step": 19195
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.944502857142857e-06,
+      "loss": 0.9239,
+      "step": 19196
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.944234285714286e-06,
+      "loss": 1.0692,
+      "step": 19197
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.943965714285714e-06,
+      "loss": 1.3422,
+      "step": 19198
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.943697142857144e-06,
+      "loss": 0.7906,
+      "step": 19199
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.943428571428572e-06,
+      "loss": 1.2621,
+      "step": 19200
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.94316e-06,
+      "loss": 0.9896,
+      "step": 19201
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.94289142857143e-06,
+      "loss": 1.0497,
+      "step": 19202
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.942622857142858e-06,
+      "loss": 1.2976,
+      "step": 19203
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.942354285714286e-06,
+      "loss": 1.37,
+      "step": 19204
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.942085714285714e-06,
+      "loss": 0.9708,
+      "step": 19205
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.941817142857144e-06,
+      "loss": 1.5127,
+      "step": 19206
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.94154857142857e-06,
+      "loss": 1.0163,
+      "step": 19207
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.94128e-06,
+      "loss": 0.7911,
+      "step": 19208
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.941011428571429e-06,
+      "loss": 1.0662,
+      "step": 19209
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.940742857142857e-06,
+      "loss": 1.4289,
+      "step": 19210
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.940474285714285e-06,
+      "loss": 1.2577,
+      "step": 19211
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.940205714285715e-06,
+      "loss": 0.8098,
+      "step": 19212
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.939937142857143e-06,
+      "loss": 1.253,
+      "step": 19213
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.939668571428571e-06,
+      "loss": 1.1277,
+      "step": 19214
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.939400000000001e-06,
+      "loss": 1.1848,
+      "step": 19215
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.93913142857143e-06,
+      "loss": 1.011,
+      "step": 19216
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.938862857142858e-06,
+      "loss": 1.2446,
+      "step": 19217
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.938594285714286e-06,
+      "loss": 1.2793,
+      "step": 19218
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.938325714285716e-06,
+      "loss": 0.9367,
+      "step": 19219
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.938057142857142e-06,
+      "loss": 1.2501,
+      "step": 19220
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.937788571428572e-06,
+      "loss": 1.2487,
+      "step": 19221
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.93752e-06,
+      "loss": 1.1682,
+      "step": 19222
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.937251428571428e-06,
+      "loss": 1.2509,
+      "step": 19223
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.936982857142856e-06,
+      "loss": 1.1361,
+      "step": 19224
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.936714285714286e-06,
+      "loss": 1.2751,
+      "step": 19225
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.936445714285715e-06,
+      "loss": 0.9192,
+      "step": 19226
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.936177142857143e-06,
+      "loss": 1.0859,
+      "step": 19227
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.935908571428573e-06,
+      "loss": 1.1828,
+      "step": 19228
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.93564e-06,
+      "loss": 1.2168,
+      "step": 19229
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.935371428571429e-06,
+      "loss": 0.8114,
+      "step": 19230
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.935102857142857e-06,
+      "loss": 1.3415,
+      "step": 19231
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.934834285714287e-06,
+      "loss": 1.3647,
+      "step": 19232
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.934565714285713e-06,
+      "loss": 1.0764,
+      "step": 19233
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.934297142857143e-06,
+      "loss": 1.1021,
+      "step": 19234
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.934028571428571e-06,
+      "loss": 1.3831,
+      "step": 19235
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.933760000000001e-06,
+      "loss": 0.8881,
+      "step": 19236
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.933491428571428e-06,
+      "loss": 1.2517,
+      "step": 19237
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.933222857142858e-06,
+      "loss": 1.2961,
+      "step": 19238
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.932954285714286e-06,
+      "loss": 0.9755,
+      "step": 19239
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.932685714285714e-06,
+      "loss": 1.6359,
+      "step": 19240
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.932417142857144e-06,
+      "loss": 1.081,
+      "step": 19241
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.932148571428572e-06,
+      "loss": 0.9602,
+      "step": 19242
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.93188e-06,
+      "loss": 1.4059,
+      "step": 19243
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.931611428571428e-06,
+      "loss": 1.3418,
+      "step": 19244
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.931342857142858e-06,
+      "loss": 1.1641,
+      "step": 19245
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.931074285714285e-06,
+      "loss": 1.4356,
+      "step": 19246
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.930805714285715e-06,
+      "loss": 1.3051,
+      "step": 19247
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.930537142857143e-06,
+      "loss": 1.4594,
+      "step": 19248
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.930268571428573e-06,
+      "loss": 0.9451,
+      "step": 19249
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.93e-06,
+      "loss": 1.5957,
+      "step": 19250
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.929731428571429e-06,
+      "loss": 1.2123,
+      "step": 19251
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.929462857142857e-06,
+      "loss": 1.343,
+      "step": 19252
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.929194285714285e-06,
+      "loss": 1.0128,
+      "step": 19253
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.928925714285715e-06,
+      "loss": 1.2489,
+      "step": 19254
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.928657142857143e-06,
+      "loss": 1.5843,
+      "step": 19255
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.928388571428572e-06,
+      "loss": 0.6214,
+      "step": 19256
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92812e-06,
+      "loss": 1.4422,
+      "step": 19257
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92785142857143e-06,
+      "loss": 1.4321,
+      "step": 19258
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.927582857142858e-06,
+      "loss": 1.101,
+      "step": 19259
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.927314285714286e-06,
+      "loss": 1.0469,
+      "step": 19260
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.927045714285714e-06,
+      "loss": 1.0152,
+      "step": 19261
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.926777142857144e-06,
+      "loss": 1.041,
+      "step": 19262
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92650857142857e-06,
+      "loss": 1.1226,
+      "step": 19263
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92624e-06,
+      "loss": 0.7213,
+      "step": 19264
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.925971428571429e-06,
+      "loss": 1.1685,
+      "step": 19265
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.925702857142857e-06,
+      "loss": 1.187,
+      "step": 19266
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.925434285714287e-06,
+      "loss": 1.0022,
+      "step": 19267
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.925165714285715e-06,
+      "loss": 0.8471,
+      "step": 19268
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.924897142857143e-06,
+      "loss": 1.5117,
+      "step": 19269
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.924628571428571e-06,
+      "loss": 1.1331,
+      "step": 19270
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.924360000000001e-06,
+      "loss": 1.7317,
+      "step": 19271
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92409142857143e-06,
+      "loss": 1.2712,
+      "step": 19272
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.923822857142857e-06,
+      "loss": 0.8407,
+      "step": 19273
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.923554285714286e-06,
+      "loss": 1.6111,
+      "step": 19274
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.923285714285715e-06,
+      "loss": 1.2767,
+      "step": 19275
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.923017142857142e-06,
+      "loss": 0.9677,
+      "step": 19276
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.922748571428572e-06,
+      "loss": 1.3964,
+      "step": 19277
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92248e-06,
+      "loss": 1.3557,
+      "step": 19278
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.922211428571428e-06,
+      "loss": 1.3951,
+      "step": 19279
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.921942857142858e-06,
+      "loss": 1.4518,
+      "step": 19280
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.921674285714286e-06,
+      "loss": 1.0774,
+      "step": 19281
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.921405714285716e-06,
+      "loss": 1.1689,
+      "step": 19282
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.921137142857143e-06,
+      "loss": 1.2638,
+      "step": 19283
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.920868571428572e-06,
+      "loss": 1.2156,
+      "step": 19284
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.9206e-06,
+      "loss": 1.4353,
+      "step": 19285
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.920331428571429e-06,
+      "loss": 1.3457,
+      "step": 19286
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.920062857142857e-06,
+      "loss": 0.8215,
+      "step": 19287
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.919794285714287e-06,
+      "loss": 1.1173,
+      "step": 19288
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.919525714285713e-06,
+      "loss": 1.3913,
+      "step": 19289
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.919257142857143e-06,
+      "loss": 1.1703,
+      "step": 19290
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.918988571428571e-06,
+      "loss": 1.115,
+      "step": 19291
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91872e-06,
+      "loss": 1.2951,
+      "step": 19292
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91845142857143e-06,
+      "loss": 1.3081,
+      "step": 19293
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.918182857142858e-06,
+      "loss": 1.0956,
+      "step": 19294
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.917914285714287e-06,
+      "loss": 0.8479,
+      "step": 19295
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.917645714285714e-06,
+      "loss": 1.092,
+      "step": 19296
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.917377142857144e-06,
+      "loss": 1.2412,
+      "step": 19297
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.917108571428572e-06,
+      "loss": 1.4212,
+      "step": 19298
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91684e-06,
+      "loss": 1.1518,
+      "step": 19299
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.916571428571428e-06,
+      "loss": 1.0849,
+      "step": 19300
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.916302857142858e-06,
+      "loss": 0.9012,
+      "step": 19301
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.916034285714285e-06,
+      "loss": 1.4154,
+      "step": 19302
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.915765714285715e-06,
+      "loss": 1.1282,
+      "step": 19303
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.915497142857143e-06,
+      "loss": 1.4565,
+      "step": 19304
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.915228571428573e-06,
+      "loss": 1.1146,
+      "step": 19305
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91496e-06,
+      "loss": 1.2624,
+      "step": 19306
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.914691428571429e-06,
+      "loss": 1.0563,
+      "step": 19307
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.914422857142859e-06,
+      "loss": 1.1279,
+      "step": 19308
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.914154285714285e-06,
+      "loss": 1.0802,
+      "step": 19309
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.913885714285715e-06,
+      "loss": 1.3384,
+      "step": 19310
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.913617142857143e-06,
+      "loss": 1.0403,
+      "step": 19311
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.913348571428572e-06,
+      "loss": 1.3162,
+      "step": 19312
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91308e-06,
+      "loss": 1.0733,
+      "step": 19313
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91281142857143e-06,
+      "loss": 1.4053,
+      "step": 19314
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.912542857142856e-06,
+      "loss": 1.2593,
+      "step": 19315
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.912274285714286e-06,
+      "loss": 1.2923,
+      "step": 19316
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.912005714285714e-06,
+      "loss": 0.8934,
+      "step": 19317
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.911737142857144e-06,
+      "loss": 1.3585,
+      "step": 19318
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.911468571428572e-06,
+      "loss": 1.0756,
+      "step": 19319
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.9112e-06,
+      "loss": 1.0089,
+      "step": 19320
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.91093142857143e-06,
+      "loss": 1.3055,
+      "step": 19321
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.910662857142857e-06,
+      "loss": 1.4085,
+      "step": 19322
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.910394285714287e-06,
+      "loss": 1.0364,
+      "step": 19323
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.910125714285715e-06,
+      "loss": 1.547,
+      "step": 19324
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.909857142857143e-06,
+      "loss": 1.3929,
+      "step": 19325
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.909588571428571e-06,
+      "loss": 1.3478,
+      "step": 19326
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.909320000000001e-06,
+      "loss": 1.0795,
+      "step": 19327
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.909051428571429e-06,
+      "loss": 1.2558,
+      "step": 19328
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.908782857142857e-06,
+      "loss": 0.8249,
+      "step": 19329
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.908514285714285e-06,
+      "loss": 1.0707,
+      "step": 19330
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.908245714285715e-06,
+      "loss": 0.7528,
+      "step": 19331
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.907977142857144e-06,
+      "loss": 1.3561,
+      "step": 19332
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.907708571428572e-06,
+      "loss": 1.4783,
+      "step": 19333
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.907440000000002e-06,
+      "loss": 1.3096,
+      "step": 19334
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.907171428571428e-06,
+      "loss": 1.1425,
+      "step": 19335
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.906902857142858e-06,
+      "loss": 1.2025,
+      "step": 19336
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.906634285714286e-06,
+      "loss": 0.8979,
+      "step": 19337
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.906365714285714e-06,
+      "loss": 1.1398,
+      "step": 19338
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.906097142857142e-06,
+      "loss": 1.2353,
+      "step": 19339
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.905828571428572e-06,
+      "loss": 1.0391,
+      "step": 19340
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.90556e-06,
+      "loss": 0.8701,
+      "step": 19341
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.905291428571429e-06,
+      "loss": 0.8227,
+      "step": 19342
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.905022857142857e-06,
+      "loss": 1.4799,
+      "step": 19343
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.904754285714287e-06,
+      "loss": 1.1532,
+      "step": 19344
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.904485714285715e-06,
+      "loss": 0.8514,
+      "step": 19345
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.904217142857143e-06,
+      "loss": 1.0977,
+      "step": 19346
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.903948571428573e-06,
+      "loss": 1.0534,
+      "step": 19347
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.90368e-06,
+      "loss": 1.2079,
+      "step": 19348
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.90341142857143e-06,
+      "loss": 0.7645,
+      "step": 19349
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.903142857142857e-06,
+      "loss": 1.1472,
+      "step": 19350
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.902874285714287e-06,
+      "loss": 1.4865,
+      "step": 19351
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.902605714285714e-06,
+      "loss": 0.9967,
+      "step": 19352
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.902337142857144e-06,
+      "loss": 0.8218,
+      "step": 19353
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.902068571428572e-06,
+      "loss": 1.4314,
+      "step": 19354
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.9018e-06,
+      "loss": 0.8891,
+      "step": 19355
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.901531428571428e-06,
+      "loss": 0.6191,
+      "step": 19356
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.901262857142858e-06,
+      "loss": 1.2367,
+      "step": 19357
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.900994285714286e-06,
+      "loss": 1.3419,
+      "step": 19358
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.900725714285714e-06,
+      "loss": 1.0405,
+      "step": 19359
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.900457142857144e-06,
+      "loss": 0.9503,
+      "step": 19360
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.90018857142857e-06,
+      "loss": 1.4326,
+      "step": 19361
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89992e-06,
+      "loss": 1.1713,
+      "step": 19362
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.899651428571429e-06,
+      "loss": 1.2025,
+      "step": 19363
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.899382857142859e-06,
+      "loss": 1.5804,
+      "step": 19364
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.899114285714285e-06,
+      "loss": 1.2164,
+      "step": 19365
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.898845714285715e-06,
+      "loss": 1.288,
+      "step": 19366
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.898577142857143e-06,
+      "loss": 1.0841,
+      "step": 19367
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.898308571428571e-06,
+      "loss": 0.9525,
+      "step": 19368
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89804e-06,
+      "loss": 1.4458,
+      "step": 19369
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89777142857143e-06,
+      "loss": 0.9414,
+      "step": 19370
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.897502857142858e-06,
+      "loss": 0.9612,
+      "step": 19371
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.897234285714286e-06,
+      "loss": 1.0694,
+      "step": 19372
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.896965714285716e-06,
+      "loss": 1.1275,
+      "step": 19373
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.896697142857142e-06,
+      "loss": 1.108,
+      "step": 19374
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.896428571428572e-06,
+      "loss": 1.1413,
+      "step": 19375
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89616e-06,
+      "loss": 1.2711,
+      "step": 19376
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89589142857143e-06,
+      "loss": 0.7556,
+      "step": 19377
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.895622857142857e-06,
+      "loss": 1.2932,
+      "step": 19378
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.895354285714286e-06,
+      "loss": 1.1154,
+      "step": 19379
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.895085714285715e-06,
+      "loss": 1.1223,
+      "step": 19380
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.894817142857143e-06,
+      "loss": 1.1671,
+      "step": 19381
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.894548571428571e-06,
+      "loss": 1.2959,
+      "step": 19382
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89428e-06,
+      "loss": 1.0391,
+      "step": 19383
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.894011428571429e-06,
+      "loss": 1.1446,
+      "step": 19384
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.893742857142857e-06,
+      "loss": 1.1571,
+      "step": 19385
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.893474285714287e-06,
+      "loss": 1.0407,
+      "step": 19386
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.893205714285715e-06,
+      "loss": 0.9984,
+      "step": 19387
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.892937142857143e-06,
+      "loss": 1.5179,
+      "step": 19388
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.892668571428572e-06,
+      "loss": 1.2328,
+      "step": 19389
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.892400000000001e-06,
+      "loss": 1.0025,
+      "step": 19390
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.892131428571428e-06,
+      "loss": 1.1201,
+      "step": 19391
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.891862857142858e-06,
+      "loss": 0.8147,
+      "step": 19392
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.891594285714286e-06,
+      "loss": 1.1743,
+      "step": 19393
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.891325714285714e-06,
+      "loss": 1.4573,
+      "step": 19394
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.891057142857142e-06,
+      "loss": 1.2732,
+      "step": 19395
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.890788571428572e-06,
+      "loss": 1.0535,
+      "step": 19396
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89052e-06,
+      "loss": 1.4656,
+      "step": 19397
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.890251428571429e-06,
+      "loss": 1.501,
+      "step": 19398
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.889982857142858e-06,
+      "loss": 1.301,
+      "step": 19399
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.889714285714287e-06,
+      "loss": 1.1374,
+      "step": 19400
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.889445714285715e-06,
+      "loss": 1.0359,
+      "step": 19401
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.889177142857143e-06,
+      "loss": 0.6176,
+      "step": 19402
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.888908571428573e-06,
+      "loss": 0.6489,
+      "step": 19403
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88864e-06,
+      "loss": 1.204,
+      "step": 19404
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88837142857143e-06,
+      "loss": 1.0297,
+      "step": 19405
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.888102857142857e-06,
+      "loss": 1.0822,
+      "step": 19406
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.887834285714286e-06,
+      "loss": 0.8197,
+      "step": 19407
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.887565714285714e-06,
+      "loss": 1.0847,
+      "step": 19408
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.887297142857144e-06,
+      "loss": 1.5546,
+      "step": 19409
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.887028571428572e-06,
+      "loss": 1.3264,
+      "step": 19410
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88676e-06,
+      "loss": 1.2246,
+      "step": 19411
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88649142857143e-06,
+      "loss": 1.1283,
+      "step": 19412
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.886222857142858e-06,
+      "loss": 0.9452,
+      "step": 19413
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.885954285714286e-06,
+      "loss": 0.9764,
+      "step": 19414
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.885685714285714e-06,
+      "loss": 1.4795,
+      "step": 19415
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.885417142857144e-06,
+      "loss": 0.9144,
+      "step": 19416
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88514857142857e-06,
+      "loss": 1.2802,
+      "step": 19417
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88488e-06,
+      "loss": 1.0919,
+      "step": 19418
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.884611428571429e-06,
+      "loss": 1.0942,
+      "step": 19419
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.884342857142857e-06,
+      "loss": 0.9794,
+      "step": 19420
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.884074285714285e-06,
+      "loss": 0.8803,
+      "step": 19421
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.883805714285715e-06,
+      "loss": 0.9487,
+      "step": 19422
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.883537142857143e-06,
+      "loss": 1.1707,
+      "step": 19423
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.883268571428571e-06,
+      "loss": 1.3867,
+      "step": 19424
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.883000000000001e-06,
+      "loss": 1.3926,
+      "step": 19425
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88273142857143e-06,
+      "loss": 1.2481,
+      "step": 19426
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.882462857142858e-06,
+      "loss": 1.1651,
+      "step": 19427
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.882194285714286e-06,
+      "loss": 1.0441,
+      "step": 19428
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.881925714285716e-06,
+      "loss": 0.9366,
+      "step": 19429
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.881657142857142e-06,
+      "loss": 1.4468,
+      "step": 19430
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.881388571428572e-06,
+      "loss": 1.2686,
+      "step": 19431
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88112e-06,
+      "loss": 1.1141,
+      "step": 19432
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.88085142857143e-06,
+      "loss": 0.8425,
+      "step": 19433
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.880582857142856e-06,
+      "loss": 1.0676,
+      "step": 19434
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.880314285714286e-06,
+      "loss": 0.9453,
+      "step": 19435
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.880045714285715e-06,
+      "loss": 1.4314,
+      "step": 19436
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.879777142857143e-06,
+      "loss": 1.2059,
+      "step": 19437
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.879508571428573e-06,
+      "loss": 1.4224,
+      "step": 19438
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.87924e-06,
+      "loss": 1.4791,
+      "step": 19439
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.878971428571429e-06,
+      "loss": 1.3135,
+      "step": 19440
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.878702857142857e-06,
+      "loss": 1.1969,
+      "step": 19441
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.878434285714287e-06,
+      "loss": 1.0855,
+      "step": 19442
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.878165714285713e-06,
+      "loss": 1.0975,
+      "step": 19443
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.877897142857143e-06,
+      "loss": 0.9968,
+      "step": 19444
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.877628571428571e-06,
+      "loss": 1.1048,
+      "step": 19445
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.877360000000001e-06,
+      "loss": 0.9499,
+      "step": 19446
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.877091428571428e-06,
+      "loss": 1.1159,
+      "step": 19447
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.876822857142858e-06,
+      "loss": 0.6674,
+      "step": 19448
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.876554285714286e-06,
+      "loss": 1.2527,
+      "step": 19449
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.876285714285714e-06,
+      "loss": 1.1366,
+      "step": 19450
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.876017142857144e-06,
+      "loss": 1.0089,
+      "step": 19451
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.875748571428572e-06,
+      "loss": 1.2043,
+      "step": 19452
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.87548e-06,
+      "loss": 1.1838,
+      "step": 19453
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.875211428571428e-06,
+      "loss": 1.5895,
+      "step": 19454
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.874942857142858e-06,
+      "loss": 1.0879,
+      "step": 19455
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.874674285714286e-06,
+      "loss": 1.2007,
+      "step": 19456
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.874405714285715e-06,
+      "loss": 1.4247,
+      "step": 19457
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.874137142857143e-06,
+      "loss": 0.834,
+      "step": 19458
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.873868571428573e-06,
+      "loss": 1.0403,
+      "step": 19459
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.8736e-06,
+      "loss": 0.9201,
+      "step": 19460
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.873331428571429e-06,
+      "loss": 1.069,
+      "step": 19461
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.873062857142857e-06,
+      "loss": 0.7238,
+      "step": 19462
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.872794285714285e-06,
+      "loss": 1.1191,
+      "step": 19463
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.872525714285715e-06,
+      "loss": 0.9564,
+      "step": 19464
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.872257142857143e-06,
+      "loss": 1.0796,
+      "step": 19465
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.871988571428572e-06,
+      "loss": 1.3328,
+      "step": 19466
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.87172e-06,
+      "loss": 0.9999,
+      "step": 19467
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.87145142857143e-06,
+      "loss": 1.4554,
+      "step": 19468
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.871182857142858e-06,
+      "loss": 1.2549,
+      "step": 19469
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.870914285714286e-06,
+      "loss": 1.4955,
+      "step": 19470
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.870645714285714e-06,
+      "loss": 1.5159,
+      "step": 19471
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.870377142857144e-06,
+      "loss": 1.1812,
+      "step": 19472
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.87010857142857e-06,
+      "loss": 1.4947,
+      "step": 19473
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.86984e-06,
+      "loss": 0.8967,
+      "step": 19474
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.869571428571429e-06,
+      "loss": 1.2531,
+      "step": 19475
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.869302857142857e-06,
+      "loss": 0.965,
+      "step": 19476
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.869034285714287e-06,
+      "loss": 1.0296,
+      "step": 19477
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.868765714285715e-06,
+      "loss": 1.0511,
+      "step": 19478
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.868497142857145e-06,
+      "loss": 0.8324,
+      "step": 19479
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.868228571428571e-06,
+      "loss": 1.5359,
+      "step": 19480
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.867960000000001e-06,
+      "loss": 1.3549,
+      "step": 19481
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.86769142857143e-06,
+      "loss": 1.3904,
+      "step": 19482
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.867422857142857e-06,
+      "loss": 1.0756,
+      "step": 19483
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.867154285714286e-06,
+      "loss": 1.2385,
+      "step": 19484
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.866885714285715e-06,
+      "loss": 0.8514,
+      "step": 19485
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.866617142857142e-06,
+      "loss": 0.8543,
+      "step": 19486
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.866348571428572e-06,
+      "loss": 1.0437,
+      "step": 19487
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.86608e-06,
+      "loss": 1.3843,
+      "step": 19488
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.865811428571428e-06,
+      "loss": 1.084,
+      "step": 19489
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.865542857142858e-06,
+      "loss": 0.916,
+      "step": 19490
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.865274285714286e-06,
+      "loss": 1.3009,
+      "step": 19491
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.865005714285716e-06,
+      "loss": 0.9768,
+      "step": 19492
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.864737142857143e-06,
+      "loss": 0.8908,
+      "step": 19493
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.864468571428572e-06,
+      "loss": 0.9467,
+      "step": 19494
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.8642e-06,
+      "loss": 1.2272,
+      "step": 19495
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.863931428571429e-06,
+      "loss": 1.1651,
+      "step": 19496
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.863662857142857e-06,
+      "loss": 1.3428,
+      "step": 19497
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.863394285714287e-06,
+      "loss": 1.71,
+      "step": 19498
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.863125714285713e-06,
+      "loss": 1.0166,
+      "step": 19499
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.862857142857143e-06,
+      "loss": 1.2419,
+      "step": 19500
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.862588571428571e-06,
+      "loss": 0.7186,
+      "step": 19501
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.862320000000001e-06,
+      "loss": 1.7039,
+      "step": 19502
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.86205142857143e-06,
+      "loss": 1.949,
+      "step": 19503
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.861782857142858e-06,
+      "loss": 0.8052,
+      "step": 19504
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.861514285714287e-06,
+      "loss": 1.1855,
+      "step": 19505
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.861245714285714e-06,
+      "loss": 1.0817,
+      "step": 19506
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.860977142857144e-06,
+      "loss": 1.2172,
+      "step": 19507
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.860708571428572e-06,
+      "loss": 0.8294,
+      "step": 19508
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.86044e-06,
+      "loss": 0.9149,
+      "step": 19509
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.860171428571428e-06,
+      "loss": 1.1176,
+      "step": 19510
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.859902857142858e-06,
+      "loss": 1.145,
+      "step": 19511
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.859634285714285e-06,
+      "loss": 0.8492,
+      "step": 19512
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.859365714285715e-06,
+      "loss": 0.8897,
+      "step": 19513
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.859097142857143e-06,
+      "loss": 1.3475,
+      "step": 19514
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.858828571428573e-06,
+      "loss": 0.7217,
+      "step": 19515
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.85856e-06,
+      "loss": 1.1691,
+      "step": 19516
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.858291428571429e-06,
+      "loss": 1.2216,
+      "step": 19517
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.858022857142859e-06,
+      "loss": 1.056,
+      "step": 19518
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.857754285714285e-06,
+      "loss": 1.3277,
+      "step": 19519
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.857485714285715e-06,
+      "loss": 1.0502,
+      "step": 19520
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.857217142857143e-06,
+      "loss": 0.9797,
+      "step": 19521
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.856948571428572e-06,
+      "loss": 1.0556,
+      "step": 19522
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.85668e-06,
+      "loss": 0.9611,
+      "step": 19523
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.85641142857143e-06,
+      "loss": 1.3347,
+      "step": 19524
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.856142857142856e-06,
+      "loss": 1.2385,
+      "step": 19525
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.855874285714286e-06,
+      "loss": 1.4228,
+      "step": 19526
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.855605714285714e-06,
+      "loss": 1.2143,
+      "step": 19527
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.855337142857144e-06,
+      "loss": 1.1654,
+      "step": 19528
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.855068571428572e-06,
+      "loss": 1.0741,
+      "step": 19529
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.8548e-06,
+      "loss": 1.305,
+      "step": 19530
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.85453142857143e-06,
+      "loss": 1.3798,
+      "step": 19531
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.854262857142857e-06,
+      "loss": 1.2291,
+      "step": 19532
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.853994285714287e-06,
+      "loss": 1.0127,
+      "step": 19533
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.853725714285715e-06,
+      "loss": 1.744,
+      "step": 19534
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.853457142857143e-06,
+      "loss": 1.3457,
+      "step": 19535
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.853188571428571e-06,
+      "loss": 1.0591,
+      "step": 19536
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.852920000000001e-06,
+      "loss": 1.2976,
+      "step": 19537
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.852651428571429e-06,
+      "loss": 0.9509,
+      "step": 19538
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.852382857142857e-06,
+      "loss": 1.1079,
+      "step": 19539
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.852114285714285e-06,
+      "loss": 0.9741,
+      "step": 19540
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.851845714285715e-06,
+      "loss": 1.0812,
+      "step": 19541
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.851577142857144e-06,
+      "loss": 0.8275,
+      "step": 19542
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.851308571428572e-06,
+      "loss": 0.9473,
+      "step": 19543
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.851040000000002e-06,
+      "loss": 0.9338,
+      "step": 19544
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.850771428571428e-06,
+      "loss": 1.1726,
+      "step": 19545
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.850502857142858e-06,
+      "loss": 0.9362,
+      "step": 19546
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.850234285714286e-06,
+      "loss": 1.434,
+      "step": 19547
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.849965714285714e-06,
+      "loss": 1.3287,
+      "step": 19548
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.849697142857142e-06,
+      "loss": 0.781,
+      "step": 19549
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.849428571428572e-06,
+      "loss": 1.3006,
+      "step": 19550
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.84916e-06,
+      "loss": 1.1509,
+      "step": 19551
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.848891428571429e-06,
+      "loss": 0.7737,
+      "step": 19552
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.848622857142857e-06,
+      "loss": 1.1501,
+      "step": 19553
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.848354285714287e-06,
+      "loss": 1.479,
+      "step": 19554
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.848085714285713e-06,
+      "loss": 1.1614,
+      "step": 19555
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.847817142857143e-06,
+      "loss": 1.2135,
+      "step": 19556
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.847548571428573e-06,
+      "loss": 1.2483,
+      "step": 19557
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.84728e-06,
+      "loss": 1.0772,
+      "step": 19558
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.84701142857143e-06,
+      "loss": 1.2961,
+      "step": 19559
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.846742857142857e-06,
+      "loss": 1.2491,
+      "step": 19560
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.846474285714287e-06,
+      "loss": 1.0865,
+      "step": 19561
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.846205714285714e-06,
+      "loss": 1.3717,
+      "step": 19562
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.845937142857144e-06,
+      "loss": 1.4645,
+      "step": 19563
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.845668571428572e-06,
+      "loss": 1.2726,
+      "step": 19564
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.8454e-06,
+      "loss": 0.9491,
+      "step": 19565
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.845131428571428e-06,
+      "loss": 1.3912,
+      "step": 19566
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.844862857142858e-06,
+      "loss": 1.2127,
+      "step": 19567
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.844594285714285e-06,
+      "loss": 1.3135,
+      "step": 19568
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.844325714285714e-06,
+      "loss": 1.0573,
+      "step": 19569
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.844057142857144e-06,
+      "loss": 1.2243,
+      "step": 19570
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.84378857142857e-06,
+      "loss": 1.0064,
+      "step": 19571
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.84352e-06,
+      "loss": 1.5454,
+      "step": 19572
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.843251428571429e-06,
+      "loss": 1.15,
+      "step": 19573
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.842982857142859e-06,
+      "loss": 0.9038,
+      "step": 19574
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.842714285714285e-06,
+      "loss": 1.4878,
+      "step": 19575
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.842445714285715e-06,
+      "loss": 1.3743,
+      "step": 19576
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.842177142857143e-06,
+      "loss": 1.0335,
+      "step": 19577
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.841908571428571e-06,
+      "loss": 0.8871,
+      "step": 19578
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.84164e-06,
+      "loss": 1.5971,
+      "step": 19579
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.84137142857143e-06,
+      "loss": 1.2149,
+      "step": 19580
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.841102857142856e-06,
+      "loss": 1.1631,
+      "step": 19581
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.840834285714286e-06,
+      "loss": 0.9525,
+      "step": 19582
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.840565714285716e-06,
+      "loss": 1.0664,
+      "step": 19583
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.840297142857144e-06,
+      "loss": 0.947,
+      "step": 19584
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.840028571428572e-06,
+      "loss": 1.1983,
+      "step": 19585
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83976e-06,
+      "loss": 1.5271,
+      "step": 19586
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83949142857143e-06,
+      "loss": 1.0926,
+      "step": 19587
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.839222857142857e-06,
+      "loss": 1.1696,
+      "step": 19588
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.838954285714286e-06,
+      "loss": 1.2235,
+      "step": 19589
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.838685714285715e-06,
+      "loss": 0.9325,
+      "step": 19590
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.838417142857143e-06,
+      "loss": 1.0975,
+      "step": 19591
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.838148571428571e-06,
+      "loss": 1.0642,
+      "step": 19592
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83788e-06,
+      "loss": 1.0114,
+      "step": 19593
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.837611428571427e-06,
+      "loss": 1.478,
+      "step": 19594
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.837342857142857e-06,
+      "loss": 1.2295,
+      "step": 19595
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.837074285714287e-06,
+      "loss": 1.2479,
+      "step": 19596
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.836805714285715e-06,
+      "loss": 1.4252,
+      "step": 19597
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.836537142857143e-06,
+      "loss": 1.3249,
+      "step": 19598
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.836268571428572e-06,
+      "loss": 1.1036,
+      "step": 19599
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.836000000000001e-06,
+      "loss": 1.026,
+      "step": 19600
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.835731428571428e-06,
+      "loss": 0.64,
+      "step": 19601
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.835462857142858e-06,
+      "loss": 1.2711,
+      "step": 19602
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.835194285714286e-06,
+      "loss": 1.4647,
+      "step": 19603
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.834925714285714e-06,
+      "loss": 1.2287,
+      "step": 19604
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.834657142857142e-06,
+      "loss": 1.1335,
+      "step": 19605
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.834388571428572e-06,
+      "loss": 1.2514,
+      "step": 19606
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83412e-06,
+      "loss": 1.184,
+      "step": 19607
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.833851428571429e-06,
+      "loss": 1.0116,
+      "step": 19608
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.833582857142858e-06,
+      "loss": 1.5807,
+      "step": 19609
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.833314285714287e-06,
+      "loss": 1.1172,
+      "step": 19610
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.833045714285715e-06,
+      "loss": 1.3168,
+      "step": 19611
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.832777142857143e-06,
+      "loss": 1.0234,
+      "step": 19612
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.832508571428573e-06,
+      "loss": 0.7411,
+      "step": 19613
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83224e-06,
+      "loss": 1.0548,
+      "step": 19614
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83197142857143e-06,
+      "loss": 1.0721,
+      "step": 19615
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.831702857142857e-06,
+      "loss": 1.165,
+      "step": 19616
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.831434285714286e-06,
+      "loss": 1.0168,
+      "step": 19617
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.831165714285714e-06,
+      "loss": 1.1881,
+      "step": 19618
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.830897142857144e-06,
+      "loss": 1.0585,
+      "step": 19619
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.830628571428572e-06,
+      "loss": 0.9903,
+      "step": 19620
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83036e-06,
+      "loss": 0.9718,
+      "step": 19621
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.83009142857143e-06,
+      "loss": 1.3794,
+      "step": 19622
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.829822857142858e-06,
+      "loss": 0.8712,
+      "step": 19623
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.829554285714286e-06,
+      "loss": 1.5411,
+      "step": 19624
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.829285714285714e-06,
+      "loss": 0.7537,
+      "step": 19625
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.829017142857144e-06,
+      "loss": 0.9172,
+      "step": 19626
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.82874857142857e-06,
+      "loss": 1.3538,
+      "step": 19627
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.82848e-06,
+      "loss": 1.5356,
+      "step": 19628
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.828211428571429e-06,
+      "loss": 1.0561,
+      "step": 19629
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.827942857142859e-06,
+      "loss": 1.2858,
+      "step": 19630
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.827674285714285e-06,
+      "loss": 0.6846,
+      "step": 19631
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.827405714285715e-06,
+      "loss": 0.9576,
+      "step": 19632
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.827137142857143e-06,
+      "loss": 1.5532,
+      "step": 19633
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.826868571428571e-06,
+      "loss": 1.1157,
+      "step": 19634
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.826600000000001e-06,
+      "loss": 1.0895,
+      "step": 19635
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.82633142857143e-06,
+      "loss": 1.1229,
+      "step": 19636
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.826062857142858e-06,
+      "loss": 0.8735,
+      "step": 19637
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.825794285714286e-06,
+      "loss": 1.3559,
+      "step": 19638
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.825525714285716e-06,
+      "loss": 1.0299,
+      "step": 19639
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.825257142857142e-06,
+      "loss": 1.3428,
+      "step": 19640
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.824988571428572e-06,
+      "loss": 0.9109,
+      "step": 19641
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.82472e-06,
+      "loss": 1.3688,
+      "step": 19642
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.82445142857143e-06,
+      "loss": 1.0922,
+      "step": 19643
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.824182857142856e-06,
+      "loss": 1.1716,
+      "step": 19644
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.823914285714286e-06,
+      "loss": 1.1672,
+      "step": 19645
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.823645714285715e-06,
+      "loss": 0.9075,
+      "step": 19646
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.823377142857143e-06,
+      "loss": 1.2807,
+      "step": 19647
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.823108571428573e-06,
+      "loss": 1.1982,
+      "step": 19648
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.82284e-06,
+      "loss": 1.1817,
+      "step": 19649
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.822571428571429e-06,
+      "loss": 0.9824,
+      "step": 19650
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.822302857142857e-06,
+      "loss": 1.3911,
+      "step": 19651
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.822034285714287e-06,
+      "loss": 1.6489,
+      "step": 19652
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.821765714285715e-06,
+      "loss": 1.3677,
+      "step": 19653
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.821497142857143e-06,
+      "loss": 0.9638,
+      "step": 19654
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.821228571428571e-06,
+      "loss": 0.9686,
+      "step": 19655
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.820960000000001e-06,
+      "loss": 0.8821,
+      "step": 19656
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.820691428571428e-06,
+      "loss": 1.3298,
+      "step": 19657
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.820422857142858e-06,
+      "loss": 1.4776,
+      "step": 19658
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.820154285714286e-06,
+      "loss": 1.239,
+      "step": 19659
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.819885714285714e-06,
+      "loss": 1.2413,
+      "step": 19660
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.819617142857142e-06,
+      "loss": 1.0633,
+      "step": 19661
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.819348571428572e-06,
+      "loss": 1.3503,
+      "step": 19662
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.81908e-06,
+      "loss": 0.7694,
+      "step": 19663
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.818811428571428e-06,
+      "loss": 0.9864,
+      "step": 19664
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.818542857142858e-06,
+      "loss": 1.3612,
+      "step": 19665
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.818274285714286e-06,
+      "loss": 1.2551,
+      "step": 19666
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.818005714285715e-06,
+      "loss": 1.5457,
+      "step": 19667
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.817737142857143e-06,
+      "loss": 1.3922,
+      "step": 19668
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.817468571428573e-06,
+      "loss": 0.7535,
+      "step": 19669
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.8172e-06,
+      "loss": 0.9648,
+      "step": 19670
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.816931428571429e-06,
+      "loss": 1.1668,
+      "step": 19671
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.816662857142857e-06,
+      "loss": 0.8715,
+      "step": 19672
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.816394285714285e-06,
+      "loss": 1.3673,
+      "step": 19673
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.816125714285714e-06,
+      "loss": 0.9342,
+      "step": 19674
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.815857142857143e-06,
+      "loss": 1.0032,
+      "step": 19675
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.815588571428573e-06,
+      "loss": 1.2152,
+      "step": 19676
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.81532e-06,
+      "loss": 1.1984,
+      "step": 19677
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.81505142857143e-06,
+      "loss": 1.4836,
+      "step": 19678
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.814782857142858e-06,
+      "loss": 0.8653,
+      "step": 19679
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.814514285714286e-06,
+      "loss": 1.3912,
+      "step": 19680
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.814245714285714e-06,
+      "loss": 1.09,
+      "step": 19681
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.813977142857144e-06,
+      "loss": 0.9153,
+      "step": 19682
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.81370857142857e-06,
+      "loss": 1.164,
+      "step": 19683
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.81344e-06,
+      "loss": 1.2415,
+      "step": 19684
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.813171428571429e-06,
+      "loss": 1.1345,
+      "step": 19685
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.812902857142857e-06,
+      "loss": 0.7496,
+      "step": 19686
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.812634285714285e-06,
+      "loss": 1.1787,
+      "step": 19687
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.812365714285715e-06,
+      "loss": 0.6871,
+      "step": 19688
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.812097142857145e-06,
+      "loss": 1.1831,
+      "step": 19689
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.811828571428571e-06,
+      "loss": 1.1614,
+      "step": 19690
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.811560000000001e-06,
+      "loss": 1.4714,
+      "step": 19691
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.81129142857143e-06,
+      "loss": 1.0043,
+      "step": 19692
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.811022857142857e-06,
+      "loss": 1.1072,
+      "step": 19693
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.810754285714286e-06,
+      "loss": 1.4523,
+      "step": 19694
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.810485714285715e-06,
+      "loss": 0.9815,
+      "step": 19695
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.810217142857142e-06,
+      "loss": 1.1786,
+      "step": 19696
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.809948571428572e-06,
+      "loss": 1.2939,
+      "step": 19697
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.80968e-06,
+      "loss": 0.732,
+      "step": 19698
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.809411428571428e-06,
+      "loss": 1.3902,
+      "step": 19699
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.809142857142856e-06,
+      "loss": 1.378,
+      "step": 19700
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.808874285714286e-06,
+      "loss": 0.8952,
+      "step": 19701
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.808605714285716e-06,
+      "loss": 1.0859,
+      "step": 19702
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.808337142857143e-06,
+      "loss": 0.9394,
+      "step": 19703
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.808068571428572e-06,
+      "loss": 1.0363,
+      "step": 19704
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.8078e-06,
+      "loss": 0.9329,
+      "step": 19705
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.807531428571429e-06,
+      "loss": 1.5661,
+      "step": 19706
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.807262857142857e-06,
+      "loss": 1.3351,
+      "step": 19707
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.806994285714287e-06,
+      "loss": 0.7833,
+      "step": 19708
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.806725714285713e-06,
+      "loss": 1.1913,
+      "step": 19709
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.806457142857143e-06,
+      "loss": 1.7266,
+      "step": 19710
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.806188571428571e-06,
+      "loss": 1.5403,
+      "step": 19711
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.805920000000001e-06,
+      "loss": 1.0534,
+      "step": 19712
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.805651428571428e-06,
+      "loss": 1.417,
+      "step": 19713
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.805382857142858e-06,
+      "loss": 1.4313,
+      "step": 19714
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.805114285714287e-06,
+      "loss": 1.277,
+      "step": 19715
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.804845714285714e-06,
+      "loss": 1.2781,
+      "step": 19716
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.804577142857144e-06,
+      "loss": 1.1264,
+      "step": 19717
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.804308571428572e-06,
+      "loss": 1.0453,
+      "step": 19718
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.80404e-06,
+      "loss": 0.7165,
+      "step": 19719
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.803771428571428e-06,
+      "loss": 1.0176,
+      "step": 19720
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.803502857142858e-06,
+      "loss": 0.8695,
+      "step": 19721
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.803234285714285e-06,
+      "loss": 1.2597,
+      "step": 19722
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.802965714285715e-06,
+      "loss": 1.1894,
+      "step": 19723
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.802697142857143e-06,
+      "loss": 1.3573,
+      "step": 19724
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.802428571428573e-06,
+      "loss": 0.8657,
+      "step": 19725
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.802159999999999e-06,
+      "loss": 1.2215,
+      "step": 19726
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.801891428571429e-06,
+      "loss": 1.3984,
+      "step": 19727
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.801622857142859e-06,
+      "loss": 1.4335,
+      "step": 19728
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.801354285714285e-06,
+      "loss": 1.2109,
+      "step": 19729
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.801085714285715e-06,
+      "loss": 1.0496,
+      "step": 19730
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.800817142857143e-06,
+      "loss": 1.2165,
+      "step": 19731
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.800548571428572e-06,
+      "loss": 1.1681,
+      "step": 19732
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.80028e-06,
+      "loss": 0.9381,
+      "step": 19733
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.80001142857143e-06,
+      "loss": 1.0459,
+      "step": 19734
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.799742857142858e-06,
+      "loss": 1.2678,
+      "step": 19735
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.799474285714286e-06,
+      "loss": 1.3309,
+      "step": 19736
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.799205714285714e-06,
+      "loss": 0.9742,
+      "step": 19737
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.798937142857144e-06,
+      "loss": 1.567,
+      "step": 19738
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.79866857142857e-06,
+      "loss": 1.0516,
+      "step": 19739
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.7984e-06,
+      "loss": 1.1861,
+      "step": 19740
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.79813142857143e-06,
+      "loss": 0.9796,
+      "step": 19741
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.797862857142857e-06,
+      "loss": 1.1577,
+      "step": 19742
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.797594285714287e-06,
+      "loss": 1.2035,
+      "step": 19743
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.797325714285715e-06,
+      "loss": 0.9914,
+      "step": 19744
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.797057142857143e-06,
+      "loss": 0.8314,
+      "step": 19745
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.796788571428571e-06,
+      "loss": 1.2338,
+      "step": 19746
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.796520000000001e-06,
+      "loss": 0.7709,
+      "step": 19747
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.796251428571429e-06,
+      "loss": 1.009,
+      "step": 19748
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.795982857142857e-06,
+      "loss": 0.9185,
+      "step": 19749
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.795714285714285e-06,
+      "loss": 1.3498,
+      "step": 19750
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.795445714285715e-06,
+      "loss": 1.2982,
+      "step": 19751
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.795177142857142e-06,
+      "loss": 1.2231,
+      "step": 19752
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.794908571428572e-06,
+      "loss": 1.2488,
+      "step": 19753
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.794640000000002e-06,
+      "loss": 1.462,
+      "step": 19754
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.794371428571428e-06,
+      "loss": 1.0004,
+      "step": 19755
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.794102857142858e-06,
+      "loss": 1.0992,
+      "step": 19756
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.793834285714286e-06,
+      "loss": 0.9084,
+      "step": 19757
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.793565714285716e-06,
+      "loss": 1.4051,
+      "step": 19758
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.793297142857142e-06,
+      "loss": 1.6625,
+      "step": 19759
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.793028571428572e-06,
+      "loss": 1.1988,
+      "step": 19760
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.79276e-06,
+      "loss": 1.2976,
+      "step": 19761
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.792491428571429e-06,
+      "loss": 1.2894,
+      "step": 19762
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.792222857142857e-06,
+      "loss": 1.2141,
+      "step": 19763
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.791954285714287e-06,
+      "loss": 1.1767,
+      "step": 19764
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.791685714285713e-06,
+      "loss": 1.2604,
+      "step": 19765
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.791417142857143e-06,
+      "loss": 0.8725,
+      "step": 19766
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.791148571428573e-06,
+      "loss": 0.9513,
+      "step": 19767
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.79088e-06,
+      "loss": 1.3892,
+      "step": 19768
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.79061142857143e-06,
+      "loss": 0.8045,
+      "step": 19769
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.790342857142857e-06,
+      "loss": 1.0713,
+      "step": 19770
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.790074285714287e-06,
+      "loss": 1.2252,
+      "step": 19771
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.789805714285714e-06,
+      "loss": 0.7758,
+      "step": 19772
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.789537142857144e-06,
+      "loss": 1.0955,
+      "step": 19773
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.789268571428572e-06,
+      "loss": 1.1788,
+      "step": 19774
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.789e-06,
+      "loss": 0.9411,
+      "step": 19775
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.788731428571428e-06,
+      "loss": 1.1473,
+      "step": 19776
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.788462857142858e-06,
+      "loss": 1.3263,
+      "step": 19777
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.788194285714285e-06,
+      "loss": 1.3265,
+      "step": 19778
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.787925714285714e-06,
+      "loss": 1.4848,
+      "step": 19779
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.787657142857143e-06,
+      "loss": 0.9414,
+      "step": 19780
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.787388571428572e-06,
+      "loss": 1.5608,
+      "step": 19781
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.78712e-06,
+      "loss": 1.4796,
+      "step": 19782
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.786851428571429e-06,
+      "loss": 1.2274,
+      "step": 19783
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.786582857142859e-06,
+      "loss": 0.869,
+      "step": 19784
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.786314285714285e-06,
+      "loss": 1.4332,
+      "step": 19785
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.786045714285715e-06,
+      "loss": 1.2021,
+      "step": 19786
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.785777142857143e-06,
+      "loss": 1.263,
+      "step": 19787
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.785508571428571e-06,
+      "loss": 1.8323,
+      "step": 19788
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.78524e-06,
+      "loss": 0.9264,
+      "step": 19789
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.78497142857143e-06,
+      "loss": 1.1509,
+      "step": 19790
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.784702857142856e-06,
+      "loss": 1.4689,
+      "step": 19791
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.784434285714286e-06,
+      "loss": 1.0719,
+      "step": 19792
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.784165714285714e-06,
+      "loss": 1.2833,
+      "step": 19793
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.783897142857144e-06,
+      "loss": 1.3037,
+      "step": 19794
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.783628571428572e-06,
+      "loss": 1.4916,
+      "step": 19795
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.78336e-06,
+      "loss": 1.0135,
+      "step": 19796
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.78309142857143e-06,
+      "loss": 1.7561,
+      "step": 19797
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.782822857142857e-06,
+      "loss": 1.0984,
+      "step": 19798
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.782554285714286e-06,
+      "loss": 1.315,
+      "step": 19799
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.782285714285715e-06,
+      "loss": 1.3377,
+      "step": 19800
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.782017142857143e-06,
+      "loss": 1.1035,
+      "step": 19801
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.781748571428571e-06,
+      "loss": 1.253,
+      "step": 19802
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.78148e-06,
+      "loss": 1.1902,
+      "step": 19803
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.781211428571429e-06,
+      "loss": 1.4329,
+      "step": 19804
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.780942857142857e-06,
+      "loss": 0.9939,
+      "step": 19805
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.780674285714285e-06,
+      "loss": 1.0842,
+      "step": 19806
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.780405714285715e-06,
+      "loss": 1.0626,
+      "step": 19807
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.780137142857143e-06,
+      "loss": 1.5188,
+      "step": 19808
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.779868571428572e-06,
+      "loss": 1.0867,
+      "step": 19809
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.779600000000001e-06,
+      "loss": 1.3744,
+      "step": 19810
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.779331428571428e-06,
+      "loss": 0.8619,
+      "step": 19811
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.779062857142858e-06,
+      "loss": 1.1412,
+      "step": 19812
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.778794285714286e-06,
+      "loss": 1.2003,
+      "step": 19813
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.778525714285714e-06,
+      "loss": 1.3294,
+      "step": 19814
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.778257142857142e-06,
+      "loss": 0.794,
+      "step": 19815
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.777988571428572e-06,
+      "loss": 0.9487,
+      "step": 19816
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.77772e-06,
+      "loss": 1.2345,
+      "step": 19817
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.777451428571429e-06,
+      "loss": 1.5254,
+      "step": 19818
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.777182857142857e-06,
+      "loss": 1.0704,
+      "step": 19819
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.776914285714287e-06,
+      "loss": 0.9427,
+      "step": 19820
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.776645714285715e-06,
+      "loss": 0.9031,
+      "step": 19821
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.776377142857143e-06,
+      "loss": 1.0364,
+      "step": 19822
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.776108571428573e-06,
+      "loss": 1.1056,
+      "step": 19823
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.77584e-06,
+      "loss": 0.9039,
+      "step": 19824
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.77557142857143e-06,
+      "loss": 1.0648,
+      "step": 19825
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.775302857142857e-06,
+      "loss": 0.7701,
+      "step": 19826
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.775034285714287e-06,
+      "loss": 1.3422,
+      "step": 19827
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.774765714285714e-06,
+      "loss": 1.3296,
+      "step": 19828
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.774497142857144e-06,
+      "loss": 1.1836,
+      "step": 19829
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.774228571428572e-06,
+      "loss": 1.1273,
+      "step": 19830
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.77396e-06,
+      "loss": 1.1389,
+      "step": 19831
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.773691428571428e-06,
+      "loss": 0.9801,
+      "step": 19832
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.773422857142858e-06,
+      "loss": 1.3535,
+      "step": 19833
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.773154285714286e-06,
+      "loss": 1.2823,
+      "step": 19834
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.772885714285714e-06,
+      "loss": 0.7636,
+      "step": 19835
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.772617142857144e-06,
+      "loss": 1.4286,
+      "step": 19836
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.77234857142857e-06,
+      "loss": 0.9369,
+      "step": 19837
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.77208e-06,
+      "loss": 0.7774,
+      "step": 19838
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.771811428571429e-06,
+      "loss": 1.2467,
+      "step": 19839
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.771542857142859e-06,
+      "loss": 1.2507,
+      "step": 19840
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.771274285714285e-06,
+      "loss": 1.4283,
+      "step": 19841
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.771005714285715e-06,
+      "loss": 1.7163,
+      "step": 19842
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.770737142857143e-06,
+      "loss": 1.0609,
+      "step": 19843
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.770468571428571e-06,
+      "loss": 0.832,
+      "step": 19844
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.7702e-06,
+      "loss": 1.8017,
+      "step": 19845
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.76993142857143e-06,
+      "loss": 1.4108,
+      "step": 19846
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.769662857142858e-06,
+      "loss": 1.4885,
+      "step": 19847
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.769394285714286e-06,
+      "loss": 1.1971,
+      "step": 19848
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.769125714285716e-06,
+      "loss": 1.3523,
+      "step": 19849
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.768857142857144e-06,
+      "loss": 1.2237,
+      "step": 19850
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.768588571428572e-06,
+      "loss": 1.2276,
+      "step": 19851
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.76832e-06,
+      "loss": 1.3117,
+      "step": 19852
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.76805142857143e-06,
+      "loss": 1.6451,
+      "step": 19853
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.767782857142856e-06,
+      "loss": 1.5247,
+      "step": 19854
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.767514285714286e-06,
+      "loss": 0.7982,
+      "step": 19855
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.767245714285715e-06,
+      "loss": 0.8132,
+      "step": 19856
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.766977142857143e-06,
+      "loss": 1.0264,
+      "step": 19857
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.766708571428571e-06,
+      "loss": 1.309,
+      "step": 19858
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.76644e-06,
+      "loss": 1.0309,
+      "step": 19859
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.766171428571429e-06,
+      "loss": 1.254,
+      "step": 19860
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.765902857142857e-06,
+      "loss": 1.3734,
+      "step": 19861
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.765634285714287e-06,
+      "loss": 1.1795,
+      "step": 19862
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.765365714285715e-06,
+      "loss": 1.1443,
+      "step": 19863
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.765097142857143e-06,
+      "loss": 1.3759,
+      "step": 19864
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.764828571428571e-06,
+      "loss": 1.0325,
+      "step": 19865
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.764560000000001e-06,
+      "loss": 0.872,
+      "step": 19866
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.764291428571428e-06,
+      "loss": 1.089,
+      "step": 19867
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.764022857142858e-06,
+      "loss": 1.08,
+      "step": 19868
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.763754285714286e-06,
+      "loss": 0.9858,
+      "step": 19869
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.763485714285714e-06,
+      "loss": 1.199,
+      "step": 19870
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.763217142857142e-06,
+      "loss": 1.0497,
+      "step": 19871
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.762948571428572e-06,
+      "loss": 1.142,
+      "step": 19872
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.76268e-06,
+      "loss": 1.1696,
+      "step": 19873
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.762411428571428e-06,
+      "loss": 0.9238,
+      "step": 19874
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.762142857142858e-06,
+      "loss": 1.4361,
+      "step": 19875
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.761874285714286e-06,
+      "loss": 1.5075,
+      "step": 19876
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.761605714285715e-06,
+      "loss": 1.2599,
+      "step": 19877
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.761337142857143e-06,
+      "loss": 1.0396,
+      "step": 19878
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.761068571428573e-06,
+      "loss": 0.9029,
+      "step": 19879
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.7608e-06,
+      "loss": 1.3859,
+      "step": 19880
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.760531428571429e-06,
+      "loss": 1.1415,
+      "step": 19881
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.760262857142857e-06,
+      "loss": 1.1393,
+      "step": 19882
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.759994285714285e-06,
+      "loss": 1.2428,
+      "step": 19883
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.759725714285714e-06,
+      "loss": 1.3686,
+      "step": 19884
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.759457142857143e-06,
+      "loss": 1.5223,
+      "step": 19885
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.759188571428572e-06,
+      "loss": 1.4659,
+      "step": 19886
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.75892e-06,
+      "loss": 1.2303,
+      "step": 19887
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.75865142857143e-06,
+      "loss": 1.2606,
+      "step": 19888
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.758382857142858e-06,
+      "loss": 1.0226,
+      "step": 19889
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.758114285714286e-06,
+      "loss": 1.1281,
+      "step": 19890
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.757845714285714e-06,
+      "loss": 1.2038,
+      "step": 19891
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.757577142857144e-06,
+      "loss": 1.3379,
+      "step": 19892
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.75730857142857e-06,
+      "loss": 1.1091,
+      "step": 19893
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.75704e-06,
+      "loss": 1.1438,
+      "step": 19894
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.756771428571429e-06,
+      "loss": 1.2193,
+      "step": 19895
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.756502857142857e-06,
+      "loss": 1.6464,
+      "step": 19896
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.756234285714285e-06,
+      "loss": 0.989,
+      "step": 19897
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.755965714285715e-06,
+      "loss": 1.2155,
+      "step": 19898
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.755697142857143e-06,
+      "loss": 1.6501,
+      "step": 19899
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.755428571428571e-06,
+      "loss": 1.1469,
+      "step": 19900
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.755160000000001e-06,
+      "loss": 1.1897,
+      "step": 19901
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.75489142857143e-06,
+      "loss": 0.7931,
+      "step": 19902
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.754622857142857e-06,
+      "loss": 1.2273,
+      "step": 19903
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.754354285714286e-06,
+      "loss": 1.1379,
+      "step": 19904
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.754085714285715e-06,
+      "loss": 0.8785,
+      "step": 19905
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.753817142857142e-06,
+      "loss": 1.418,
+      "step": 19906
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.753548571428572e-06,
+      "loss": 0.8218,
+      "step": 19907
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.75328e-06,
+      "loss": 1.1081,
+      "step": 19908
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.75301142857143e-06,
+      "loss": 1.537,
+      "step": 19909
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.752742857142856e-06,
+      "loss": 0.8333,
+      "step": 19910
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.752474285714286e-06,
+      "loss": 1.4313,
+      "step": 19911
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.752205714285714e-06,
+      "loss": 1.5674,
+      "step": 19912
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.751937142857143e-06,
+      "loss": 1.5051,
+      "step": 19913
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.751668571428572e-06,
+      "loss": 1.2398,
+      "step": 19914
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.7514e-06,
+      "loss": 0.8665,
+      "step": 19915
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.751131428571429e-06,
+      "loss": 0.8489,
+      "step": 19916
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.750862857142857e-06,
+      "loss": 0.925,
+      "step": 19917
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.750594285714287e-06,
+      "loss": 1.2863,
+      "step": 19918
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.750325714285713e-06,
+      "loss": 1.1003,
+      "step": 19919
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.750057142857143e-06,
+      "loss": 1.5819,
+      "step": 19920
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.749788571428571e-06,
+      "loss": 1.6271,
+      "step": 19921
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.749520000000001e-06,
+      "loss": 1.4285,
+      "step": 19922
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.749251428571428e-06,
+      "loss": 1.3025,
+      "step": 19923
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.748982857142858e-06,
+      "loss": 1.0492,
+      "step": 19924
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.748714285714286e-06,
+      "loss": 1.0368,
+      "step": 19925
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.748445714285714e-06,
+      "loss": 1.1828,
+      "step": 19926
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.748177142857144e-06,
+      "loss": 1.1751,
+      "step": 19927
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.747908571428572e-06,
+      "loss": 0.8163,
+      "step": 19928
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.74764e-06,
+      "loss": 1.0549,
+      "step": 19929
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.747371428571428e-06,
+      "loss": 1.7444,
+      "step": 19930
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.747102857142858e-06,
+      "loss": 0.961,
+      "step": 19931
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.746834285714286e-06,
+      "loss": 0.9786,
+      "step": 19932
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.746565714285715e-06,
+      "loss": 1.2958,
+      "step": 19933
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.746297142857143e-06,
+      "loss": 1.0802,
+      "step": 19934
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.746028571428573e-06,
+      "loss": 1.0998,
+      "step": 19935
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.745759999999999e-06,
+      "loss": 1.3445,
+      "step": 19936
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.745491428571429e-06,
+      "loss": 1.2642,
+      "step": 19937
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.745222857142857e-06,
+      "loss": 1.1608,
+      "step": 19938
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.744954285714285e-06,
+      "loss": 0.8764,
+      "step": 19939
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.744685714285715e-06,
+      "loss": 1.1342,
+      "step": 19940
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.744417142857143e-06,
+      "loss": 0.6799,
+      "step": 19941
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.744148571428572e-06,
+      "loss": 1.1031,
+      "step": 19942
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.74388e-06,
+      "loss": 1.3237,
+      "step": 19943
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.74361142857143e-06,
+      "loss": 1.2194,
+      "step": 19944
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.743342857142858e-06,
+      "loss": 1.229,
+      "step": 19945
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.743074285714286e-06,
+      "loss": 0.9301,
+      "step": 19946
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.742805714285714e-06,
+      "loss": 1.2125,
+      "step": 19947
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.742537142857144e-06,
+      "loss": 1.3529,
+      "step": 19948
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.74226857142857e-06,
+      "loss": 1.7776,
+      "step": 19949
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.742e-06,
+      "loss": 0.9298,
+      "step": 19950
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.741731428571429e-06,
+      "loss": 0.6289,
+      "step": 19951
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.741462857142857e-06,
+      "loss": 1.2526,
+      "step": 19952
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.741194285714287e-06,
+      "loss": 0.7133,
+      "step": 19953
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.740925714285715e-06,
+      "loss": 1.1675,
+      "step": 19954
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.740657142857145e-06,
+      "loss": 1.1194,
+      "step": 19955
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.740388571428571e-06,
+      "loss": 0.8496,
+      "step": 19956
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.740120000000001e-06,
+      "loss": 0.9537,
+      "step": 19957
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.739851428571429e-06,
+      "loss": 1.1398,
+      "step": 19958
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.739582857142857e-06,
+      "loss": 1.4701,
+      "step": 19959
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.739314285714285e-06,
+      "loss": 1.2967,
+      "step": 19960
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.739045714285715e-06,
+      "loss": 1.1995,
+      "step": 19961
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.738777142857142e-06,
+      "loss": 1.1178,
+      "step": 19962
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.738508571428572e-06,
+      "loss": 1.2882,
+      "step": 19963
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.73824e-06,
+      "loss": 0.8782,
+      "step": 19964
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.737971428571428e-06,
+      "loss": 1.0711,
+      "step": 19965
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.737702857142858e-06,
+      "loss": 1.4424,
+      "step": 19966
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.737434285714286e-06,
+      "loss": 1.196,
+      "step": 19967
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.737165714285716e-06,
+      "loss": 1.5501,
+      "step": 19968
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.736897142857142e-06,
+      "loss": 1.2145,
+      "step": 19969
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.736628571428572e-06,
+      "loss": 1.2312,
+      "step": 19970
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.73636e-06,
+      "loss": 1.2408,
+      "step": 19971
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.736091428571429e-06,
+      "loss": 1.306,
+      "step": 19972
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.735822857142857e-06,
+      "loss": 1.1008,
+      "step": 19973
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.735554285714287e-06,
+      "loss": 1.3137,
+      "step": 19974
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.735285714285713e-06,
+      "loss": 0.8822,
+      "step": 19975
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.735017142857143e-06,
+      "loss": 0.9607,
+      "step": 19976
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.734748571428571e-06,
+      "loss": 1.0729,
+      "step": 19977
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.734480000000001e-06,
+      "loss": 0.8724,
+      "step": 19978
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.73421142857143e-06,
+      "loss": 1.3193,
+      "step": 19979
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.733942857142857e-06,
+      "loss": 1.4702,
+      "step": 19980
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.733674285714287e-06,
+      "loss": 1.1961,
+      "step": 19981
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.733405714285714e-06,
+      "loss": 1.1913,
+      "step": 19982
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.733137142857144e-06,
+      "loss": 1.4306,
+      "step": 19983
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.732868571428572e-06,
+      "loss": 0.9921,
+      "step": 19984
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.7326e-06,
+      "loss": 1.0106,
+      "step": 19985
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.732331428571428e-06,
+      "loss": 1.2432,
+      "step": 19986
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.732062857142858e-06,
+      "loss": 1.5136,
+      "step": 19987
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.731794285714285e-06,
+      "loss": 1.3083,
+      "step": 19988
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.731525714285714e-06,
+      "loss": 1.1976,
+      "step": 19989
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.731257142857143e-06,
+      "loss": 0.9009,
+      "step": 19990
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.730988571428572e-06,
+      "loss": 1.1464,
+      "step": 19991
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.73072e-06,
+      "loss": 1.2315,
+      "step": 19992
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.730451428571429e-06,
+      "loss": 1.3837,
+      "step": 19993
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.730182857142859e-06,
+      "loss": 1.2032,
+      "step": 19994
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.729914285714285e-06,
+      "loss": 1.5061,
+      "step": 19995
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.729645714285715e-06,
+      "loss": 1.9405,
+      "step": 19996
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.729377142857143e-06,
+      "loss": 0.6869,
+      "step": 19997
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.729108571428571e-06,
+      "loss": 1.2015,
+      "step": 19998
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.72884e-06,
+      "loss": 1.2307,
+      "step": 19999
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.72857142857143e-06,
+      "loss": 1.3022,
+      "step": 20000
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.728302857142858e-06,
+      "loss": 0.8802,
+      "step": 20001
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.728034285714286e-06,
+      "loss": 1.1495,
+      "step": 20002
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.727765714285714e-06,
+      "loss": 1.0215,
+      "step": 20003
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.727497142857144e-06,
+      "loss": 1.1715,
+      "step": 20004
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.727228571428572e-06,
+      "loss": 1.1419,
+      "step": 20005
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.72696e-06,
+      "loss": 0.8518,
+      "step": 20006
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.72669142857143e-06,
+      "loss": 1.3579,
+      "step": 20007
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.726422857142857e-06,
+      "loss": 1.0584,
+      "step": 20008
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.726154285714286e-06,
+      "loss": 1.128,
+      "step": 20009
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.725885714285715e-06,
+      "loss": 1.4566,
+      "step": 20010
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.725617142857143e-06,
+      "loss": 1.1068,
+      "step": 20011
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.725348571428571e-06,
+      "loss": 1.1568,
+      "step": 20012
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.72508e-06,
+      "loss": 1.4196,
+      "step": 20013
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.724811428571429e-06,
+      "loss": 0.9312,
+      "step": 20014
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.724542857142857e-06,
+      "loss": 1.3217,
+      "step": 20015
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.724274285714285e-06,
+      "loss": 1.5434,
+      "step": 20016
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.724005714285715e-06,
+      "loss": 1.5451,
+      "step": 20017
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.723737142857143e-06,
+      "loss": 1.2489,
+      "step": 20018
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.723468571428572e-06,
+      "loss": 1.1723,
+      "step": 20019
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.723200000000001e-06,
+      "loss": 1.0656,
+      "step": 20020
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.722931428571428e-06,
+      "loss": 1.1521,
+      "step": 20021
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.722662857142858e-06,
+      "loss": 1.6103,
+      "step": 20022
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.722394285714286e-06,
+      "loss": 1.314,
+      "step": 20023
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.722125714285714e-06,
+      "loss": 1.0191,
+      "step": 20024
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.721857142857142e-06,
+      "loss": 1.2507,
+      "step": 20025
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.721588571428572e-06,
+      "loss": 1.5258,
+      "step": 20026
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.72132e-06,
+      "loss": 1.0109,
+      "step": 20027
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.721051428571429e-06,
+      "loss": 1.2104,
+      "step": 20028
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.720782857142857e-06,
+      "loss": 1.2103,
+      "step": 20029
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.720514285714287e-06,
+      "loss": 1.3546,
+      "step": 20030
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.720245714285715e-06,
+      "loss": 1.4013,
+      "step": 20031
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.719977142857143e-06,
+      "loss": 1.4273,
+      "step": 20032
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.719708571428573e-06,
+      "loss": 0.7447,
+      "step": 20033
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.71944e-06,
+      "loss": 1.3494,
+      "step": 20034
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.71917142857143e-06,
+      "loss": 0.7234,
+      "step": 20035
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.718902857142857e-06,
+      "loss": 1.2426,
+      "step": 20036
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.718634285714287e-06,
+      "loss": 1.2067,
+      "step": 20037
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.718365714285714e-06,
+      "loss": 1.3322,
+      "step": 20038
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.718097142857144e-06,
+      "loss": 0.859,
+      "step": 20039
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.717828571428572e-06,
+      "loss": 0.73,
+      "step": 20040
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71756e-06,
+      "loss": 1.363,
+      "step": 20041
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.717291428571428e-06,
+      "loss": 1.43,
+      "step": 20042
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.717022857142858e-06,
+      "loss": 1.1793,
+      "step": 20043
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.716754285714286e-06,
+      "loss": 1.0837,
+      "step": 20044
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.716485714285714e-06,
+      "loss": 1.1114,
+      "step": 20045
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.716217142857144e-06,
+      "loss": 1.3206,
+      "step": 20046
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71594857142857e-06,
+      "loss": 1.4052,
+      "step": 20047
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71568e-06,
+      "loss": 1.0849,
+      "step": 20048
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.715411428571429e-06,
+      "loss": 0.9766,
+      "step": 20049
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.715142857142859e-06,
+      "loss": 1.2565,
+      "step": 20050
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.714874285714285e-06,
+      "loss": 0.889,
+      "step": 20051
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.714605714285715e-06,
+      "loss": 0.9555,
+      "step": 20052
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.714337142857143e-06,
+      "loss": 1.463,
+      "step": 20053
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.714068571428571e-06,
+      "loss": 1.0437,
+      "step": 20054
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.7138e-06,
+      "loss": 1.5246,
+      "step": 20055
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71353142857143e-06,
+      "loss": 1.5978,
+      "step": 20056
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.713262857142858e-06,
+      "loss": 0.8683,
+      "step": 20057
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.712994285714286e-06,
+      "loss": 1.541,
+      "step": 20058
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.712725714285716e-06,
+      "loss": 1.515,
+      "step": 20059
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.712457142857144e-06,
+      "loss": 1.1183,
+      "step": 20060
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.712188571428572e-06,
+      "loss": 0.7853,
+      "step": 20061
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71192e-06,
+      "loss": 1.1707,
+      "step": 20062
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71165142857143e-06,
+      "loss": 0.9193,
+      "step": 20063
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.711382857142856e-06,
+      "loss": 0.7649,
+      "step": 20064
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.711114285714286e-06,
+      "loss": 1.1307,
+      "step": 20065
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.710845714285714e-06,
+      "loss": 1.2622,
+      "step": 20066
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.710577142857143e-06,
+      "loss": 1.0857,
+      "step": 20067
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.710308571428571e-06,
+      "loss": 1.4625,
+      "step": 20068
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71004e-06,
+      "loss": 1.8545,
+      "step": 20069
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.709771428571429e-06,
+      "loss": 1.1845,
+      "step": 20070
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.709502857142857e-06,
+      "loss": 1.293,
+      "step": 20071
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.709234285714287e-06,
+      "loss": 0.9764,
+      "step": 20072
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.708965714285715e-06,
+      "loss": 1.2663,
+      "step": 20073
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.708697142857143e-06,
+      "loss": 1.4204,
+      "step": 20074
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.708428571428571e-06,
+      "loss": 0.8573,
+      "step": 20075
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.708160000000001e-06,
+      "loss": 1.3829,
+      "step": 20076
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.707891428571428e-06,
+      "loss": 0.6258,
+      "step": 20077
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.707622857142858e-06,
+      "loss": 1.0124,
+      "step": 20078
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.707354285714286e-06,
+      "loss": 1.1111,
+      "step": 20079
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.707085714285714e-06,
+      "loss": 1.3807,
+      "step": 20080
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.706817142857142e-06,
+      "loss": 1.1573,
+      "step": 20081
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.706548571428572e-06,
+      "loss": 1.796,
+      "step": 20082
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.70628e-06,
+      "loss": 1.2701,
+      "step": 20083
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.706011428571428e-06,
+      "loss": 1.3452,
+      "step": 20084
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.705742857142858e-06,
+      "loss": 1.1254,
+      "step": 20085
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.705474285714286e-06,
+      "loss": 1.3994,
+      "step": 20086
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.705205714285715e-06,
+      "loss": 1.572,
+      "step": 20087
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.704937142857143e-06,
+      "loss": 0.973,
+      "step": 20088
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.704668571428573e-06,
+      "loss": 1.3504,
+      "step": 20089
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.7044e-06,
+      "loss": 1.6418,
+      "step": 20090
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.704131428571429e-06,
+      "loss": 1.3913,
+      "step": 20091
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.703862857142857e-06,
+      "loss": 1.0762,
+      "step": 20092
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.703594285714285e-06,
+      "loss": 1.2802,
+      "step": 20093
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.703325714285714e-06,
+      "loss": 1.2529,
+      "step": 20094
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.703057142857143e-06,
+      "loss": 1.0876,
+      "step": 20095
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.702788571428572e-06,
+      "loss": 1.1991,
+      "step": 20096
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.70252e-06,
+      "loss": 1.3707,
+      "step": 20097
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.70225142857143e-06,
+      "loss": 1.6012,
+      "step": 20098
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.701982857142858e-06,
+      "loss": 0.889,
+      "step": 20099
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.701714285714286e-06,
+      "loss": 1.0611,
+      "step": 20100
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.701445714285714e-06,
+      "loss": 0.8604,
+      "step": 20101
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.701177142857144e-06,
+      "loss": 0.9991,
+      "step": 20102
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.70090857142857e-06,
+      "loss": 0.7928,
+      "step": 20103
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.70064e-06,
+      "loss": 1.0244,
+      "step": 20104
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.700371428571429e-06,
+      "loss": 0.7392,
+      "step": 20105
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.700102857142858e-06,
+      "loss": 1.1295,
+      "step": 20106
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.699834285714285e-06,
+      "loss": 1.4327,
+      "step": 20107
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.699565714285715e-06,
+      "loss": 1.3643,
+      "step": 20108
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.699297142857143e-06,
+      "loss": 1.5043,
+      "step": 20109
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.699028571428571e-06,
+      "loss": 1.2738,
+      "step": 20110
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.698760000000001e-06,
+      "loss": 1.2995,
+      "step": 20111
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.69849142857143e-06,
+      "loss": 1.2551,
+      "step": 20112
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.698222857142857e-06,
+      "loss": 1.555,
+      "step": 20113
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.697954285714286e-06,
+      "loss": 0.8293,
+      "step": 20114
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.697685714285715e-06,
+      "loss": 0.8484,
+      "step": 20115
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.697417142857142e-06,
+      "loss": 1.3969,
+      "step": 20116
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.697148571428572e-06,
+      "loss": 0.93,
+      "step": 20117
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.69688e-06,
+      "loss": 0.641,
+      "step": 20118
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.69661142857143e-06,
+      "loss": 1.1902,
+      "step": 20119
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.696342857142856e-06,
+      "loss": 0.8769,
+      "step": 20120
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.696074285714286e-06,
+      "loss": 1.1134,
+      "step": 20121
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.695805714285714e-06,
+      "loss": 1.196,
+      "step": 20122
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.695537142857143e-06,
+      "loss": 1.1568,
+      "step": 20123
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.695268571428572e-06,
+      "loss": 1.4578,
+      "step": 20124
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.695e-06,
+      "loss": 1.281,
+      "step": 20125
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.694731428571429e-06,
+      "loss": 1.1581,
+      "step": 20126
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.694462857142857e-06,
+      "loss": 1.7824,
+      "step": 20127
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.694194285714287e-06,
+      "loss": 1.1618,
+      "step": 20128
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.693925714285715e-06,
+      "loss": 1.0648,
+      "step": 20129
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.693657142857143e-06,
+      "loss": 1.2015,
+      "step": 20130
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.693388571428571e-06,
+      "loss": 1.1429,
+      "step": 20131
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.693120000000001e-06,
+      "loss": 1.3105,
+      "step": 20132
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.692851428571428e-06,
+      "loss": 1.5483,
+      "step": 20133
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.692582857142858e-06,
+      "loss": 1.1282,
+      "step": 20134
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.692314285714286e-06,
+      "loss": 1.1145,
+      "step": 20135
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.692045714285714e-06,
+      "loss": 1.2532,
+      "step": 20136
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.691777142857144e-06,
+      "loss": 1.5687,
+      "step": 20137
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.691508571428572e-06,
+      "loss": 0.9595,
+      "step": 20138
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.69124e-06,
+      "loss": 1.4988,
+      "step": 20139
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.690971428571428e-06,
+      "loss": 0.9745,
+      "step": 20140
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.690702857142858e-06,
+      "loss": 1.3045,
+      "step": 20141
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.690434285714286e-06,
+      "loss": 1.3963,
+      "step": 20142
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.690165714285715e-06,
+      "loss": 0.9171,
+      "step": 20143
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.689897142857143e-06,
+      "loss": 1.0734,
+      "step": 20144
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.689628571428573e-06,
+      "loss": 1.2013,
+      "step": 20145
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.689359999999999e-06,
+      "loss": 1.1143,
+      "step": 20146
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.689091428571429e-06,
+      "loss": 1.1265,
+      "step": 20147
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.688822857142857e-06,
+      "loss": 1.4739,
+      "step": 20148
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.688554285714285e-06,
+      "loss": 1.0318,
+      "step": 20149
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.688285714285715e-06,
+      "loss": 1.4778,
+      "step": 20150
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.688017142857143e-06,
+      "loss": 1.3277,
+      "step": 20151
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.687748571428573e-06,
+      "loss": 1.1362,
+      "step": 20152
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.68748e-06,
+      "loss": 1.1335,
+      "step": 20153
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.68721142857143e-06,
+      "loss": 1.4346,
+      "step": 20154
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.686942857142858e-06,
+      "loss": 1.1052,
+      "step": 20155
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.686674285714286e-06,
+      "loss": 1.3215,
+      "step": 20156
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.686405714285714e-06,
+      "loss": 1.3488,
+      "step": 20157
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.686137142857144e-06,
+      "loss": 1.032,
+      "step": 20158
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.68586857142857e-06,
+      "loss": 1.1107,
+      "step": 20159
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.6856e-06,
+      "loss": 1.3021,
+      "step": 20160
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.685331428571429e-06,
+      "loss": 1.1139,
+      "step": 20161
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.685062857142857e-06,
+      "loss": 1.0307,
+      "step": 20162
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.684794285714287e-06,
+      "loss": 1.0192,
+      "step": 20163
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.684525714285715e-06,
+      "loss": 1.0372,
+      "step": 20164
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.684257142857145e-06,
+      "loss": 1.1913,
+      "step": 20165
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.683988571428571e-06,
+      "loss": 0.8871,
+      "step": 20166
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.683720000000001e-06,
+      "loss": 1.0168,
+      "step": 20167
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.683451428571429e-06,
+      "loss": 1.1126,
+      "step": 20168
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.683182857142857e-06,
+      "loss": 1.3809,
+      "step": 20169
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.682914285714285e-06,
+      "loss": 1.4106,
+      "step": 20170
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.682645714285715e-06,
+      "loss": 0.9612,
+      "step": 20171
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.682377142857142e-06,
+      "loss": 1.2974,
+      "step": 20172
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.682108571428572e-06,
+      "loss": 0.8588,
+      "step": 20173
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.68184e-06,
+      "loss": 1.4683,
+      "step": 20174
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.68157142857143e-06,
+      "loss": 1.626,
+      "step": 20175
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.681302857142858e-06,
+      "loss": 1.0916,
+      "step": 20176
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.681034285714286e-06,
+      "loss": 1.0301,
+      "step": 20177
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.680765714285716e-06,
+      "loss": 1.2442,
+      "step": 20178
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.680497142857142e-06,
+      "loss": 1.6218,
+      "step": 20179
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.680228571428572e-06,
+      "loss": 0.7645,
+      "step": 20180
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.67996e-06,
+      "loss": 0.9165,
+      "step": 20181
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.679691428571429e-06,
+      "loss": 0.9573,
+      "step": 20182
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.679422857142857e-06,
+      "loss": 1.4238,
+      "step": 20183
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.679154285714287e-06,
+      "loss": 0.8684,
+      "step": 20184
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.678885714285713e-06,
+      "loss": 1.3138,
+      "step": 20185
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.678617142857143e-06,
+      "loss": 1.1694,
+      "step": 20186
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.678348571428571e-06,
+      "loss": 1.2369,
+      "step": 20187
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.678080000000001e-06,
+      "loss": 1.3169,
+      "step": 20188
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.67781142857143e-06,
+      "loss": 0.7294,
+      "step": 20189
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.677542857142857e-06,
+      "loss": 1.3279,
+      "step": 20190
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.677274285714287e-06,
+      "loss": 1.0812,
+      "step": 20191
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.677005714285714e-06,
+      "loss": 0.821,
+      "step": 20192
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.676737142857144e-06,
+      "loss": 1.0032,
+      "step": 20193
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.676468571428572e-06,
+      "loss": 1.1039,
+      "step": 20194
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.6762e-06,
+      "loss": 0.8359,
+      "step": 20195
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.675931428571428e-06,
+      "loss": 0.9039,
+      "step": 20196
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.675662857142858e-06,
+      "loss": 1.5631,
+      "step": 20197
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.675394285714285e-06,
+      "loss": 1.4084,
+      "step": 20198
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.675125714285714e-06,
+      "loss": 1.2634,
+      "step": 20199
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.674857142857143e-06,
+      "loss": 1.1854,
+      "step": 20200
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.674588571428572e-06,
+      "loss": 1.6385,
+      "step": 20201
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.67432e-06,
+      "loss": 1.3601,
+      "step": 20202
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.674051428571429e-06,
+      "loss": 0.8848,
+      "step": 20203
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.673782857142859e-06,
+      "loss": 0.8896,
+      "step": 20204
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.673514285714285e-06,
+      "loss": 1.1975,
+      "step": 20205
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.673245714285715e-06,
+      "loss": 0.8171,
+      "step": 20206
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.672977142857143e-06,
+      "loss": 1.1581,
+      "step": 20207
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.672708571428571e-06,
+      "loss": 1.351,
+      "step": 20208
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.67244e-06,
+      "loss": 1.2082,
+      "step": 20209
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.67217142857143e-06,
+      "loss": 1.4133,
+      "step": 20210
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.671902857142858e-06,
+      "loss": 1.1577,
+      "step": 20211
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.671634285714286e-06,
+      "loss": 1.1735,
+      "step": 20212
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.671365714285714e-06,
+      "loss": 1.0002,
+      "step": 20213
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.671097142857144e-06,
+      "loss": 0.8509,
+      "step": 20214
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.670828571428572e-06,
+      "loss": 1.1942,
+      "step": 20215
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.67056e-06,
+      "loss": 1.2784,
+      "step": 20216
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.67029142857143e-06,
+      "loss": 1.4403,
+      "step": 20217
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.670022857142857e-06,
+      "loss": 1.391,
+      "step": 20218
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.669754285714286e-06,
+      "loss": 1.2578,
+      "step": 20219
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.669485714285715e-06,
+      "loss": 1.3539,
+      "step": 20220
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.669217142857143e-06,
+      "loss": 1.0033,
+      "step": 20221
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.668948571428571e-06,
+      "loss": 0.9882,
+      "step": 20222
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66868e-06,
+      "loss": 0.6968,
+      "step": 20223
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.668411428571429e-06,
+      "loss": 1.2294,
+      "step": 20224
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.668142857142857e-06,
+      "loss": 1.0983,
+      "step": 20225
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.667874285714285e-06,
+      "loss": 1.3809,
+      "step": 20226
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.667605714285715e-06,
+      "loss": 1.1058,
+      "step": 20227
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.667337142857143e-06,
+      "loss": 1.0561,
+      "step": 20228
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.667068571428572e-06,
+      "loss": 1.2946,
+      "step": 20229
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.666800000000001e-06,
+      "loss": 1.3019,
+      "step": 20230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.666531428571428e-06,
+      "loss": 1.1643,
+      "step": 20231
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.666262857142858e-06,
+      "loss": 1.0183,
+      "step": 20232
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.665994285714286e-06,
+      "loss": 1.0217,
+      "step": 20233
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.665725714285716e-06,
+      "loss": 1.2904,
+      "step": 20234
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.665457142857142e-06,
+      "loss": 0.6979,
+      "step": 20235
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.665188571428572e-06,
+      "loss": 0.8391,
+      "step": 20236
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66492e-06,
+      "loss": 1.3016,
+      "step": 20237
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.664651428571429e-06,
+      "loss": 0.845,
+      "step": 20238
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.664382857142857e-06,
+      "loss": 1.5075,
+      "step": 20239
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.664114285714287e-06,
+      "loss": 1.2138,
+      "step": 20240
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.663845714285715e-06,
+      "loss": 1.502,
+      "step": 20241
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.663577142857143e-06,
+      "loss": 1.2233,
+      "step": 20242
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.663308571428573e-06,
+      "loss": 1.554,
+      "step": 20243
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66304e-06,
+      "loss": 0.8594,
+      "step": 20244
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66277142857143e-06,
+      "loss": 0.8659,
+      "step": 20245
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662502857142857e-06,
+      "loss": 0.9975,
+      "step": 20246
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662234285714287e-06,
+      "loss": 1.1552,
+      "step": 20247
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.661965714285714e-06,
+      "loss": 1.2556,
+      "step": 20248
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.661697142857144e-06,
+      "loss": 1.3966,
+      "step": 20249
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.661428571428572e-06,
+      "loss": 1.1092,
+      "step": 20250
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66116e-06,
+      "loss": 1.2491,
+      "step": 20251
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.660891428571428e-06,
+      "loss": 1.1255,
+      "step": 20252
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.660622857142858e-06,
+      "loss": 1.1089,
+      "step": 20253
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.660354285714286e-06,
+      "loss": 1.0577,
+      "step": 20254
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.660085714285714e-06,
+      "loss": 0.8839,
+      "step": 20255
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.659817142857144e-06,
+      "loss": 1.3973,
+      "step": 20256
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.659548571428572e-06,
+      "loss": 1.4258,
+      "step": 20257
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.65928e-06,
+      "loss": 1.3698,
+      "step": 20258
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.659011428571429e-06,
+      "loss": 1.4131,
+      "step": 20259
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.658742857142859e-06,
+      "loss": 0.9941,
+      "step": 20260
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.658474285714285e-06,
+      "loss": 1.6742,
+      "step": 20261
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.658205714285715e-06,
+      "loss": 1.5036,
+      "step": 20262
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.657937142857143e-06,
+      "loss": 1.2668,
+      "step": 20263
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.657668571428571e-06,
+      "loss": 1.3498,
+      "step": 20264
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.6574e-06,
+      "loss": 1.672,
+      "step": 20265
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.65713142857143e-06,
+      "loss": 1.1829,
+      "step": 20266
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.656862857142858e-06,
+      "loss": 0.9081,
+      "step": 20267
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.656594285714286e-06,
+      "loss": 1.3255,
+      "step": 20268
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.656325714285716e-06,
+      "loss": 1.5728,
+      "step": 20269
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.656057142857144e-06,
+      "loss": 1.3072,
+      "step": 20270
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.655788571428572e-06,
+      "loss": 0.943,
+      "step": 20271
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.65552e-06,
+      "loss": 1.139,
+      "step": 20272
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.65525142857143e-06,
+      "loss": 1.094,
+      "step": 20273
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.654982857142856e-06,
+      "loss": 1.4975,
+      "step": 20274
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.654714285714286e-06,
+      "loss": 0.8417,
+      "step": 20275
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.654445714285714e-06,
+      "loss": 1.534,
+      "step": 20276
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.654177142857143e-06,
+      "loss": 1.3288,
+      "step": 20277
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.653908571428571e-06,
+      "loss": 1.1117,
+      "step": 20278
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.65364e-06,
+      "loss": 1.4866,
+      "step": 20279
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.653371428571429e-06,
+      "loss": 1.4422,
+      "step": 20280
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.653102857142857e-06,
+      "loss": 1.2108,
+      "step": 20281
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.652834285714287e-06,
+      "loss": 1.0125,
+      "step": 20282
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.652565714285715e-06,
+      "loss": 1.1888,
+      "step": 20283
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.652297142857143e-06,
+      "loss": 1.4401,
+      "step": 20284
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.652028571428571e-06,
+      "loss": 0.8186,
+      "step": 20285
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.651760000000001e-06,
+      "loss": 1.1702,
+      "step": 20286
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.651491428571428e-06,
+      "loss": 1.0176,
+      "step": 20287
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.651222857142858e-06,
+      "loss": 1.2334,
+      "step": 20288
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.650954285714286e-06,
+      "loss": 0.9888,
+      "step": 20289
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.650685714285714e-06,
+      "loss": 1.1836,
+      "step": 20290
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.650417142857142e-06,
+      "loss": 1.317,
+      "step": 20291
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.650148571428572e-06,
+      "loss": 1.0713,
+      "step": 20292
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64988e-06,
+      "loss": 1.0147,
+      "step": 20293
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.649611428571428e-06,
+      "loss": 1.1225,
+      "step": 20294
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.649342857142858e-06,
+      "loss": 1.0428,
+      "step": 20295
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.649074285714286e-06,
+      "loss": 1.5271,
+      "step": 20296
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.648805714285715e-06,
+      "loss": 0.9857,
+      "step": 20297
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.648537142857143e-06,
+      "loss": 1.0603,
+      "step": 20298
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.648268571428573e-06,
+      "loss": 0.9465,
+      "step": 20299
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.648e-06,
+      "loss": 1.4637,
+      "step": 20300
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.647731428571429e-06,
+      "loss": 0.865,
+      "step": 20301
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.647462857142857e-06,
+      "loss": 0.9207,
+      "step": 20302
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.647194285714287e-06,
+      "loss": 1.1666,
+      "step": 20303
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.646925714285714e-06,
+      "loss": 1.5856,
+      "step": 20304
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.646657142857143e-06,
+      "loss": 1.3064,
+      "step": 20305
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.646388571428572e-06,
+      "loss": 0.6473,
+      "step": 20306
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64612e-06,
+      "loss": 1.2245,
+      "step": 20307
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64585142857143e-06,
+      "loss": 0.8852,
+      "step": 20308
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.645582857142858e-06,
+      "loss": 1.0999,
+      "step": 20309
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.645314285714286e-06,
+      "loss": 1.354,
+      "step": 20310
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.645045714285714e-06,
+      "loss": 1.2479,
+      "step": 20311
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.644777142857144e-06,
+      "loss": 0.8794,
+      "step": 20312
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64450857142857e-06,
+      "loss": 1.1568,
+      "step": 20313
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64424e-06,
+      "loss": 1.3518,
+      "step": 20314
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.643971428571429e-06,
+      "loss": 1.5314,
+      "step": 20315
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.643702857142858e-06,
+      "loss": 1.1897,
+      "step": 20316
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.643434285714285e-06,
+      "loss": 1.3281,
+      "step": 20317
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.643165714285715e-06,
+      "loss": 1.5375,
+      "step": 20318
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.642897142857143e-06,
+      "loss": 1.2629,
+      "step": 20319
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.642628571428571e-06,
+      "loss": 1.2064,
+      "step": 20320
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.642360000000001e-06,
+      "loss": 1.2106,
+      "step": 20321
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64209142857143e-06,
+      "loss": 1.3511,
+      "step": 20322
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641822857142857e-06,
+      "loss": 0.9843,
+      "step": 20323
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641554285714286e-06,
+      "loss": 1.5327,
+      "step": 20324
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641285714285715e-06,
+      "loss": 0.7397,
+      "step": 20325
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641017142857144e-06,
+      "loss": 1.2552,
+      "step": 20326
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.640748571428572e-06,
+      "loss": 1.3655,
+      "step": 20327
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64048e-06,
+      "loss": 1.7422,
+      "step": 20328
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64021142857143e-06,
+      "loss": 1.4842,
+      "step": 20329
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.639942857142856e-06,
+      "loss": 1.1978,
+      "step": 20330
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.639674285714286e-06,
+      "loss": 1.2248,
+      "step": 20331
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.639405714285714e-06,
+      "loss": 1.0242,
+      "step": 20332
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.639137142857143e-06,
+      "loss": 1.1597,
+      "step": 20333
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.638868571428572e-06,
+      "loss": 1.3074,
+      "step": 20334
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.6386e-06,
+      "loss": 1.3826,
+      "step": 20335
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.638331428571429e-06,
+      "loss": 1.4461,
+      "step": 20336
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.638062857142857e-06,
+      "loss": 1.4116,
+      "step": 20337
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.637794285714287e-06,
+      "loss": 1.5081,
+      "step": 20338
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.637525714285715e-06,
+      "loss": 1.2353,
+      "step": 20339
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.637257142857143e-06,
+      "loss": 0.9822,
+      "step": 20340
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.636988571428571e-06,
+      "loss": 1.221,
+      "step": 20341
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.636720000000001e-06,
+      "loss": 1.0937,
+      "step": 20342
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.636451428571428e-06,
+      "loss": 1.5429,
+      "step": 20343
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.636182857142858e-06,
+      "loss": 0.8671,
+      "step": 20344
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.635914285714286e-06,
+      "loss": 1.4729,
+      "step": 20345
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.635645714285714e-06,
+      "loss": 1.0293,
+      "step": 20346
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.635377142857144e-06,
+      "loss": 0.7852,
+      "step": 20347
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.635108571428572e-06,
+      "loss": 1.3606,
+      "step": 20348
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.634840000000002e-06,
+      "loss": 1.0234,
+      "step": 20349
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.634571428571428e-06,
+      "loss": 0.9768,
+      "step": 20350
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.634302857142858e-06,
+      "loss": 1.0254,
+      "step": 20351
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.634034285714286e-06,
+      "loss": 0.8344,
+      "step": 20352
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.633765714285715e-06,
+      "loss": 1.2825,
+      "step": 20353
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.633497142857143e-06,
+      "loss": 1.2539,
+      "step": 20354
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.633228571428573e-06,
+      "loss": 1.0252,
+      "step": 20355
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.632959999999999e-06,
+      "loss": 1.5535,
+      "step": 20356
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.632691428571429e-06,
+      "loss": 1.3109,
+      "step": 20357
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.632422857142857e-06,
+      "loss": 1.2326,
+      "step": 20358
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.632154285714285e-06,
+      "loss": 0.7771,
+      "step": 20359
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.631885714285715e-06,
+      "loss": 1.2191,
+      "step": 20360
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.631617142857143e-06,
+      "loss": 1.1444,
+      "step": 20361
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.631348571428573e-06,
+      "loss": 1.3542,
+      "step": 20362
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.63108e-06,
+      "loss": 0.8937,
+      "step": 20363
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.63081142857143e-06,
+      "loss": 1.4875,
+      "step": 20364
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.630542857142858e-06,
+      "loss": 0.918,
+      "step": 20365
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.630274285714286e-06,
+      "loss": 0.7591,
+      "step": 20366
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.630005714285714e-06,
+      "loss": 1.0512,
+      "step": 20367
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.629737142857144e-06,
+      "loss": 1.2165,
+      "step": 20368
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.62946857142857e-06,
+      "loss": 1.1027,
+      "step": 20369
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.6292e-06,
+      "loss": 1.1802,
+      "step": 20370
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.628931428571428e-06,
+      "loss": 1.5515,
+      "step": 20371
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.628662857142857e-06,
+      "loss": 1.2209,
+      "step": 20372
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.628394285714287e-06,
+      "loss": 1.4465,
+      "step": 20373
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.628125714285715e-06,
+      "loss": 1.2034,
+      "step": 20374
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.627857142857145e-06,
+      "loss": 1.0636,
+      "step": 20375
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.627588571428571e-06,
+      "loss": 0.6776,
+      "step": 20376
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.627320000000001e-06,
+      "loss": 1.0729,
+      "step": 20377
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.627051428571429e-06,
+      "loss": 1.0417,
+      "step": 20378
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.626782857142857e-06,
+      "loss": 1.2561,
+      "step": 20379
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.626514285714285e-06,
+      "loss": 1.0876,
+      "step": 20380
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.626245714285715e-06,
+      "loss": 0.8733,
+      "step": 20381
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.625977142857142e-06,
+      "loss": 1.4789,
+      "step": 20382
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.625708571428572e-06,
+      "loss": 1.0815,
+      "step": 20383
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.62544e-06,
+      "loss": 1.1489,
+      "step": 20384
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.62517142857143e-06,
+      "loss": 0.9921,
+      "step": 20385
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.624902857142858e-06,
+      "loss": 1.6598,
+      "step": 20386
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.624634285714286e-06,
+      "loss": 1.5488,
+      "step": 20387
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.624365714285716e-06,
+      "loss": 0.9001,
+      "step": 20388
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.624097142857142e-06,
+      "loss": 1.3727,
+      "step": 20389
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.623828571428572e-06,
+      "loss": 1.3347,
+      "step": 20390
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.62356e-06,
+      "loss": 1.2713,
+      "step": 20391
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.623291428571429e-06,
+      "loss": 1.3162,
+      "step": 20392
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.623022857142857e-06,
+      "loss": 1.0618,
+      "step": 20393
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.622754285714287e-06,
+      "loss": 1.1253,
+      "step": 20394
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.622485714285713e-06,
+      "loss": 1.4885,
+      "step": 20395
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.622217142857143e-06,
+      "loss": 1.3199,
+      "step": 20396
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.621948571428571e-06,
+      "loss": 0.7637,
+      "step": 20397
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.621680000000001e-06,
+      "loss": 1.4338,
+      "step": 20398
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.62141142857143e-06,
+      "loss": 1.1266,
+      "step": 20399
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.621142857142857e-06,
+      "loss": 1.4371,
+      "step": 20400
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.620874285714287e-06,
+      "loss": 1.0602,
+      "step": 20401
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.620605714285714e-06,
+      "loss": 1.3745,
+      "step": 20402
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.620337142857144e-06,
+      "loss": 1.4416,
+      "step": 20403
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.620068571428572e-06,
+      "loss": 0.8787,
+      "step": 20404
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.6198e-06,
+      "loss": 1.0028,
+      "step": 20405
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.619531428571428e-06,
+      "loss": 1.1742,
+      "step": 20406
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.619262857142858e-06,
+      "loss": 0.8271,
+      "step": 20407
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.618994285714286e-06,
+      "loss": 1.3797,
+      "step": 20408
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.618725714285714e-06,
+      "loss": 1.4822,
+      "step": 20409
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.618457142857143e-06,
+      "loss": 1.5084,
+      "step": 20410
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.618188571428572e-06,
+      "loss": 1.181,
+      "step": 20411
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.61792e-06,
+      "loss": 0.7763,
+      "step": 20412
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.617651428571429e-06,
+      "loss": 1.4381,
+      "step": 20413
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.617382857142859e-06,
+      "loss": 1.1109,
+      "step": 20414
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.617114285714285e-06,
+      "loss": 1.6596,
+      "step": 20415
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.616845714285715e-06,
+      "loss": 1.145,
+      "step": 20416
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.616577142857143e-06,
+      "loss": 1.2949,
+      "step": 20417
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.616308571428571e-06,
+      "loss": 1.1329,
+      "step": 20418
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.61604e-06,
+      "loss": 1.3189,
+      "step": 20419
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.61577142857143e-06,
+      "loss": 1.2763,
+      "step": 20420
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.615502857142858e-06,
+      "loss": 1.4584,
+      "step": 20421
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.615234285714286e-06,
+      "loss": 1.179,
+      "step": 20422
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.614965714285714e-06,
+      "loss": 1.0046,
+      "step": 20423
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.614697142857144e-06,
+      "loss": 1.0596,
+      "step": 20424
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.614428571428572e-06,
+      "loss": 0.8735,
+      "step": 20425
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.61416e-06,
+      "loss": 1.6549,
+      "step": 20426
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.61389142857143e-06,
+      "loss": 1.353,
+      "step": 20427
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.613622857142857e-06,
+      "loss": 1.3019,
+      "step": 20428
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.613354285714286e-06,
+      "loss": 0.9909,
+      "step": 20429
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.613085714285715e-06,
+      "loss": 1.1771,
+      "step": 20430
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.612817142857144e-06,
+      "loss": 1.1083,
+      "step": 20431
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.612548571428571e-06,
+      "loss": 0.9402,
+      "step": 20432
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.61228e-06,
+      "loss": 1.046,
+      "step": 20433
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.612011428571429e-06,
+      "loss": 0.7897,
+      "step": 20434
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.611742857142857e-06,
+      "loss": 0.8011,
+      "step": 20435
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.611474285714285e-06,
+      "loss": 1.0475,
+      "step": 20436
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.611205714285715e-06,
+      "loss": 1.2049,
+      "step": 20437
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.610937142857143e-06,
+      "loss": 0.8472,
+      "step": 20438
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.610668571428572e-06,
+      "loss": 1.2015,
+      "step": 20439
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.610400000000001e-06,
+      "loss": 1.4335,
+      "step": 20440
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.610131428571428e-06,
+      "loss": 0.855,
+      "step": 20441
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.609862857142858e-06,
+      "loss": 0.93,
+      "step": 20442
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.609594285714286e-06,
+      "loss": 1.2601,
+      "step": 20443
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.609325714285716e-06,
+      "loss": 0.8375,
+      "step": 20444
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.609057142857142e-06,
+      "loss": 1.2105,
+      "step": 20445
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.608788571428572e-06,
+      "loss": 1.2165,
+      "step": 20446
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60852e-06,
+      "loss": 1.0445,
+      "step": 20447
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.608251428571429e-06,
+      "loss": 0.8971,
+      "step": 20448
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.607982857142857e-06,
+      "loss": 1.3291,
+      "step": 20449
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.607714285714287e-06,
+      "loss": 1.2964,
+      "step": 20450
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.607445714285715e-06,
+      "loss": 1.3147,
+      "step": 20451
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.607177142857143e-06,
+      "loss": 1.0833,
+      "step": 20452
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.606908571428573e-06,
+      "loss": 1.3259,
+      "step": 20453
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.606640000000001e-06,
+      "loss": 1.6666,
+      "step": 20454
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60637142857143e-06,
+      "loss": 0.8641,
+      "step": 20455
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.606102857142857e-06,
+      "loss": 1.0232,
+      "step": 20456
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.605834285714287e-06,
+      "loss": 0.7753,
+      "step": 20457
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.605565714285714e-06,
+      "loss": 0.9972,
+      "step": 20458
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.605297142857144e-06,
+      "loss": 1.2547,
+      "step": 20459
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.605028571428572e-06,
+      "loss": 1.3284,
+      "step": 20460
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60476e-06,
+      "loss": 0.8825,
+      "step": 20461
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.604491428571428e-06,
+      "loss": 0.9593,
+      "step": 20462
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.604222857142858e-06,
+      "loss": 1.1079,
+      "step": 20463
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.603954285714284e-06,
+      "loss": 1.4483,
+      "step": 20464
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.603685714285714e-06,
+      "loss": 0.9538,
+      "step": 20465
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.603417142857144e-06,
+      "loss": 1.2126,
+      "step": 20466
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.603148571428572e-06,
+      "loss": 0.9187,
+      "step": 20467
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60288e-06,
+      "loss": 1.139,
+      "step": 20468
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.602611428571429e-06,
+      "loss": 0.8721,
+      "step": 20469
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.602342857142859e-06,
+      "loss": 1.2037,
+      "step": 20470
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.602074285714285e-06,
+      "loss": 1.1861,
+      "step": 20471
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.601805714285715e-06,
+      "loss": 1.6142,
+      "step": 20472
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.601537142857143e-06,
+      "loss": 1.3568,
+      "step": 20473
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.601268571428571e-06,
+      "loss": 1.2104,
+      "step": 20474
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.601e-06,
+      "loss": 1.2658,
+      "step": 20475
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60073142857143e-06,
+      "loss": 1.2144,
+      "step": 20476
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.600462857142858e-06,
+      "loss": 1.102,
+      "step": 20477
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.600194285714286e-06,
+      "loss": 1.1763,
+      "step": 20478
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.599925714285716e-06,
+      "loss": 1.4076,
+      "step": 20479
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.599657142857144e-06,
+      "loss": 0.9143,
+      "step": 20480
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.599388571428572e-06,
+      "loss": 1.3752,
+      "step": 20481
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59912e-06,
+      "loss": 1.0715,
+      "step": 20482
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59885142857143e-06,
+      "loss": 0.7617,
+      "step": 20483
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.598582857142856e-06,
+      "loss": 1.0525,
+      "step": 20484
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.598314285714286e-06,
+      "loss": 0.9373,
+      "step": 20485
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.598045714285714e-06,
+      "loss": 1.2333,
+      "step": 20486
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.597777142857143e-06,
+      "loss": 1.0718,
+      "step": 20487
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59750857142857e-06,
+      "loss": 1.2777,
+      "step": 20488
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59724e-06,
+      "loss": 1.3664,
+      "step": 20489
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.596971428571429e-06,
+      "loss": 1.3445,
+      "step": 20490
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.596702857142857e-06,
+      "loss": 1.0247,
+      "step": 20491
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.596434285714287e-06,
+      "loss": 1.4671,
+      "step": 20492
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.596165714285715e-06,
+      "loss": 1.0343,
+      "step": 20493
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.595897142857143e-06,
+      "loss": 1.0702,
+      "step": 20494
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.595628571428571e-06,
+      "loss": 1.3604,
+      "step": 20495
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.595360000000001e-06,
+      "loss": 1.344,
+      "step": 20496
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.595091428571428e-06,
+      "loss": 0.9509,
+      "step": 20497
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.594822857142858e-06,
+      "loss": 1.1544,
+      "step": 20498
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.594554285714286e-06,
+      "loss": 1.0855,
+      "step": 20499
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.594285714285716e-06,
+      "loss": 0.7624,
+      "step": 20500
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.594017142857142e-06,
+      "loss": 1.2973,
+      "step": 20501
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.593748571428572e-06,
+      "loss": 1.3975,
+      "step": 20502
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59348e-06,
+      "loss": 0.9459,
+      "step": 20503
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.593211428571428e-06,
+      "loss": 1.126,
+      "step": 20504
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.592942857142858e-06,
+      "loss": 0.9192,
+      "step": 20505
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.592674285714286e-06,
+      "loss": 1.1807,
+      "step": 20506
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.592405714285715e-06,
+      "loss": 0.7764,
+      "step": 20507
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.592137142857143e-06,
+      "loss": 1.2102,
+      "step": 20508
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.591868571428573e-06,
+      "loss": 0.9602,
+      "step": 20509
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.5916e-06,
+      "loss": 0.7967,
+      "step": 20510
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.591331428571429e-06,
+      "loss": 1.47,
+      "step": 20511
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.591062857142857e-06,
+      "loss": 0.8993,
+      "step": 20512
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.590794285714287e-06,
+      "loss": 1.2506,
+      "step": 20513
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.590525714285714e-06,
+      "loss": 1.4271,
+      "step": 20514
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.590257142857143e-06,
+      "loss": 1.0781,
+      "step": 20515
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.589988571428572e-06,
+      "loss": 1.4617,
+      "step": 20516
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.58972e-06,
+      "loss": 1.1094,
+      "step": 20517
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.58945142857143e-06,
+      "loss": 1.5589,
+      "step": 20518
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.589182857142858e-06,
+      "loss": 1.1167,
+      "step": 20519
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.588914285714286e-06,
+      "loss": 0.7913,
+      "step": 20520
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.588645714285714e-06,
+      "loss": 1.2686,
+      "step": 20521
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.588377142857144e-06,
+      "loss": 1.3107,
+      "step": 20522
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.58810857142857e-06,
+      "loss": 0.9212,
+      "step": 20523
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.58784e-06,
+      "loss": 1.4118,
+      "step": 20524
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.587571428571429e-06,
+      "loss": 1.2572,
+      "step": 20525
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.587302857142858e-06,
+      "loss": 1.0651,
+      "step": 20526
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.587034285714285e-06,
+      "loss": 1.0265,
+      "step": 20527
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.586765714285715e-06,
+      "loss": 1.4319,
+      "step": 20528
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.586497142857143e-06,
+      "loss": 1.0524,
+      "step": 20529
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.586228571428571e-06,
+      "loss": 0.6801,
+      "step": 20530
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.585960000000001e-06,
+      "loss": 0.8786,
+      "step": 20531
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.58569142857143e-06,
+      "loss": 0.933,
+      "step": 20532
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.585422857142857e-06,
+      "loss": 1.0552,
+      "step": 20533
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.585154285714286e-06,
+      "loss": 1.0684,
+      "step": 20534
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.584885714285715e-06,
+      "loss": 1.084,
+      "step": 20535
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.584617142857144e-06,
+      "loss": 1.236,
+      "step": 20536
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.584348571428572e-06,
+      "loss": 1.3097,
+      "step": 20537
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.58408e-06,
+      "loss": 1.1817,
+      "step": 20538
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.58381142857143e-06,
+      "loss": 1.1402,
+      "step": 20539
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.583542857142856e-06,
+      "loss": 1.2801,
+      "step": 20540
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.583274285714286e-06,
+      "loss": 1.2165,
+      "step": 20541
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.583005714285714e-06,
+      "loss": 1.4209,
+      "step": 20542
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.582737142857143e-06,
+      "loss": 1.4422,
+      "step": 20543
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.582468571428572e-06,
+      "loss": 1.5832,
+      "step": 20544
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.5822e-06,
+      "loss": 1.3182,
+      "step": 20545
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.581931428571429e-06,
+      "loss": 1.3951,
+      "step": 20546
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.581662857142857e-06,
+      "loss": 1.569,
+      "step": 20547
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.581394285714287e-06,
+      "loss": 1.0886,
+      "step": 20548
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.581125714285715e-06,
+      "loss": 1.2248,
+      "step": 20549
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580857142857143e-06,
+      "loss": 1.1458,
+      "step": 20550
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580588571428571e-06,
+      "loss": 1.0445,
+      "step": 20551
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580320000000001e-06,
+      "loss": 0.6905,
+      "step": 20552
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580051428571428e-06,
+      "loss": 1.228,
+      "step": 20553
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.579782857142858e-06,
+      "loss": 1.2925,
+      "step": 20554
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.579514285714286e-06,
+      "loss": 1.3838,
+      "step": 20555
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.579245714285714e-06,
+      "loss": 1.2749,
+      "step": 20556
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.578977142857144e-06,
+      "loss": 1.1283,
+      "step": 20557
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.578708571428572e-06,
+      "loss": 1.5247,
+      "step": 20558
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.578440000000002e-06,
+      "loss": 1.24,
+      "step": 20559
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.578171428571428e-06,
+      "loss": 1.0711,
+      "step": 20560
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.577902857142858e-06,
+      "loss": 0.9276,
+      "step": 20561
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.577634285714286e-06,
+      "loss": 1.0476,
+      "step": 20562
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.577365714285715e-06,
+      "loss": 0.8633,
+      "step": 20563
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.577097142857143e-06,
+      "loss": 1.4338,
+      "step": 20564
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.576828571428573e-06,
+      "loss": 1.0225,
+      "step": 20565
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.576559999999999e-06,
+      "loss": 1.0001,
+      "step": 20566
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.576291428571429e-06,
+      "loss": 1.4104,
+      "step": 20567
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.576022857142857e-06,
+      "loss": 1.1687,
+      "step": 20568
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.575754285714285e-06,
+      "loss": 0.8124,
+      "step": 20569
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.575485714285713e-06,
+      "loss": 1.049,
+      "step": 20570
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.575217142857143e-06,
+      "loss": 1.4387,
+      "step": 20571
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.574948571428573e-06,
+      "loss": 1.2181,
+      "step": 20572
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.57468e-06,
+      "loss": 1.0404,
+      "step": 20573
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.57441142857143e-06,
+      "loss": 1.4304,
+      "step": 20574
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.574142857142858e-06,
+      "loss": 1.406,
+      "step": 20575
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.573874285714286e-06,
+      "loss": 1.3598,
+      "step": 20576
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.573605714285714e-06,
+      "loss": 0.7967,
+      "step": 20577
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.573337142857144e-06,
+      "loss": 1.0634,
+      "step": 20578
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.57306857142857e-06,
+      "loss": 1.4081,
+      "step": 20579
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.5728e-06,
+      "loss": 1.4195,
+      "step": 20580
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.572531428571428e-06,
+      "loss": 1.5508,
+      "step": 20581
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.572262857142858e-06,
+      "loss": 1.1231,
+      "step": 20582
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.571994285714285e-06,
+      "loss": 1.2344,
+      "step": 20583
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.571725714285715e-06,
+      "loss": 1.4043,
+      "step": 20584
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.571457142857145e-06,
+      "loss": 1.0631,
+      "step": 20585
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.571188571428571e-06,
+      "loss": 0.9865,
+      "step": 20586
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.570920000000001e-06,
+      "loss": 1.3207,
+      "step": 20587
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.570651428571429e-06,
+      "loss": 1.6091,
+      "step": 20588
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.570382857142857e-06,
+      "loss": 1.1652,
+      "step": 20589
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.570114285714285e-06,
+      "loss": 1.2032,
+      "step": 20590
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.569845714285715e-06,
+      "loss": 1.2244,
+      "step": 20591
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.569577142857142e-06,
+      "loss": 1.3381,
+      "step": 20592
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.569308571428572e-06,
+      "loss": 1.0046,
+      "step": 20593
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.56904e-06,
+      "loss": 1.2189,
+      "step": 20594
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.56877142857143e-06,
+      "loss": 1.3281,
+      "step": 20595
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.568502857142856e-06,
+      "loss": 1.1679,
+      "step": 20596
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.568234285714286e-06,
+      "loss": 1.2767,
+      "step": 20597
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.567965714285716e-06,
+      "loss": 1.1819,
+      "step": 20598
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.567697142857142e-06,
+      "loss": 1.4808,
+      "step": 20599
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.567428571428572e-06,
+      "loss": 1.2957,
+      "step": 20600
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.56716e-06,
+      "loss": 1.1023,
+      "step": 20601
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.566891428571429e-06,
+      "loss": 1.0043,
+      "step": 20602
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.566622857142857e-06,
+      "loss": 0.8848,
+      "step": 20603
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.566354285714287e-06,
+      "loss": 0.919,
+      "step": 20604
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.566085714285715e-06,
+      "loss": 1.3174,
+      "step": 20605
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.565817142857143e-06,
+      "loss": 0.9433,
+      "step": 20606
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.565548571428571e-06,
+      "loss": 1.1466,
+      "step": 20607
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.565280000000001e-06,
+      "loss": 1.9316,
+      "step": 20608
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.565011428571428e-06,
+      "loss": 0.8697,
+      "step": 20609
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.564742857142857e-06,
+      "loss": 0.9655,
+      "step": 20610
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.564474285714287e-06,
+      "loss": 1.088,
+      "step": 20611
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.564205714285714e-06,
+      "loss": 0.928,
+      "step": 20612
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.563937142857144e-06,
+      "loss": 1.1623,
+      "step": 20613
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.563668571428572e-06,
+      "loss": 0.9211,
+      "step": 20614
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.5634e-06,
+      "loss": 1.1973,
+      "step": 20615
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.563131428571428e-06,
+      "loss": 1.1864,
+      "step": 20616
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.562862857142858e-06,
+      "loss": 1.187,
+      "step": 20617
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.562594285714286e-06,
+      "loss": 1.0737,
+      "step": 20618
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.562325714285714e-06,
+      "loss": 1.2981,
+      "step": 20619
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.562057142857143e-06,
+      "loss": 1.4155,
+      "step": 20620
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.561788571428572e-06,
+      "loss": 1.2823,
+      "step": 20621
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.561519999999999e-06,
+      "loss": 1.4711,
+      "step": 20622
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.561251428571429e-06,
+      "loss": 1.2101,
+      "step": 20623
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.560982857142859e-06,
+      "loss": 1.2544,
+      "step": 20624
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.560714285714285e-06,
+      "loss": 1.3415,
+      "step": 20625
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.560445714285715e-06,
+      "loss": 0.9261,
+      "step": 20626
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.560177142857143e-06,
+      "loss": 1.551,
+      "step": 20627
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.559908571428573e-06,
+      "loss": 1.3577,
+      "step": 20628
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55964e-06,
+      "loss": 1.1242,
+      "step": 20629
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55937142857143e-06,
+      "loss": 1.0809,
+      "step": 20630
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.559102857142858e-06,
+      "loss": 1.6005,
+      "step": 20631
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.558834285714286e-06,
+      "loss": 0.8991,
+      "step": 20632
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.558565714285714e-06,
+      "loss": 1.6863,
+      "step": 20633
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.558297142857144e-06,
+      "loss": 1.4097,
+      "step": 20634
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55802857142857e-06,
+      "loss": 1.2174,
+      "step": 20635
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55776e-06,
+      "loss": 1.4008,
+      "step": 20636
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55749142857143e-06,
+      "loss": 1.48,
+      "step": 20637
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.557222857142857e-06,
+      "loss": 1.4583,
+      "step": 20638
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.556954285714286e-06,
+      "loss": 1.2535,
+      "step": 20639
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.556685714285715e-06,
+      "loss": 1.4861,
+      "step": 20640
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.556417142857144e-06,
+      "loss": 1.6665,
+      "step": 20641
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.556148571428571e-06,
+      "loss": 1.0583,
+      "step": 20642
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55588e-06,
+      "loss": 0.6631,
+      "step": 20643
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.555611428571429e-06,
+      "loss": 1.0831,
+      "step": 20644
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.555342857142857e-06,
+      "loss": 1.5901,
+      "step": 20645
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.555074285714285e-06,
+      "loss": 0.9345,
+      "step": 20646
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.554805714285715e-06,
+      "loss": 1.2256,
+      "step": 20647
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.554537142857142e-06,
+      "loss": 1.1126,
+      "step": 20648
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.554268571428572e-06,
+      "loss": 1.265,
+      "step": 20649
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.554000000000001e-06,
+      "loss": 0.9258,
+      "step": 20650
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55373142857143e-06,
+      "loss": 1.2544,
+      "step": 20651
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.553462857142858e-06,
+      "loss": 0.8445,
+      "step": 20652
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.553194285714286e-06,
+      "loss": 0.9814,
+      "step": 20653
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.552925714285716e-06,
+      "loss": 1.0617,
+      "step": 20654
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.552657142857142e-06,
+      "loss": 1.1283,
+      "step": 20655
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.552388571428572e-06,
+      "loss": 1.352,
+      "step": 20656
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.55212e-06,
+      "loss": 1.2552,
+      "step": 20657
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.551851428571429e-06,
+      "loss": 0.7284,
+      "step": 20658
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.551582857142857e-06,
+      "loss": 1.0799,
+      "step": 20659
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.551314285714287e-06,
+      "loss": 1.3782,
+      "step": 20660
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.551045714285713e-06,
+      "loss": 0.9273,
+      "step": 20661
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.550777142857143e-06,
+      "loss": 1.0199,
+      "step": 20662
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.550508571428573e-06,
+      "loss": 1.0773,
+      "step": 20663
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.550240000000001e-06,
+      "loss": 1.0547,
+      "step": 20664
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54997142857143e-06,
+      "loss": 1.3408,
+      "step": 20665
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.549702857142857e-06,
+      "loss": 1.3252,
+      "step": 20666
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.549434285714287e-06,
+      "loss": 1.0582,
+      "step": 20667
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.549165714285714e-06,
+      "loss": 1.4924,
+      "step": 20668
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.548897142857144e-06,
+      "loss": 0.9811,
+      "step": 20669
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.548628571428572e-06,
+      "loss": 1.2017,
+      "step": 20670
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54836e-06,
+      "loss": 1.091,
+      "step": 20671
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.548091428571428e-06,
+      "loss": 0.997,
+      "step": 20672
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.547822857142858e-06,
+      "loss": 1.4287,
+      "step": 20673
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.547554285714286e-06,
+      "loss": 1.2178,
+      "step": 20674
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.547285714285714e-06,
+      "loss": 0.9052,
+      "step": 20675
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.547017142857142e-06,
+      "loss": 1.0973,
+      "step": 20676
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.546748571428572e-06,
+      "loss": 1.4692,
+      "step": 20677
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54648e-06,
+      "loss": 1.2202,
+      "step": 20678
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.546211428571429e-06,
+      "loss": 1.2476,
+      "step": 20679
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.545942857142859e-06,
+      "loss": 1.0804,
+      "step": 20680
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.545674285714285e-06,
+      "loss": 1.2428,
+      "step": 20681
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.545405714285715e-06,
+      "loss": 0.9481,
+      "step": 20682
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.545137142857143e-06,
+      "loss": 0.8945,
+      "step": 20683
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.544868571428571e-06,
+      "loss": 1.3108,
+      "step": 20684
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.5446e-06,
+      "loss": 1.0754,
+      "step": 20685
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54433142857143e-06,
+      "loss": 1.3168,
+      "step": 20686
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.544062857142858e-06,
+      "loss": 0.6905,
+      "step": 20687
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.543794285714286e-06,
+      "loss": 1.5402,
+      "step": 20688
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.543525714285714e-06,
+      "loss": 0.9565,
+      "step": 20689
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.543257142857144e-06,
+      "loss": 1.3009,
+      "step": 20690
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.542988571428572e-06,
+      "loss": 1.5758,
+      "step": 20691
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54272e-06,
+      "loss": 0.8596,
+      "step": 20692
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54245142857143e-06,
+      "loss": 0.9629,
+      "step": 20693
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.542182857142856e-06,
+      "loss": 0.8456,
+      "step": 20694
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.541914285714286e-06,
+      "loss": 1.1868,
+      "step": 20695
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.541645714285714e-06,
+      "loss": 0.9531,
+      "step": 20696
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.541377142857143e-06,
+      "loss": 1.2338,
+      "step": 20697
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54110857142857e-06,
+      "loss": 1.3263,
+      "step": 20698
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.54084e-06,
+      "loss": 0.9966,
+      "step": 20699
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.540571428571429e-06,
+      "loss": 1.2512,
+      "step": 20700
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.540302857142857e-06,
+      "loss": 1.2753,
+      "step": 20701
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.540034285714285e-06,
+      "loss": 1.6258,
+      "step": 20702
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.539765714285715e-06,
+      "loss": 1.2672,
+      "step": 20703
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.539497142857143e-06,
+      "loss": 0.8387,
+      "step": 20704
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.539228571428571e-06,
+      "loss": 0.8478,
+      "step": 20705
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.538960000000001e-06,
+      "loss": 1.0001,
+      "step": 20706
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.538691428571428e-06,
+      "loss": 1.1164,
+      "step": 20707
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.538422857142858e-06,
+      "loss": 1.7293,
+      "step": 20708
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.538154285714286e-06,
+      "loss": 0.9231,
+      "step": 20709
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.537885714285716e-06,
+      "loss": 1.0178,
+      "step": 20710
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.537617142857142e-06,
+      "loss": 1.053,
+      "step": 20711
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.537348571428572e-06,
+      "loss": 1.3243,
+      "step": 20712
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.53708e-06,
+      "loss": 1.3,
+      "step": 20713
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.536811428571428e-06,
+      "loss": 1.21,
+      "step": 20714
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.536542857142857e-06,
+      "loss": 1.5139,
+      "step": 20715
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.536274285714286e-06,
+      "loss": 1.1549,
+      "step": 20716
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.536005714285715e-06,
+      "loss": 1.2378,
+      "step": 20717
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.535737142857143e-06,
+      "loss": 1.3099,
+      "step": 20718
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.535468571428573e-06,
+      "loss": 1.414,
+      "step": 20719
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.5352e-06,
+      "loss": 0.991,
+      "step": 20720
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.534931428571429e-06,
+      "loss": 1.6334,
+      "step": 20721
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.534662857142857e-06,
+      "loss": 1.1031,
+      "step": 20722
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.534394285714287e-06,
+      "loss": 1.2419,
+      "step": 20723
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.534125714285714e-06,
+      "loss": 1.1113,
+      "step": 20724
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.533857142857143e-06,
+      "loss": 0.8087,
+      "step": 20725
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.533588571428572e-06,
+      "loss": 1.0157,
+      "step": 20726
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.53332e-06,
+      "loss": 1.3097,
+      "step": 20727
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.533051428571428e-06,
+      "loss": 1.0415,
+      "step": 20728
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.532782857142858e-06,
+      "loss": 1.4359,
+      "step": 20729
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.532514285714286e-06,
+      "loss": 1.4424,
+      "step": 20730
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.532245714285714e-06,
+      "loss": 1.1095,
+      "step": 20731
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.531977142857144e-06,
+      "loss": 1.4287,
+      "step": 20732
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.531708571428572e-06,
+      "loss": 1.2683,
+      "step": 20733
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.53144e-06,
+      "loss": 0.9899,
+      "step": 20734
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.531171428571429e-06,
+      "loss": 1.2005,
+      "step": 20735
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.530902857142858e-06,
+      "loss": 1.2046,
+      "step": 20736
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.530634285714285e-06,
+      "loss": 1.1532,
+      "step": 20737
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.530365714285715e-06,
+      "loss": 1.3093,
+      "step": 20738
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.530097142857143e-06,
+      "loss": 1.4223,
+      "step": 20739
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.529828571428571e-06,
+      "loss": 1.2642,
+      "step": 20740
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.52956e-06,
+      "loss": 0.85,
+      "step": 20741
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.52929142857143e-06,
+      "loss": 0.9973,
+      "step": 20742
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.529022857142857e-06,
+      "loss": 1.1045,
+      "step": 20743
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.528754285714286e-06,
+      "loss": 1.038,
+      "step": 20744
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.528485714285715e-06,
+      "loss": 1.2343,
+      "step": 20745
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.528217142857144e-06,
+      "loss": 1.3295,
+      "step": 20746
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.527948571428572e-06,
+      "loss": 0.8607,
+      "step": 20747
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.52768e-06,
+      "loss": 0.8704,
+      "step": 20748
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.52741142857143e-06,
+      "loss": 1.5273,
+      "step": 20749
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.527142857142856e-06,
+      "loss": 1.3084,
+      "step": 20750
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.526874285714286e-06,
+      "loss": 1.0917,
+      "step": 20751
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.526605714285714e-06,
+      "loss": 1.4889,
+      "step": 20752
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.526337142857143e-06,
+      "loss": 1.1323,
+      "step": 20753
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.52606857142857e-06,
+      "loss": 0.5846,
+      "step": 20754
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.5258e-06,
+      "loss": 1.189,
+      "step": 20755
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.52553142857143e-06,
+      "loss": 1.4898,
+      "step": 20756
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.525262857142857e-06,
+      "loss": 1.2051,
+      "step": 20757
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.524994285714287e-06,
+      "loss": 0.8685,
+      "step": 20758
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.524725714285715e-06,
+      "loss": 1.0271,
+      "step": 20759
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.524457142857143e-06,
+      "loss": 1.1404,
+      "step": 20760
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.524188571428571e-06,
+      "loss": 1.1683,
+      "step": 20761
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.523920000000001e-06,
+      "loss": 1.5026,
+      "step": 20762
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.523651428571428e-06,
+      "loss": 1.1654,
+      "step": 20763
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.523382857142858e-06,
+      "loss": 1.3585,
+      "step": 20764
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.523114285714286e-06,
+      "loss": 1.1589,
+      "step": 20765
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.522845714285714e-06,
+      "loss": 1.0472,
+      "step": 20766
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.522577142857142e-06,
+      "loss": 0.8273,
+      "step": 20767
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.522308571428572e-06,
+      "loss": 1.0424,
+      "step": 20768
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.522040000000002e-06,
+      "loss": 1.3179,
+      "step": 20769
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.521771428571428e-06,
+      "loss": 1.2752,
+      "step": 20770
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.521502857142858e-06,
+      "loss": 0.9159,
+      "step": 20771
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.521234285714286e-06,
+      "loss": 1.0565,
+      "step": 20772
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.520965714285715e-06,
+      "loss": 1.3197,
+      "step": 20773
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.520697142857143e-06,
+      "loss": 1.2322,
+      "step": 20774
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.520428571428573e-06,
+      "loss": 1.3943,
+      "step": 20775
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.520159999999999e-06,
+      "loss": 0.9959,
+      "step": 20776
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.519891428571429e-06,
+      "loss": 1.3296,
+      "step": 20777
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.519622857142857e-06,
+      "loss": 1.4309,
+      "step": 20778
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.519354285714287e-06,
+      "loss": 1.233,
+      "step": 20779
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.519085714285713e-06,
+      "loss": 1.1706,
+      "step": 20780
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.518817142857143e-06,
+      "loss": 1.0571,
+      "step": 20781
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.518548571428572e-06,
+      "loss": 0.9556,
+      "step": 20782
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.51828e-06,
+      "loss": 0.8698,
+      "step": 20783
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.51801142857143e-06,
+      "loss": 1.3116,
+      "step": 20784
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.517742857142858e-06,
+      "loss": 1.0726,
+      "step": 20785
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.517474285714286e-06,
+      "loss": 0.9283,
+      "step": 20786
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.517205714285714e-06,
+      "loss": 0.719,
+      "step": 20787
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.516937142857144e-06,
+      "loss": 1.0823,
+      "step": 20788
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.51666857142857e-06,
+      "loss": 0.8427,
+      "step": 20789
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.5164e-06,
+      "loss": 0.7905,
+      "step": 20790
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.516131428571428e-06,
+      "loss": 1.3348,
+      "step": 20791
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.515862857142858e-06,
+      "loss": 0.9929,
+      "step": 20792
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.515594285714285e-06,
+      "loss": 1.2519,
+      "step": 20793
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.515325714285715e-06,
+      "loss": 0.5547,
+      "step": 20794
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.515057142857143e-06,
+      "loss": 0.9999,
+      "step": 20795
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.514788571428571e-06,
+      "loss": 0.8673,
+      "step": 20796
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.514520000000001e-06,
+      "loss": 1.1001,
+      "step": 20797
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.514251428571429e-06,
+      "loss": 1.2792,
+      "step": 20798
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.513982857142857e-06,
+      "loss": 1.2687,
+      "step": 20799
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.513714285714285e-06,
+      "loss": 1.0053,
+      "step": 20800
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.513445714285715e-06,
+      "loss": 1.0072,
+      "step": 20801
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.513177142857144e-06,
+      "loss": 1.1807,
+      "step": 20802
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.512908571428572e-06,
+      "loss": 0.9603,
+      "step": 20803
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.51264e-06,
+      "loss": 1.0302,
+      "step": 20804
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.51237142857143e-06,
+      "loss": 1.2971,
+      "step": 20805
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.512102857142856e-06,
+      "loss": 1.4454,
+      "step": 20806
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.511834285714286e-06,
+      "loss": 1.1223,
+      "step": 20807
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.511565714285714e-06,
+      "loss": 1.2636,
+      "step": 20808
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.511297142857142e-06,
+      "loss": 1.0832,
+      "step": 20809
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.511028571428572e-06,
+      "loss": 1.0108,
+      "step": 20810
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.51076e-06,
+      "loss": 1.2202,
+      "step": 20811
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.510491428571429e-06,
+      "loss": 1.3409,
+      "step": 20812
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.510222857142857e-06,
+      "loss": 0.8853,
+      "step": 20813
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.509954285714287e-06,
+      "loss": 1.4091,
+      "step": 20814
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.509685714285715e-06,
+      "loss": 1.4431,
+      "step": 20815
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.509417142857143e-06,
+      "loss": 0.7671,
+      "step": 20816
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.509148571428571e-06,
+      "loss": 1.0784,
+      "step": 20817
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.508880000000001e-06,
+      "loss": 0.9547,
+      "step": 20818
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.508611428571428e-06,
+      "loss": 1.2568,
+      "step": 20819
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.508342857142857e-06,
+      "loss": 1.7197,
+      "step": 20820
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.508074285714286e-06,
+      "loss": 1.2441,
+      "step": 20821
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.507805714285714e-06,
+      "loss": 1.2558,
+      "step": 20822
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.507537142857144e-06,
+      "loss": 1.6376,
+      "step": 20823
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.507268571428572e-06,
+      "loss": 1.1892,
+      "step": 20824
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.507000000000002e-06,
+      "loss": 0.7888,
+      "step": 20825
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.506731428571428e-06,
+      "loss": 1.276,
+      "step": 20826
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.506462857142858e-06,
+      "loss": 1.7461,
+      "step": 20827
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.506194285714286e-06,
+      "loss": 0.8354,
+      "step": 20828
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.505925714285714e-06,
+      "loss": 0.8648,
+      "step": 20829
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.505657142857143e-06,
+      "loss": 1.4541,
+      "step": 20830
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.505388571428572e-06,
+      "loss": 0.7354,
+      "step": 20831
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.505119999999999e-06,
+      "loss": 0.9652,
+      "step": 20832
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.504851428571429e-06,
+      "loss": 1.0221,
+      "step": 20833
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.504582857142857e-06,
+      "loss": 0.8504,
+      "step": 20834
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.504314285714285e-06,
+      "loss": 1.4639,
+      "step": 20835
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.504045714285715e-06,
+      "loss": 1.165,
+      "step": 20836
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.503777142857143e-06,
+      "loss": 1.5358,
+      "step": 20837
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.503508571428573e-06,
+      "loss": 1.207,
+      "step": 20838
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.50324e-06,
+      "loss": 1.1687,
+      "step": 20839
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.50297142857143e-06,
+      "loss": 1.0142,
+      "step": 20840
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.502702857142858e-06,
+      "loss": 1.5594,
+      "step": 20841
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.502434285714286e-06,
+      "loss": 1.0273,
+      "step": 20842
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.502165714285714e-06,
+      "loss": 1.6711,
+      "step": 20843
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.501897142857144e-06,
+      "loss": 1.3171,
+      "step": 20844
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.50162857142857e-06,
+      "loss": 1.1885,
+      "step": 20845
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.50136e-06,
+      "loss": 0.8877,
+      "step": 20846
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.501091428571428e-06,
+      "loss": 1.3418,
+      "step": 20847
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.500822857142858e-06,
+      "loss": 1.037,
+      "step": 20848
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.500554285714286e-06,
+      "loss": 0.9862,
+      "step": 20849
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.500285714285715e-06,
+      "loss": 1.1827,
+      "step": 20850
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.500017142857144e-06,
+      "loss": 1.3121,
+      "step": 20851
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.499748571428571e-06,
+      "loss": 0.8217,
+      "step": 20852
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.49948e-06,
+      "loss": 1.1593,
+      "step": 20853
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.499211428571429e-06,
+      "loss": 0.9687,
+      "step": 20854
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.498942857142857e-06,
+      "loss": 0.8569,
+      "step": 20855
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.498674285714285e-06,
+      "loss": 1.1901,
+      "step": 20856
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.498405714285715e-06,
+      "loss": 1.2277,
+      "step": 20857
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.498137142857142e-06,
+      "loss": 1.2101,
+      "step": 20858
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.497868571428572e-06,
+      "loss": 1.7289,
+      "step": 20859
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.4976e-06,
+      "loss": 1.5997,
+      "step": 20860
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.49733142857143e-06,
+      "loss": 1.2165,
+      "step": 20861
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.497062857142858e-06,
+      "loss": 1.1236,
+      "step": 20862
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.496794285714286e-06,
+      "loss": 0.9034,
+      "step": 20863
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.496525714285716e-06,
+      "loss": 1.2982,
+      "step": 20864
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.496257142857142e-06,
+      "loss": 1.3464,
+      "step": 20865
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.495988571428572e-06,
+      "loss": 1.178,
+      "step": 20866
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.49572e-06,
+      "loss": 1.5589,
+      "step": 20867
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.495451428571429e-06,
+      "loss": 1.3444,
+      "step": 20868
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.495182857142857e-06,
+      "loss": 1.0806,
+      "step": 20869
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.494914285714287e-06,
+      "loss": 1.0371,
+      "step": 20870
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.494645714285713e-06,
+      "loss": 1.3223,
+      "step": 20871
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.494377142857143e-06,
+      "loss": 0.7337,
+      "step": 20872
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.494108571428571e-06,
+      "loss": 1.2308,
+      "step": 20873
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.493840000000001e-06,
+      "loss": 1.37,
+      "step": 20874
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.49357142857143e-06,
+      "loss": 1.3998,
+      "step": 20875
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.493302857142857e-06,
+      "loss": 1.2096,
+      "step": 20876
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.493034285714287e-06,
+      "loss": 0.9756,
+      "step": 20877
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.492765714285714e-06,
+      "loss": 1.9122,
+      "step": 20878
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.492497142857144e-06,
+      "loss": 1.0572,
+      "step": 20879
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.492228571428572e-06,
+      "loss": 1.3338,
+      "step": 20880
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.49196e-06,
+      "loss": 1.19,
+      "step": 20881
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.491691428571428e-06,
+      "loss": 1.2665,
+      "step": 20882
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.491422857142858e-06,
+      "loss": 1.0564,
+      "step": 20883
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.491154285714286e-06,
+      "loss": 0.7953,
+      "step": 20884
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.490885714285714e-06,
+      "loss": 1.285,
+      "step": 20885
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.490617142857142e-06,
+      "loss": 1.1101,
+      "step": 20886
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.490348571428572e-06,
+      "loss": 1.623,
+      "step": 20887
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.49008e-06,
+      "loss": 1.0064,
+      "step": 20888
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.489811428571429e-06,
+      "loss": 1.1546,
+      "step": 20889
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.489542857142859e-06,
+      "loss": 1.6547,
+      "step": 20890
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.489274285714285e-06,
+      "loss": 1.0329,
+      "step": 20891
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.489005714285715e-06,
+      "loss": 1.0767,
+      "step": 20892
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.488737142857143e-06,
+      "loss": 1.2829,
+      "step": 20893
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.488468571428571e-06,
+      "loss": 0.949,
+      "step": 20894
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.4882e-06,
+      "loss": 1.1177,
+      "step": 20895
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.48793142857143e-06,
+      "loss": 0.8359,
+      "step": 20896
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.487662857142858e-06,
+      "loss": 1.0743,
+      "step": 20897
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.487394285714286e-06,
+      "loss": 0.8211,
+      "step": 20898
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.487125714285714e-06,
+      "loss": 1.3524,
+      "step": 20899
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.486857142857144e-06,
+      "loss": 1.3741,
+      "step": 20900
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.486588571428572e-06,
+      "loss": 0.9065,
+      "step": 20901
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.48632e-06,
+      "loss": 1.4114,
+      "step": 20902
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.48605142857143e-06,
+      "loss": 1.178,
+      "step": 20903
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.485782857142856e-06,
+      "loss": 1.1408,
+      "step": 20904
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.485514285714286e-06,
+      "loss": 1.4041,
+      "step": 20905
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.485245714285714e-06,
+      "loss": 0.9617,
+      "step": 20906
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.484977142857144e-06,
+      "loss": 0.8452,
+      "step": 20907
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.48470857142857e-06,
+      "loss": 1.4504,
+      "step": 20908
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.48444e-06,
+      "loss": 1.123,
+      "step": 20909
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.484171428571429e-06,
+      "loss": 1.3502,
+      "step": 20910
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.483902857142857e-06,
+      "loss": 1.2544,
+      "step": 20911
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.483634285714285e-06,
+      "loss": 1.1802,
+      "step": 20912
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.483365714285715e-06,
+      "loss": 1.3616,
+      "step": 20913
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.483097142857143e-06,
+      "loss": 1.4132,
+      "step": 20914
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.482828571428571e-06,
+      "loss": 1.0211,
+      "step": 20915
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.482560000000001e-06,
+      "loss": 0.8034,
+      "step": 20916
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.482291428571428e-06,
+      "loss": 0.5962,
+      "step": 20917
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.482022857142858e-06,
+      "loss": 1.4906,
+      "step": 20918
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.481754285714286e-06,
+      "loss": 1.4643,
+      "step": 20919
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.481485714285716e-06,
+      "loss": 0.9828,
+      "step": 20920
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.481217142857142e-06,
+      "loss": 1.4111,
+      "step": 20921
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.480948571428572e-06,
+      "loss": 1.6453,
+      "step": 20922
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.48068e-06,
+      "loss": 1.3445,
+      "step": 20923
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.480411428571428e-06,
+      "loss": 1.5729,
+      "step": 20924
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.480142857142857e-06,
+      "loss": 1.1473,
+      "step": 20925
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.479874285714286e-06,
+      "loss": 1.1487,
+      "step": 20926
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.479605714285715e-06,
+      "loss": 1.4001,
+      "step": 20927
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.479337142857143e-06,
+      "loss": 1.3989,
+      "step": 20928
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.479068571428573e-06,
+      "loss": 0.9807,
+      "step": 20929
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.478800000000001e-06,
+      "loss": 1.5383,
+      "step": 20930
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.478531428571429e-06,
+      "loss": 1.3571,
+      "step": 20931
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.478262857142857e-06,
+      "loss": 1.2557,
+      "step": 20932
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.477994285714287e-06,
+      "loss": 1.2867,
+      "step": 20933
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.477725714285714e-06,
+      "loss": 1.5353,
+      "step": 20934
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.477457142857143e-06,
+      "loss": 1.1408,
+      "step": 20935
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.477188571428572e-06,
+      "loss": 1.3206,
+      "step": 20936
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.47692e-06,
+      "loss": 1.6748,
+      "step": 20937
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.476651428571428e-06,
+      "loss": 0.8222,
+      "step": 20938
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.476382857142858e-06,
+      "loss": 1.2215,
+      "step": 20939
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.476114285714286e-06,
+      "loss": 1.243,
+      "step": 20940
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.475845714285714e-06,
+      "loss": 0.9587,
+      "step": 20941
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.475577142857144e-06,
+      "loss": 1.3018,
+      "step": 20942
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.475308571428572e-06,
+      "loss": 1.1362,
+      "step": 20943
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.47504e-06,
+      "loss": 1.5097,
+      "step": 20944
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.474771428571429e-06,
+      "loss": 1.3515,
+      "step": 20945
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.474502857142858e-06,
+      "loss": 1.0442,
+      "step": 20946
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.474234285714285e-06,
+      "loss": 1.1925,
+      "step": 20947
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.473965714285715e-06,
+      "loss": 1.2256,
+      "step": 20948
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.473697142857143e-06,
+      "loss": 1.0384,
+      "step": 20949
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.473428571428571e-06,
+      "loss": 1.5385,
+      "step": 20950
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.47316e-06,
+      "loss": 1.4829,
+      "step": 20951
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.47289142857143e-06,
+      "loss": 0.9956,
+      "step": 20952
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.472622857142857e-06,
+      "loss": 1.024,
+      "step": 20953
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.472354285714286e-06,
+      "loss": 1.0536,
+      "step": 20954
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.472085714285715e-06,
+      "loss": 1.101,
+      "step": 20955
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.471817142857144e-06,
+      "loss": 1.173,
+      "step": 20956
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.471548571428572e-06,
+      "loss": 1.1943,
+      "step": 20957
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.47128e-06,
+      "loss": 1.2305,
+      "step": 20958
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.47101142857143e-06,
+      "loss": 1.0107,
+      "step": 20959
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.470742857142856e-06,
+      "loss": 1.451,
+      "step": 20960
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.470474285714286e-06,
+      "loss": 1.1316,
+      "step": 20961
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.470205714285714e-06,
+      "loss": 1.1958,
+      "step": 20962
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.469937142857143e-06,
+      "loss": 1.0495,
+      "step": 20963
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.46966857142857e-06,
+      "loss": 1.2012,
+      "step": 20964
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.4694e-06,
+      "loss": 1.0101,
+      "step": 20965
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.469131428571429e-06,
+      "loss": 1.0251,
+      "step": 20966
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.468862857142857e-06,
+      "loss": 1.1234,
+      "step": 20967
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.468594285714287e-06,
+      "loss": 0.6848,
+      "step": 20968
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.468325714285715e-06,
+      "loss": 0.9517,
+      "step": 20969
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.468057142857143e-06,
+      "loss": 1.2645,
+      "step": 20970
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.467788571428571e-06,
+      "loss": 0.7966,
+      "step": 20971
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.467520000000001e-06,
+      "loss": 1.2988,
+      "step": 20972
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.467251428571428e-06,
+      "loss": 1.4572,
+      "step": 20973
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.466982857142858e-06,
+      "loss": 1.0571,
+      "step": 20974
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.466714285714286e-06,
+      "loss": 1.1143,
+      "step": 20975
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.466445714285716e-06,
+      "loss": 1.2074,
+      "step": 20976
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.466177142857142e-06,
+      "loss": 0.9621,
+      "step": 20977
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.465908571428572e-06,
+      "loss": 1.245,
+      "step": 20978
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.46564e-06,
+      "loss": 1.1583,
+      "step": 20979
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.465371428571428e-06,
+      "loss": 1.2069,
+      "step": 20980
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.465102857142858e-06,
+      "loss": 1.1665,
+      "step": 20981
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.464834285714286e-06,
+      "loss": 1.0169,
+      "step": 20982
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.464565714285715e-06,
+      "loss": 0.9514,
+      "step": 20983
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.464297142857143e-06,
+      "loss": 1.1234,
+      "step": 20984
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.464028571428573e-06,
+      "loss": 1.4368,
+      "step": 20985
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.463759999999999e-06,
+      "loss": 1.2722,
+      "step": 20986
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.463491428571429e-06,
+      "loss": 0.7687,
+      "step": 20987
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.463222857142857e-06,
+      "loss": 1.7053,
+      "step": 20988
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.462954285714287e-06,
+      "loss": 1.0783,
+      "step": 20989
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.462685714285713e-06,
+      "loss": 0.8566,
+      "step": 20990
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.462417142857143e-06,
+      "loss": 0.9142,
+      "step": 20991
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.462148571428572e-06,
+      "loss": 0.7967,
+      "step": 20992
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.46188e-06,
+      "loss": 1.4986,
+      "step": 20993
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.46161142857143e-06,
+      "loss": 0.7882,
+      "step": 20994
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.461342857142858e-06,
+      "loss": 1.1894,
+      "step": 20995
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.461074285714286e-06,
+      "loss": 0.9292,
+      "step": 20996
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.460805714285714e-06,
+      "loss": 0.8992,
+      "step": 20997
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.460537142857144e-06,
+      "loss": 1.3194,
+      "step": 20998
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.460268571428572e-06,
+      "loss": 1.0126,
+      "step": 20999
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.46e-06,
+      "loss": 1.0913,
+      "step": 21000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.459731428571428e-06,
+      "loss": 1.023,
+      "step": 21001
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.459462857142858e-06,
+      "loss": 1.4568,
+      "step": 21002
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.459194285714285e-06,
+      "loss": 1.0413,
+      "step": 21003
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.458925714285715e-06,
+      "loss": 1.0403,
+      "step": 21004
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.458657142857143e-06,
+      "loss": 1.043,
+      "step": 21005
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.458388571428571e-06,
+      "loss": 1.1942,
+      "step": 21006
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.458120000000001e-06,
+      "loss": 0.7542,
+      "step": 21007
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.457851428571429e-06,
+      "loss": 1.1776,
+      "step": 21008
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.457582857142857e-06,
+      "loss": 1.3003,
+      "step": 21009
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.457314285714285e-06,
+      "loss": 1.46,
+      "step": 21010
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.457045714285715e-06,
+      "loss": 1.1095,
+      "step": 21011
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.456777142857144e-06,
+      "loss": 1.1141,
+      "step": 21012
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.456508571428572e-06,
+      "loss": 0.8124,
+      "step": 21013
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.45624e-06,
+      "loss": 0.9819,
+      "step": 21014
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.45597142857143e-06,
+      "loss": 1.197,
+      "step": 21015
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.455702857142856e-06,
+      "loss": 1.2327,
+      "step": 21016
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.455434285714286e-06,
+      "loss": 1.2713,
+      "step": 21017
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.455165714285714e-06,
+      "loss": 1.4233,
+      "step": 21018
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.454897142857142e-06,
+      "loss": 1.1111,
+      "step": 21019
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.454628571428572e-06,
+      "loss": 0.8813,
+      "step": 21020
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.45436e-06,
+      "loss": 0.9124,
+      "step": 21021
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.454091428571429e-06,
+      "loss": 1.2791,
+      "step": 21022
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.453822857142857e-06,
+      "loss": 1.2119,
+      "step": 21023
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.453554285714287e-06,
+      "loss": 1.4302,
+      "step": 21024
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.453285714285715e-06,
+      "loss": 1.1565,
+      "step": 21025
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.453017142857143e-06,
+      "loss": 0.8475,
+      "step": 21026
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.452748571428571e-06,
+      "loss": 1.236,
+      "step": 21027
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.452480000000001e-06,
+      "loss": 1.1088,
+      "step": 21028
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.452211428571428e-06,
+      "loss": 1.2696,
+      "step": 21029
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.451942857142857e-06,
+      "loss": 0.9629,
+      "step": 21030
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.451674285714286e-06,
+      "loss": 0.8865,
+      "step": 21031
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.451405714285714e-06,
+      "loss": 1.0031,
+      "step": 21032
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.451137142857144e-06,
+      "loss": 1.2961,
+      "step": 21033
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.450868571428572e-06,
+      "loss": 1.0881,
+      "step": 21034
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.450600000000002e-06,
+      "loss": 0.9975,
+      "step": 21035
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.450331428571428e-06,
+      "loss": 1.2229,
+      "step": 21036
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.450062857142858e-06,
+      "loss": 1.1907,
+      "step": 21037
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.449794285714286e-06,
+      "loss": 1.1823,
+      "step": 21038
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.449525714285714e-06,
+      "loss": 1.2657,
+      "step": 21039
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.449257142857143e-06,
+      "loss": 0.9238,
+      "step": 21040
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.448988571428572e-06,
+      "loss": 1.0248,
+      "step": 21041
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.448719999999999e-06,
+      "loss": 0.7886,
+      "step": 21042
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.448451428571429e-06,
+      "loss": 0.8575,
+      "step": 21043
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.448182857142857e-06,
+      "loss": 1.28,
+      "step": 21044
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.447914285714285e-06,
+      "loss": 0.9399,
+      "step": 21045
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.447645714285715e-06,
+      "loss": 1.4048,
+      "step": 21046
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.447377142857143e-06,
+      "loss": 1.1644,
+      "step": 21047
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.447108571428573e-06,
+      "loss": 0.8515,
+      "step": 21048
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.44684e-06,
+      "loss": 1.4089,
+      "step": 21049
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.44657142857143e-06,
+      "loss": 1.1564,
+      "step": 21050
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.446302857142858e-06,
+      "loss": 1.0784,
+      "step": 21051
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.446034285714286e-06,
+      "loss": 1.1304,
+      "step": 21052
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.445765714285714e-06,
+      "loss": 1.2974,
+      "step": 21053
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.445497142857144e-06,
+      "loss": 1.0953,
+      "step": 21054
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.44522857142857e-06,
+      "loss": 1.3742,
+      "step": 21055
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.44496e-06,
+      "loss": 1.1616,
+      "step": 21056
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.444691428571428e-06,
+      "loss": 0.7878,
+      "step": 21057
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.444422857142858e-06,
+      "loss": 1.2949,
+      "step": 21058
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.444154285714286e-06,
+      "loss": 0.8826,
+      "step": 21059
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.443885714285715e-06,
+      "loss": 1.0593,
+      "step": 21060
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.443617142857144e-06,
+      "loss": 1.0919,
+      "step": 21061
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.443348571428571e-06,
+      "loss": 1.0505,
+      "step": 21062
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.44308e-06,
+      "loss": 1.2686,
+      "step": 21063
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.442811428571429e-06,
+      "loss": 0.9545,
+      "step": 21064
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.442542857142857e-06,
+      "loss": 1.0627,
+      "step": 21065
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.442274285714285e-06,
+      "loss": 1.3273,
+      "step": 21066
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.442005714285715e-06,
+      "loss": 1.3295,
+      "step": 21067
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.441737142857142e-06,
+      "loss": 0.994,
+      "step": 21068
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.441468571428572e-06,
+      "loss": 1.5871,
+      "step": 21069
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.4412e-06,
+      "loss": 1.2084,
+      "step": 21070
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.44093142857143e-06,
+      "loss": 1.0944,
+      "step": 21071
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.440662857142858e-06,
+      "loss": 0.8305,
+      "step": 21072
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.440394285714286e-06,
+      "loss": 0.9217,
+      "step": 21073
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.440125714285716e-06,
+      "loss": 1.6596,
+      "step": 21074
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.439857142857142e-06,
+      "loss": 1.0552,
+      "step": 21075
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.439588571428572e-06,
+      "loss": 0.9744,
+      "step": 21076
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.43932e-06,
+      "loss": 1.1294,
+      "step": 21077
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.439051428571429e-06,
+      "loss": 1.2308,
+      "step": 21078
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.438782857142857e-06,
+      "loss": 1.0446,
+      "step": 21079
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.438514285714287e-06,
+      "loss": 0.9501,
+      "step": 21080
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.438245714285715e-06,
+      "loss": 1.4146,
+      "step": 21081
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.437977142857143e-06,
+      "loss": 0.9319,
+      "step": 21082
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.437708571428571e-06,
+      "loss": 0.9979,
+      "step": 21083
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.437440000000001e-06,
+      "loss": 1.1377,
+      "step": 21084
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.43717142857143e-06,
+      "loss": 1.3377,
+      "step": 21085
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.436902857142857e-06,
+      "loss": 1.2582,
+      "step": 21086
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.436634285714287e-06,
+      "loss": 1.0487,
+      "step": 21087
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.436365714285714e-06,
+      "loss": 0.9507,
+      "step": 21088
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.436097142857144e-06,
+      "loss": 1.0888,
+      "step": 21089
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.435828571428572e-06,
+      "loss": 1.1746,
+      "step": 21090
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.43556e-06,
+      "loss": 1.1582,
+      "step": 21091
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.435291428571428e-06,
+      "loss": 1.4265,
+      "step": 21092
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.435022857142858e-06,
+      "loss": 0.8065,
+      "step": 21093
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.434754285714286e-06,
+      "loss": 1.0358,
+      "step": 21094
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.434485714285714e-06,
+      "loss": 1.2506,
+      "step": 21095
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.434217142857142e-06,
+      "loss": 1.3501,
+      "step": 21096
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.433948571428572e-06,
+      "loss": 1.2035,
+      "step": 21097
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.43368e-06,
+      "loss": 1.2278,
+      "step": 21098
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.433411428571429e-06,
+      "loss": 1.5321,
+      "step": 21099
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.433142857142859e-06,
+      "loss": 1.3073,
+      "step": 21100
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.432874285714285e-06,
+      "loss": 1.2745,
+      "step": 21101
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.432605714285715e-06,
+      "loss": 0.9672,
+      "step": 21102
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.432337142857143e-06,
+      "loss": 1.3603,
+      "step": 21103
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.432068571428573e-06,
+      "loss": 1.0162,
+      "step": 21104
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.4318e-06,
+      "loss": 1.2741,
+      "step": 21105
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.43153142857143e-06,
+      "loss": 1.1246,
+      "step": 21106
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.431262857142858e-06,
+      "loss": 1.2017,
+      "step": 21107
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.430994285714286e-06,
+      "loss": 1.3368,
+      "step": 21108
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.430725714285714e-06,
+      "loss": 1.0722,
+      "step": 21109
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.430457142857144e-06,
+      "loss": 1.4884,
+      "step": 21110
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.430188571428572e-06,
+      "loss": 0.9508,
+      "step": 21111
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.42992e-06,
+      "loss": 1.1327,
+      "step": 21112
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.42965142857143e-06,
+      "loss": 1.1096,
+      "step": 21113
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.429382857142856e-06,
+      "loss": 1.322,
+      "step": 21114
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.429114285714286e-06,
+      "loss": 1.0102,
+      "step": 21115
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.428845714285714e-06,
+      "loss": 1.1339,
+      "step": 21116
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.428577142857144e-06,
+      "loss": 1.1548,
+      "step": 21117
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.42830857142857e-06,
+      "loss": 1.4477,
+      "step": 21118
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.42804e-06,
+      "loss": 1.0556,
+      "step": 21119
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.427771428571429e-06,
+      "loss": 1.4753,
+      "step": 21120
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.427502857142857e-06,
+      "loss": 1.0893,
+      "step": 21121
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.427234285714285e-06,
+      "loss": 1.0209,
+      "step": 21122
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.426965714285715e-06,
+      "loss": 0.8751,
+      "step": 21123
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.426697142857143e-06,
+      "loss": 1.1635,
+      "step": 21124
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.426428571428571e-06,
+      "loss": 1.4482,
+      "step": 21125
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.426160000000001e-06,
+      "loss": 1.0633,
+      "step": 21126
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.42589142857143e-06,
+      "loss": 1.2705,
+      "step": 21127
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.425622857142858e-06,
+      "loss": 1.0182,
+      "step": 21128
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.425354285714286e-06,
+      "loss": 0.9056,
+      "step": 21129
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.425085714285716e-06,
+      "loss": 0.835,
+      "step": 21130
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.424817142857142e-06,
+      "loss": 1.1322,
+      "step": 21131
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.424548571428572e-06,
+      "loss": 1.3163,
+      "step": 21132
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.42428e-06,
+      "loss": 1.1038,
+      "step": 21133
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.424011428571428e-06,
+      "loss": 1.2135,
+      "step": 21134
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.423742857142857e-06,
+      "loss": 1.1553,
+      "step": 21135
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.423474285714286e-06,
+      "loss": 1.1502,
+      "step": 21136
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.423205714285715e-06,
+      "loss": 1.1452,
+      "step": 21137
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.422937142857143e-06,
+      "loss": 1.1916,
+      "step": 21138
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.422668571428573e-06,
+      "loss": 1.3418,
+      "step": 21139
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.422400000000001e-06,
+      "loss": 1.4255,
+      "step": 21140
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.422131428571429e-06,
+      "loss": 1.339,
+      "step": 21141
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.421862857142857e-06,
+      "loss": 1.5114,
+      "step": 21142
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.421594285714287e-06,
+      "loss": 1.2233,
+      "step": 21143
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.421325714285714e-06,
+      "loss": 1.0447,
+      "step": 21144
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.421057142857143e-06,
+      "loss": 1.4644,
+      "step": 21145
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.420788571428572e-06,
+      "loss": 0.9499,
+      "step": 21146
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.42052e-06,
+      "loss": 1.032,
+      "step": 21147
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.420251428571428e-06,
+      "loss": 0.9283,
+      "step": 21148
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.419982857142858e-06,
+      "loss": 1.0555,
+      "step": 21149
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.419714285714286e-06,
+      "loss": 1.3988,
+      "step": 21150
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.419445714285714e-06,
+      "loss": 0.9689,
+      "step": 21151
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.419177142857144e-06,
+      "loss": 0.9557,
+      "step": 21152
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.418908571428572e-06,
+      "loss": 1.3411,
+      "step": 21153
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.41864e-06,
+      "loss": 1.0698,
+      "step": 21154
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.418371428571429e-06,
+      "loss": 0.9748,
+      "step": 21155
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.418102857142858e-06,
+      "loss": 1.0666,
+      "step": 21156
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.417834285714285e-06,
+      "loss": 1.1485,
+      "step": 21157
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.417565714285715e-06,
+      "loss": 1.2287,
+      "step": 21158
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.417297142857143e-06,
+      "loss": 1.1556,
+      "step": 21159
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.417028571428571e-06,
+      "loss": 0.9738,
+      "step": 21160
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.41676e-06,
+      "loss": 0.6243,
+      "step": 21161
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.41649142857143e-06,
+      "loss": 0.8737,
+      "step": 21162
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.416222857142857e-06,
+      "loss": 1.0951,
+      "step": 21163
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.415954285714286e-06,
+      "loss": 1.0394,
+      "step": 21164
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.415685714285715e-06,
+      "loss": 1.3772,
+      "step": 21165
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.415417142857144e-06,
+      "loss": 1.5843,
+      "step": 21166
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.415148571428572e-06,
+      "loss": 1.3564,
+      "step": 21167
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.41488e-06,
+      "loss": 1.168,
+      "step": 21168
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.41461142857143e-06,
+      "loss": 1.3412,
+      "step": 21169
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.414342857142856e-06,
+      "loss": 1.4735,
+      "step": 21170
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.414074285714286e-06,
+      "loss": 1.1392,
+      "step": 21171
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.413805714285714e-06,
+      "loss": 1.2109,
+      "step": 21172
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.413537142857144e-06,
+      "loss": 1.5666,
+      "step": 21173
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.41326857142857e-06,
+      "loss": 1.5483,
+      "step": 21174
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.413e-06,
+      "loss": 1.0176,
+      "step": 21175
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.412731428571429e-06,
+      "loss": 1.4579,
+      "step": 21176
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.412462857142857e-06,
+      "loss": 1.5544,
+      "step": 21177
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.412194285714287e-06,
+      "loss": 1.2988,
+      "step": 21178
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.411925714285715e-06,
+      "loss": 0.9797,
+      "step": 21179
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.411657142857143e-06,
+      "loss": 1.1186,
+      "step": 21180
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.411388571428571e-06,
+      "loss": 1.1523,
+      "step": 21181
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.411120000000001e-06,
+      "loss": 0.9298,
+      "step": 21182
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.410851428571428e-06,
+      "loss": 1.1797,
+      "step": 21183
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.410582857142858e-06,
+      "loss": 1.1887,
+      "step": 21184
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.410314285714286e-06,
+      "loss": 1.4016,
+      "step": 21185
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.410045714285716e-06,
+      "loss": 1.19,
+      "step": 21186
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.409777142857142e-06,
+      "loss": 1.3977,
+      "step": 21187
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.409508571428572e-06,
+      "loss": 0.878,
+      "step": 21188
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.40924e-06,
+      "loss": 1.2105,
+      "step": 21189
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.408971428571428e-06,
+      "loss": 1.1526,
+      "step": 21190
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.408702857142858e-06,
+      "loss": 1.2416,
+      "step": 21191
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.408434285714286e-06,
+      "loss": 1.107,
+      "step": 21192
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.408165714285715e-06,
+      "loss": 1.3252,
+      "step": 21193
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.407897142857143e-06,
+      "loss": 0.8415,
+      "step": 21194
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.407628571428573e-06,
+      "loss": 1.1062,
+      "step": 21195
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.407359999999999e-06,
+      "loss": 0.8193,
+      "step": 21196
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.407091428571429e-06,
+      "loss": 1.1672,
+      "step": 21197
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.406822857142857e-06,
+      "loss": 1.2986,
+      "step": 21198
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.406554285714287e-06,
+      "loss": 0.7512,
+      "step": 21199
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.406285714285713e-06,
+      "loss": 0.9293,
+      "step": 21200
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.406017142857143e-06,
+      "loss": 1.0699,
+      "step": 21201
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.405748571428572e-06,
+      "loss": 1.2223,
+      "step": 21202
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.40548e-06,
+      "loss": 1.0813,
+      "step": 21203
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.40521142857143e-06,
+      "loss": 1.0036,
+      "step": 21204
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.404942857142858e-06,
+      "loss": 1.3198,
+      "step": 21205
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.404674285714286e-06,
+      "loss": 1.2886,
+      "step": 21206
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.404405714285714e-06,
+      "loss": 0.8091,
+      "step": 21207
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.404137142857144e-06,
+      "loss": 1.1309,
+      "step": 21208
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.403868571428572e-06,
+      "loss": 1.3447,
+      "step": 21209
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.4036e-06,
+      "loss": 1.783,
+      "step": 21210
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.403331428571428e-06,
+      "loss": 0.8883,
+      "step": 21211
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.403062857142858e-06,
+      "loss": 1.1929,
+      "step": 21212
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.402794285714285e-06,
+      "loss": 1.081,
+      "step": 21213
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.402525714285715e-06,
+      "loss": 1.1188,
+      "step": 21214
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.402257142857143e-06,
+      "loss": 1.0779,
+      "step": 21215
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.401988571428571e-06,
+      "loss": 1.0104,
+      "step": 21216
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.401720000000001e-06,
+      "loss": 1.4504,
+      "step": 21217
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.401451428571429e-06,
+      "loss": 0.9534,
+      "step": 21218
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.401182857142857e-06,
+      "loss": 1.1376,
+      "step": 21219
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.400914285714285e-06,
+      "loss": 0.9617,
+      "step": 21220
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.400645714285715e-06,
+      "loss": 0.9411,
+      "step": 21221
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.400377142857144e-06,
+      "loss": 1.4345,
+      "step": 21222
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.400108571428572e-06,
+      "loss": 0.8261,
+      "step": 21223
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.39984e-06,
+      "loss": 1.589,
+      "step": 21224
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.39957142857143e-06,
+      "loss": 1.2957,
+      "step": 21225
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.399302857142856e-06,
+      "loss": 1.3821,
+      "step": 21226
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.399034285714286e-06,
+      "loss": 0.9961,
+      "step": 21227
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.398765714285714e-06,
+      "loss": 0.8859,
+      "step": 21228
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.398497142857142e-06,
+      "loss": 1.4426,
+      "step": 21229
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.398228571428572e-06,
+      "loss": 1.2881,
+      "step": 21230
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.39796e-06,
+      "loss": 1.2441,
+      "step": 21231
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.39769142857143e-06,
+      "loss": 1.0647,
+      "step": 21232
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.397422857142857e-06,
+      "loss": 1.0821,
+      "step": 21233
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.397154285714287e-06,
+      "loss": 1.4424,
+      "step": 21234
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.396885714285715e-06,
+      "loss": 1.4994,
+      "step": 21235
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.396617142857143e-06,
+      "loss": 1.328,
+      "step": 21236
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.396348571428571e-06,
+      "loss": 0.9436,
+      "step": 21237
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.396080000000001e-06,
+      "loss": 1.4203,
+      "step": 21238
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.395811428571428e-06,
+      "loss": 1.0507,
+      "step": 21239
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.395542857142857e-06,
+      "loss": 1.4259,
+      "step": 21240
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.395274285714286e-06,
+      "loss": 1.414,
+      "step": 21241
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.395005714285714e-06,
+      "loss": 1.1348,
+      "step": 21242
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.394737142857144e-06,
+      "loss": 0.9073,
+      "step": 21243
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.394468571428572e-06,
+      "loss": 1.4613,
+      "step": 21244
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.394200000000002e-06,
+      "loss": 1.4898,
+      "step": 21245
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.393931428571428e-06,
+      "loss": 1.025,
+      "step": 21246
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.393662857142858e-06,
+      "loss": 1.1191,
+      "step": 21247
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.393394285714286e-06,
+      "loss": 1.3045,
+      "step": 21248
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.393125714285714e-06,
+      "loss": 1.1638,
+      "step": 21249
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.392857142857143e-06,
+      "loss": 1.0793,
+      "step": 21250
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.392588571428572e-06,
+      "loss": 1.0941,
+      "step": 21251
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.392319999999999e-06,
+      "loss": 0.7385,
+      "step": 21252
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.392051428571429e-06,
+      "loss": 1.2218,
+      "step": 21253
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.391782857142857e-06,
+      "loss": 1.3494,
+      "step": 21254
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.391514285714287e-06,
+      "loss": 1.0501,
+      "step": 21255
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.391245714285715e-06,
+      "loss": 1.1428,
+      "step": 21256
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.390977142857143e-06,
+      "loss": 0.9178,
+      "step": 21257
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.390708571428573e-06,
+      "loss": 1.1527,
+      "step": 21258
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.39044e-06,
+      "loss": 1.2457,
+      "step": 21259
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.39017142857143e-06,
+      "loss": 1.3258,
+      "step": 21260
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.389902857142858e-06,
+      "loss": 1.4271,
+      "step": 21261
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.389634285714286e-06,
+      "loss": 0.7296,
+      "step": 21262
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.389365714285714e-06,
+      "loss": 0.8931,
+      "step": 21263
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.389097142857144e-06,
+      "loss": 1.2457,
+      "step": 21264
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.38882857142857e-06,
+      "loss": 1.3709,
+      "step": 21265
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.38856e-06,
+      "loss": 1.2796,
+      "step": 21266
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.388291428571428e-06,
+      "loss": 1.5654,
+      "step": 21267
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.388022857142858e-06,
+      "loss": 1.1063,
+      "step": 21268
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.387754285714286e-06,
+      "loss": 1.21,
+      "step": 21269
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.387485714285715e-06,
+      "loss": 1.0414,
+      "step": 21270
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.387217142857144e-06,
+      "loss": 1.043,
+      "step": 21271
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.386948571428571e-06,
+      "loss": 1.1723,
+      "step": 21272
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.38668e-06,
+      "loss": 0.9512,
+      "step": 21273
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.386411428571429e-06,
+      "loss": 1.0942,
+      "step": 21274
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.386142857142857e-06,
+      "loss": 1.5611,
+      "step": 21275
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.385874285714285e-06,
+      "loss": 1.0561,
+      "step": 21276
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.385605714285715e-06,
+      "loss": 1.2478,
+      "step": 21277
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.385337142857143e-06,
+      "loss": 1.1793,
+      "step": 21278
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.385068571428572e-06,
+      "loss": 1.2526,
+      "step": 21279
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.3848e-06,
+      "loss": 1.6912,
+      "step": 21280
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.38453142857143e-06,
+      "loss": 0.9975,
+      "step": 21281
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.384262857142858e-06,
+      "loss": 1.4749,
+      "step": 21282
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.383994285714286e-06,
+      "loss": 1.0315,
+      "step": 21283
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.383725714285716e-06,
+      "loss": 1.1259,
+      "step": 21284
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.383457142857142e-06,
+      "loss": 1.2643,
+      "step": 21285
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.383188571428572e-06,
+      "loss": 0.9983,
+      "step": 21286
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.38292e-06,
+      "loss": 1.3635,
+      "step": 21287
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.382651428571429e-06,
+      "loss": 1.0804,
+      "step": 21288
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.382382857142857e-06,
+      "loss": 1.4412,
+      "step": 21289
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.382114285714287e-06,
+      "loss": 1.1107,
+      "step": 21290
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.381845714285715e-06,
+      "loss": 1.1396,
+      "step": 21291
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.381577142857143e-06,
+      "loss": 1.1582,
+      "step": 21292
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.381308571428571e-06,
+      "loss": 1.4837,
+      "step": 21293
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.381040000000001e-06,
+      "loss": 1.0386,
+      "step": 21294
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.38077142857143e-06,
+      "loss": 0.8973,
+      "step": 21295
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.380502857142857e-06,
+      "loss": 1.1937,
+      "step": 21296
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.380234285714287e-06,
+      "loss": 1.5408,
+      "step": 21297
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.379965714285714e-06,
+      "loss": 1.2312,
+      "step": 21298
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.379697142857144e-06,
+      "loss": 0.7338,
+      "step": 21299
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.379428571428572e-06,
+      "loss": 0.8076,
+      "step": 21300
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.379160000000002e-06,
+      "loss": 1.3855,
+      "step": 21301
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.378891428571428e-06,
+      "loss": 0.8191,
+      "step": 21302
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.378622857142858e-06,
+      "loss": 1.4002,
+      "step": 21303
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.378354285714286e-06,
+      "loss": 1.4494,
+      "step": 21304
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.378085714285714e-06,
+      "loss": 1.3854,
+      "step": 21305
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.377817142857142e-06,
+      "loss": 1.0979,
+      "step": 21306
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.377548571428572e-06,
+      "loss": 1.1312,
+      "step": 21307
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.37728e-06,
+      "loss": 1.4018,
+      "step": 21308
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.377011428571429e-06,
+      "loss": 1.504,
+      "step": 21309
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.376742857142859e-06,
+      "loss": 1.3527,
+      "step": 21310
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.376474285714285e-06,
+      "loss": 1.3011,
+      "step": 21311
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.376205714285715e-06,
+      "loss": 1.2144,
+      "step": 21312
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.375937142857143e-06,
+      "loss": 1.3964,
+      "step": 21313
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.375668571428573e-06,
+      "loss": 1.4263,
+      "step": 21314
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.3754e-06,
+      "loss": 1.4154,
+      "step": 21315
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.37513142857143e-06,
+      "loss": 1.1044,
+      "step": 21316
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.374862857142858e-06,
+      "loss": 1.4072,
+      "step": 21317
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.374594285714286e-06,
+      "loss": 1.2863,
+      "step": 21318
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.374325714285714e-06,
+      "loss": 0.5077,
+      "step": 21319
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.374057142857144e-06,
+      "loss": 1.1862,
+      "step": 21320
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.373788571428572e-06,
+      "loss": 0.9625,
+      "step": 21321
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.37352e-06,
+      "loss": 1.1792,
+      "step": 21322
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.37325142857143e-06,
+      "loss": 0.9248,
+      "step": 21323
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.372982857142858e-06,
+      "loss": 1.1248,
+      "step": 21324
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.372714285714286e-06,
+      "loss": 1.1199,
+      "step": 21325
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.372445714285714e-06,
+      "loss": 0.8737,
+      "step": 21326
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.372177142857144e-06,
+      "loss": 1.2366,
+      "step": 21327
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.37190857142857e-06,
+      "loss": 1.3035,
+      "step": 21328
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.37164e-06,
+      "loss": 1.4525,
+      "step": 21329
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.371371428571429e-06,
+      "loss": 1.3103,
+      "step": 21330
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.371102857142857e-06,
+      "loss": 1.2368,
+      "step": 21331
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.370834285714285e-06,
+      "loss": 1.2064,
+      "step": 21332
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.370565714285715e-06,
+      "loss": 0.9045,
+      "step": 21333
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.370297142857143e-06,
+      "loss": 1.2339,
+      "step": 21334
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.370028571428571e-06,
+      "loss": 0.9707,
+      "step": 21335
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.369760000000001e-06,
+      "loss": 0.8388,
+      "step": 21336
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.36949142857143e-06,
+      "loss": 1.5201,
+      "step": 21337
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.369222857142858e-06,
+      "loss": 0.8709,
+      "step": 21338
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.368954285714286e-06,
+      "loss": 1.0018,
+      "step": 21339
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.368685714285716e-06,
+      "loss": 0.7764,
+      "step": 21340
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.368417142857142e-06,
+      "loss": 0.9561,
+      "step": 21341
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.368148571428572e-06,
+      "loss": 1.6327,
+      "step": 21342
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.36788e-06,
+      "loss": 1.5588,
+      "step": 21343
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.367611428571428e-06,
+      "loss": 0.9166,
+      "step": 21344
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.367342857142857e-06,
+      "loss": 1.1825,
+      "step": 21345
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.367074285714286e-06,
+      "loss": 1.2734,
+      "step": 21346
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.366805714285715e-06,
+      "loss": 1.2715,
+      "step": 21347
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.366537142857143e-06,
+      "loss": 1.0694,
+      "step": 21348
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.366268571428573e-06,
+      "loss": 1.2703,
+      "step": 21349
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.366000000000001e-06,
+      "loss": 1.729,
+      "step": 21350
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.365731428571429e-06,
+      "loss": 0.6661,
+      "step": 21351
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.365462857142857e-06,
+      "loss": 1.0198,
+      "step": 21352
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.365194285714287e-06,
+      "loss": 1.0848,
+      "step": 21353
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.364925714285714e-06,
+      "loss": 0.7661,
+      "step": 21354
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.364657142857143e-06,
+      "loss": 1.1285,
+      "step": 21355
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.364388571428572e-06,
+      "loss": 1.2056,
+      "step": 21356
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.36412e-06,
+      "loss": 0.9752,
+      "step": 21357
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.363851428571428e-06,
+      "loss": 1.2096,
+      "step": 21358
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.363582857142858e-06,
+      "loss": 1.0303,
+      "step": 21359
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.363314285714286e-06,
+      "loss": 1.0139,
+      "step": 21360
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.363045714285714e-06,
+      "loss": 1.5767,
+      "step": 21361
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.362777142857144e-06,
+      "loss": 1.4155,
+      "step": 21362
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.362508571428572e-06,
+      "loss": 1.1589,
+      "step": 21363
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.36224e-06,
+      "loss": 1.3231,
+      "step": 21364
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.361971428571429e-06,
+      "loss": 1.2527,
+      "step": 21365
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.361702857142858e-06,
+      "loss": 1.4783,
+      "step": 21366
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.361434285714285e-06,
+      "loss": 1.3371,
+      "step": 21367
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.361165714285715e-06,
+      "loss": 1.3689,
+      "step": 21368
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.360897142857143e-06,
+      "loss": 1.2487,
+      "step": 21369
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.360628571428571e-06,
+      "loss": 1.3161,
+      "step": 21370
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.36036e-06,
+      "loss": 0.8763,
+      "step": 21371
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.36009142857143e-06,
+      "loss": 1.2461,
+      "step": 21372
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.359822857142857e-06,
+      "loss": 1.2283,
+      "step": 21373
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.359554285714286e-06,
+      "loss": 1.6869,
+      "step": 21374
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.359285714285715e-06,
+      "loss": 1.0394,
+      "step": 21375
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.359017142857144e-06,
+      "loss": 1.3214,
+      "step": 21376
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.358748571428572e-06,
+      "loss": 0.7422,
+      "step": 21377
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.35848e-06,
+      "loss": 1.3984,
+      "step": 21378
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.35821142857143e-06,
+      "loss": 1.3633,
+      "step": 21379
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.357942857142856e-06,
+      "loss": 1.2007,
+      "step": 21380
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.357674285714286e-06,
+      "loss": 0.9483,
+      "step": 21381
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.357405714285714e-06,
+      "loss": 0.9324,
+      "step": 21382
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.357137142857144e-06,
+      "loss": 0.6192,
+      "step": 21383
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.35686857142857e-06,
+      "loss": 1.0519,
+      "step": 21384
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.3566e-06,
+      "loss": 1.173,
+      "step": 21385
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.356331428571429e-06,
+      "loss": 0.6557,
+      "step": 21386
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.356062857142857e-06,
+      "loss": 1.1055,
+      "step": 21387
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.355794285714287e-06,
+      "loss": 1.6174,
+      "step": 21388
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.355525714285715e-06,
+      "loss": 1.0518,
+      "step": 21389
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.355257142857143e-06,
+      "loss": 1.0408,
+      "step": 21390
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.354988571428571e-06,
+      "loss": 1.0982,
+      "step": 21391
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.354720000000001e-06,
+      "loss": 1.16,
+      "step": 21392
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.354451428571428e-06,
+      "loss": 1.2546,
+      "step": 21393
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.354182857142858e-06,
+      "loss": 1.3965,
+      "step": 21394
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.353914285714286e-06,
+      "loss": 1.3294,
+      "step": 21395
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.353645714285716e-06,
+      "loss": 1.3393,
+      "step": 21396
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.353377142857142e-06,
+      "loss": 0.7345,
+      "step": 21397
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.353108571428572e-06,
+      "loss": 1.2492,
+      "step": 21398
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.35284e-06,
+      "loss": 1.086,
+      "step": 21399
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.352571428571428e-06,
+      "loss": 0.7586,
+      "step": 21400
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.352302857142858e-06,
+      "loss": 1.4889,
+      "step": 21401
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.352034285714286e-06,
+      "loss": 1.3093,
+      "step": 21402
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.351765714285715e-06,
+      "loss": 0.9338,
+      "step": 21403
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.351497142857143e-06,
+      "loss": 1.2766,
+      "step": 21404
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.351228571428573e-06,
+      "loss": 1.2322,
+      "step": 21405
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.35096e-06,
+      "loss": 1.1371,
+      "step": 21406
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.350691428571429e-06,
+      "loss": 1.3439,
+      "step": 21407
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.350422857142857e-06,
+      "loss": 1.1686,
+      "step": 21408
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.350154285714287e-06,
+      "loss": 0.7503,
+      "step": 21409
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.349885714285713e-06,
+      "loss": 1.4681,
+      "step": 21410
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.349617142857143e-06,
+      "loss": 0.9973,
+      "step": 21411
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.349348571428572e-06,
+      "loss": 1.463,
+      "step": 21412
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.34908e-06,
+      "loss": 0.8927,
+      "step": 21413
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.34881142857143e-06,
+      "loss": 1.0288,
+      "step": 21414
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.348542857142858e-06,
+      "loss": 0.721,
+      "step": 21415
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.348274285714286e-06,
+      "loss": 1.083,
+      "step": 21416
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.348005714285714e-06,
+      "loss": 1.1237,
+      "step": 21417
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.347737142857144e-06,
+      "loss": 1.0401,
+      "step": 21418
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.347468571428572e-06,
+      "loss": 0.8521,
+      "step": 21419
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.3472e-06,
+      "loss": 1.0507,
+      "step": 21420
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.346931428571428e-06,
+      "loss": 0.8606,
+      "step": 21421
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.346662857142858e-06,
+      "loss": 0.8676,
+      "step": 21422
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.346394285714285e-06,
+      "loss": 0.9989,
+      "step": 21423
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.346125714285715e-06,
+      "loss": 1.1591,
+      "step": 21424
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.345857142857143e-06,
+      "loss": 1.2919,
+      "step": 21425
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.345588571428571e-06,
+      "loss": 1.4766,
+      "step": 21426
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.345320000000001e-06,
+      "loss": 1.0845,
+      "step": 21427
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.345051428571429e-06,
+      "loss": 1.3625,
+      "step": 21428
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.344782857142859e-06,
+      "loss": 1.2797,
+      "step": 21429
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.344514285714285e-06,
+      "loss": 1.3223,
+      "step": 21430
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.344245714285715e-06,
+      "loss": 1.3103,
+      "step": 21431
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.343977142857144e-06,
+      "loss": 1.6701,
+      "step": 21432
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.343708571428572e-06,
+      "loss": 1.4888,
+      "step": 21433
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.34344e-06,
+      "loss": 1.416,
+      "step": 21434
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.34317142857143e-06,
+      "loss": 0.942,
+      "step": 21435
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.342902857142856e-06,
+      "loss": 1.196,
+      "step": 21436
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.342634285714286e-06,
+      "loss": 1.1389,
+      "step": 21437
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.342365714285714e-06,
+      "loss": 0.7084,
+      "step": 21438
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.342097142857142e-06,
+      "loss": 0.9103,
+      "step": 21439
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.341828571428572e-06,
+      "loss": 1.2282,
+      "step": 21440
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.34156e-06,
+      "loss": 1.0393,
+      "step": 21441
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.34129142857143e-06,
+      "loss": 0.9515,
+      "step": 21442
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.341022857142857e-06,
+      "loss": 0.9862,
+      "step": 21443
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.340754285714287e-06,
+      "loss": 0.7923,
+      "step": 21444
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.340485714285715e-06,
+      "loss": 0.7076,
+      "step": 21445
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.340217142857143e-06,
+      "loss": 0.7996,
+      "step": 21446
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.339948571428571e-06,
+      "loss": 1.386,
+      "step": 21447
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.339680000000001e-06,
+      "loss": 1.295,
+      "step": 21448
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.339411428571428e-06,
+      "loss": 1.2293,
+      "step": 21449
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.339142857142857e-06,
+      "loss": 1.4562,
+      "step": 21450
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.338874285714286e-06,
+      "loss": 0.7145,
+      "step": 21451
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.338605714285716e-06,
+      "loss": 1.4255,
+      "step": 21452
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.338337142857144e-06,
+      "loss": 1.0151,
+      "step": 21453
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.338068571428572e-06,
+      "loss": 1.1967,
+      "step": 21454
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.337800000000002e-06,
+      "loss": 1.0045,
+      "step": 21455
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.337531428571428e-06,
+      "loss": 1.612,
+      "step": 21456
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.337262857142858e-06,
+      "loss": 1.6689,
+      "step": 21457
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.336994285714286e-06,
+      "loss": 0.9427,
+      "step": 21458
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.336725714285714e-06,
+      "loss": 0.7842,
+      "step": 21459
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.336457142857143e-06,
+      "loss": 0.9364,
+      "step": 21460
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.336188571428572e-06,
+      "loss": 1.1043,
+      "step": 21461
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.335919999999999e-06,
+      "loss": 1.0273,
+      "step": 21462
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.335651428571429e-06,
+      "loss": 0.9153,
+      "step": 21463
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.335382857142857e-06,
+      "loss": 1.1342,
+      "step": 21464
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.335114285714287e-06,
+      "loss": 1.1377,
+      "step": 21465
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.334845714285715e-06,
+      "loss": 1.3717,
+      "step": 21466
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.334577142857143e-06,
+      "loss": 0.8639,
+      "step": 21467
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.334308571428573e-06,
+      "loss": 0.982,
+      "step": 21468
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.33404e-06,
+      "loss": 1.4951,
+      "step": 21469
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.33377142857143e-06,
+      "loss": 1.1769,
+      "step": 21470
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.333502857142858e-06,
+      "loss": 1.1421,
+      "step": 21471
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.333234285714286e-06,
+      "loss": 0.8542,
+      "step": 21472
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.332965714285714e-06,
+      "loss": 1.3789,
+      "step": 21473
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.332697142857144e-06,
+      "loss": 0.9703,
+      "step": 21474
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.332428571428572e-06,
+      "loss": 0.8526,
+      "step": 21475
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.33216e-06,
+      "loss": 1.14,
+      "step": 21476
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.331891428571428e-06,
+      "loss": 1.1083,
+      "step": 21477
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.331622857142858e-06,
+      "loss": 0.7315,
+      "step": 21478
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.331354285714285e-06,
+      "loss": 1.1638,
+      "step": 21479
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.331085714285715e-06,
+      "loss": 1.2763,
+      "step": 21480
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.330817142857144e-06,
+      "loss": 0.9353,
+      "step": 21481
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.330548571428571e-06,
+      "loss": 0.8227,
+      "step": 21482
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.33028e-06,
+      "loss": 1.2764,
+      "step": 21483
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.330011428571429e-06,
+      "loss": 0.9238,
+      "step": 21484
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.329742857142857e-06,
+      "loss": 0.7444,
+      "step": 21485
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.329474285714285e-06,
+      "loss": 1.2615,
+      "step": 21486
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.329205714285715e-06,
+      "loss": 1.1188,
+      "step": 21487
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.328937142857143e-06,
+      "loss": 0.9761,
+      "step": 21488
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.328668571428572e-06,
+      "loss": 1.4408,
+      "step": 21489
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.3284e-06,
+      "loss": 1.3118,
+      "step": 21490
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.32813142857143e-06,
+      "loss": 1.2994,
+      "step": 21491
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.327862857142856e-06,
+      "loss": 0.8481,
+      "step": 21492
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.327594285714286e-06,
+      "loss": 1.0835,
+      "step": 21493
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.327325714285716e-06,
+      "loss": 1.2011,
+      "step": 21494
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.327057142857142e-06,
+      "loss": 0.9321,
+      "step": 21495
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.326788571428572e-06,
+      "loss": 1.1721,
+      "step": 21496
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.32652e-06,
+      "loss": 0.7242,
+      "step": 21497
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.32625142857143e-06,
+      "loss": 0.8886,
+      "step": 21498
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.325982857142857e-06,
+      "loss": 1.1686,
+      "step": 21499
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.325714285714287e-06,
+      "loss": 1.7764,
+      "step": 21500
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.325445714285715e-06,
+      "loss": 1.203,
+      "step": 21501
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.325177142857143e-06,
+      "loss": 1.256,
+      "step": 21502
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.324908571428571e-06,
+      "loss": 0.9733,
+      "step": 21503
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.324640000000001e-06,
+      "loss": 1.2909,
+      "step": 21504
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.324371428571427e-06,
+      "loss": 1.566,
+      "step": 21505
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.324102857142857e-06,
+      "loss": 1.2471,
+      "step": 21506
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.323834285714287e-06,
+      "loss": 1.3025,
+      "step": 21507
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.323565714285714e-06,
+      "loss": 1.0193,
+      "step": 21508
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.323297142857144e-06,
+      "loss": 0.8279,
+      "step": 21509
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.323028571428572e-06,
+      "loss": 1.3317,
+      "step": 21510
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.322760000000002e-06,
+      "loss": 1.0657,
+      "step": 21511
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.322491428571428e-06,
+      "loss": 1.459,
+      "step": 21512
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.322222857142858e-06,
+      "loss": 1.5599,
+      "step": 21513
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.321954285714286e-06,
+      "loss": 1.3266,
+      "step": 21514
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.321685714285714e-06,
+      "loss": 1.3695,
+      "step": 21515
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.321417142857142e-06,
+      "loss": 1.1798,
+      "step": 21516
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.321148571428572e-06,
+      "loss": 1.1771,
+      "step": 21517
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.320879999999999e-06,
+      "loss": 1.2563,
+      "step": 21518
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.320611428571429e-06,
+      "loss": 1.2582,
+      "step": 21519
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.320342857142859e-06,
+      "loss": 0.9231,
+      "step": 21520
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.320074285714285e-06,
+      "loss": 0.8498,
+      "step": 21521
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.319805714285715e-06,
+      "loss": 0.8506,
+      "step": 21522
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.319537142857143e-06,
+      "loss": 0.9674,
+      "step": 21523
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.319268571428573e-06,
+      "loss": 1.2629,
+      "step": 21524
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.319e-06,
+      "loss": 0.9757,
+      "step": 21525
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31873142857143e-06,
+      "loss": 1.2357,
+      "step": 21526
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.318462857142858e-06,
+      "loss": 1.4069,
+      "step": 21527
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.318194285714286e-06,
+      "loss": 1.2748,
+      "step": 21528
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.317925714285714e-06,
+      "loss": 0.9457,
+      "step": 21529
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.317657142857144e-06,
+      "loss": 1.4031,
+      "step": 21530
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31738857142857e-06,
+      "loss": 1.3047,
+      "step": 21531
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31712e-06,
+      "loss": 1.6067,
+      "step": 21532
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31685142857143e-06,
+      "loss": 0.9265,
+      "step": 21533
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.316582857142858e-06,
+      "loss": 1.4036,
+      "step": 21534
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.316314285714286e-06,
+      "loss": 1.2149,
+      "step": 21535
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.316045714285714e-06,
+      "loss": 1.2255,
+      "step": 21536
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.315777142857144e-06,
+      "loss": 1.129,
+      "step": 21537
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31550857142857e-06,
+      "loss": 1.3001,
+      "step": 21538
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31524e-06,
+      "loss": 1.3142,
+      "step": 21539
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.314971428571429e-06,
+      "loss": 1.1994,
+      "step": 21540
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.314702857142857e-06,
+      "loss": 0.9121,
+      "step": 21541
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.314434285714285e-06,
+      "loss": 1.229,
+      "step": 21542
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.314165714285715e-06,
+      "loss": 0.9526,
+      "step": 21543
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.313897142857142e-06,
+      "loss": 0.8282,
+      "step": 21544
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.313628571428571e-06,
+      "loss": 1.0785,
+      "step": 21545
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.313360000000001e-06,
+      "loss": 0.7026,
+      "step": 21546
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31309142857143e-06,
+      "loss": 0.8218,
+      "step": 21547
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.312822857142858e-06,
+      "loss": 1.1629,
+      "step": 21548
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.312554285714286e-06,
+      "loss": 1.1902,
+      "step": 21549
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.312285714285716e-06,
+      "loss": 1.0237,
+      "step": 21550
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.312017142857142e-06,
+      "loss": 0.9427,
+      "step": 21551
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.311748571428572e-06,
+      "loss": 0.9616,
+      "step": 21552
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.31148e-06,
+      "loss": 1.0665,
+      "step": 21553
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.311211428571428e-06,
+      "loss": 1.4877,
+      "step": 21554
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.310942857142857e-06,
+      "loss": 1.4124,
+      "step": 21555
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.310674285714286e-06,
+      "loss": 1.1156,
+      "step": 21556
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.310405714285715e-06,
+      "loss": 1.1104,
+      "step": 21557
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.310137142857143e-06,
+      "loss": 0.8539,
+      "step": 21558
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.309868571428573e-06,
+      "loss": 1.2261,
+      "step": 21559
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.309600000000001e-06,
+      "loss": 1.1223,
+      "step": 21560
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.309331428571429e-06,
+      "loss": 0.9002,
+      "step": 21561
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.309062857142857e-06,
+      "loss": 1.2158,
+      "step": 21562
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.308794285714287e-06,
+      "loss": 1.7488,
+      "step": 21563
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.308525714285714e-06,
+      "loss": 0.8477,
+      "step": 21564
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.308257142857143e-06,
+      "loss": 1.3857,
+      "step": 21565
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.307988571428572e-06,
+      "loss": 0.8577,
+      "step": 21566
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.30772e-06,
+      "loss": 0.9777,
+      "step": 21567
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.307451428571428e-06,
+      "loss": 1.2334,
+      "step": 21568
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.307182857142858e-06,
+      "loss": 1.4488,
+      "step": 21569
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.306914285714286e-06,
+      "loss": 1.0476,
+      "step": 21570
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.306645714285714e-06,
+      "loss": 1.4526,
+      "step": 21571
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.306377142857144e-06,
+      "loss": 1.3609,
+      "step": 21572
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.306108571428572e-06,
+      "loss": 1.4569,
+      "step": 21573
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.30584e-06,
+      "loss": 1.8342,
+      "step": 21574
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.305571428571429e-06,
+      "loss": 1.0443,
+      "step": 21575
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.305302857142858e-06,
+      "loss": 0.8752,
+      "step": 21576
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.305034285714285e-06,
+      "loss": 1.1227,
+      "step": 21577
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.304765714285715e-06,
+      "loss": 0.6489,
+      "step": 21578
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.304497142857143e-06,
+      "loss": 1.4241,
+      "step": 21579
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.304228571428573e-06,
+      "loss": 1.3725,
+      "step": 21580
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.30396e-06,
+      "loss": 1.2277,
+      "step": 21581
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.30369142857143e-06,
+      "loss": 1.513,
+      "step": 21582
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.303422857142857e-06,
+      "loss": 1.208,
+      "step": 21583
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.303154285714286e-06,
+      "loss": 1.2787,
+      "step": 21584
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.302885714285714e-06,
+      "loss": 1.1746,
+      "step": 21585
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.302617142857144e-06,
+      "loss": 1.1734,
+      "step": 21586
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.302348571428572e-06,
+      "loss": 1.5862,
+      "step": 21587
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.30208e-06,
+      "loss": 0.9402,
+      "step": 21588
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.30181142857143e-06,
+      "loss": 1.0213,
+      "step": 21589
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.301542857142856e-06,
+      "loss": 1.3827,
+      "step": 21590
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.301274285714286e-06,
+      "loss": 1.1422,
+      "step": 21591
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.301005714285714e-06,
+      "loss": 1.652,
+      "step": 21592
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.300737142857144e-06,
+      "loss": 1.0131,
+      "step": 21593
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.30046857142857e-06,
+      "loss": 1.3324,
+      "step": 21594
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.3002e-06,
+      "loss": 0.9804,
+      "step": 21595
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.299931428571429e-06,
+      "loss": 1.0192,
+      "step": 21596
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.299662857142857e-06,
+      "loss": 1.1051,
+      "step": 21597
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.299394285714285e-06,
+      "loss": 1.143,
+      "step": 21598
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.299125714285715e-06,
+      "loss": 1.0029,
+      "step": 21599
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.298857142857143e-06,
+      "loss": 1.091,
+      "step": 21600
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.298588571428571e-06,
+      "loss": 1.0164,
+      "step": 21601
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.298320000000001e-06,
+      "loss": 0.8204,
+      "step": 21602
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.29805142857143e-06,
+      "loss": 1.208,
+      "step": 21603
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.297782857142858e-06,
+      "loss": 1.2416,
+      "step": 21604
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.297514285714286e-06,
+      "loss": 1.1904,
+      "step": 21605
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.297245714285716e-06,
+      "loss": 1.1831,
+      "step": 21606
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.296977142857142e-06,
+      "loss": 1.1843,
+      "step": 21607
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.296708571428572e-06,
+      "loss": 1.0014,
+      "step": 21608
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.29644e-06,
+      "loss": 1.0771,
+      "step": 21609
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.296171428571428e-06,
+      "loss": 1.1639,
+      "step": 21610
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.295902857142856e-06,
+      "loss": 1.101,
+      "step": 21611
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.295634285714286e-06,
+      "loss": 1.3663,
+      "step": 21612
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.295365714285715e-06,
+      "loss": 1.1141,
+      "step": 21613
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.295097142857143e-06,
+      "loss": 1.2221,
+      "step": 21614
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.294828571428573e-06,
+      "loss": 0.9537,
+      "step": 21615
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.29456e-06,
+      "loss": 1.1746,
+      "step": 21616
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.294291428571429e-06,
+      "loss": 1.0085,
+      "step": 21617
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.294022857142857e-06,
+      "loss": 1.2995,
+      "step": 21618
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.293754285714287e-06,
+      "loss": 0.7764,
+      "step": 21619
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.293485714285713e-06,
+      "loss": 1.1314,
+      "step": 21620
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.293217142857143e-06,
+      "loss": 1.1038,
+      "step": 21621
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.292948571428572e-06,
+      "loss": 1.3949,
+      "step": 21622
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.29268e-06,
+      "loss": 1.1534,
+      "step": 21623
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.292411428571428e-06,
+      "loss": 0.8373,
+      "step": 21624
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.292142857142858e-06,
+      "loss": 1.185,
+      "step": 21625
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.291874285714288e-06,
+      "loss": 1.1993,
+      "step": 21626
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.291605714285714e-06,
+      "loss": 1.3086,
+      "step": 21627
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.291337142857144e-06,
+      "loss": 1.2776,
+      "step": 21628
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.291068571428572e-06,
+      "loss": 1.1171,
+      "step": 21629
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.2908e-06,
+      "loss": 1.3143,
+      "step": 21630
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.290531428571428e-06,
+      "loss": 0.7393,
+      "step": 21631
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.290262857142858e-06,
+      "loss": 1.3801,
+      "step": 21632
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.289994285714285e-06,
+      "loss": 0.8798,
+      "step": 21633
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.289725714285715e-06,
+      "loss": 1.2866,
+      "step": 21634
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.289457142857143e-06,
+      "loss": 1.3674,
+      "step": 21635
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.289188571428571e-06,
+      "loss": 1.6704,
+      "step": 21636
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.28892e-06,
+      "loss": 1.1939,
+      "step": 21637
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.288651428571429e-06,
+      "loss": 1.1785,
+      "step": 21638
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.288382857142859e-06,
+      "loss": 0.8612,
+      "step": 21639
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.288114285714285e-06,
+      "loss": 0.9374,
+      "step": 21640
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.287845714285715e-06,
+      "loss": 0.9401,
+      "step": 21641
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.287577142857144e-06,
+      "loss": 1.2706,
+      "step": 21642
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.287308571428572e-06,
+      "loss": 0.8833,
+      "step": 21643
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.28704e-06,
+      "loss": 1.2312,
+      "step": 21644
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.28677142857143e-06,
+      "loss": 1.1069,
+      "step": 21645
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.286502857142856e-06,
+      "loss": 1.3103,
+      "step": 21646
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.286234285714286e-06,
+      "loss": 0.9564,
+      "step": 21647
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.285965714285714e-06,
+      "loss": 1.257,
+      "step": 21648
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.285697142857144e-06,
+      "loss": 1.0238,
+      "step": 21649
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.28542857142857e-06,
+      "loss": 1.0914,
+      "step": 21650
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.28516e-06,
+      "loss": 1.091,
+      "step": 21651
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.28489142857143e-06,
+      "loss": 1.2506,
+      "step": 21652
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.284622857142857e-06,
+      "loss": 1.2823,
+      "step": 21653
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.284354285714287e-06,
+      "loss": 1.284,
+      "step": 21654
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.284085714285715e-06,
+      "loss": 1.2623,
+      "step": 21655
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.283817142857143e-06,
+      "loss": 1.405,
+      "step": 21656
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.283548571428571e-06,
+      "loss": 1.4154,
+      "step": 21657
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.283280000000001e-06,
+      "loss": 1.1438,
+      "step": 21658
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.283011428571428e-06,
+      "loss": 1.1706,
+      "step": 21659
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.282742857142857e-06,
+      "loss": 1.6849,
+      "step": 21660
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.282474285714286e-06,
+      "loss": 1.184,
+      "step": 21661
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.282205714285715e-06,
+      "loss": 1.6692,
+      "step": 21662
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.281937142857142e-06,
+      "loss": 1.3262,
+      "step": 21663
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.281668571428572e-06,
+      "loss": 1.3476,
+      "step": 21664
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.281400000000002e-06,
+      "loss": 1.3897,
+      "step": 21665
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.281131428571428e-06,
+      "loss": 1.5388,
+      "step": 21666
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.280862857142858e-06,
+      "loss": 0.6905,
+      "step": 21667
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.280594285714286e-06,
+      "loss": 1.149,
+      "step": 21668
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.280325714285714e-06,
+      "loss": 1.0568,
+      "step": 21669
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.280057142857143e-06,
+      "loss": 1.1789,
+      "step": 21670
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.279788571428572e-06,
+      "loss": 1.5702,
+      "step": 21671
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.27952e-06,
+      "loss": 1.1057,
+      "step": 21672
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.279251428571429e-06,
+      "loss": 1.2676,
+      "step": 21673
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.278982857142857e-06,
+      "loss": 1.4293,
+      "step": 21674
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.278714285714287e-06,
+      "loss": 1.0789,
+      "step": 21675
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.278445714285713e-06,
+      "loss": 1.2814,
+      "step": 21676
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.278177142857143e-06,
+      "loss": 0.8584,
+      "step": 21677
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.277908571428573e-06,
+      "loss": 0.6381,
+      "step": 21678
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.27764e-06,
+      "loss": 1.2493,
+      "step": 21679
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.27737142857143e-06,
+      "loss": 1.3018,
+      "step": 21680
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.277102857142858e-06,
+      "loss": 1.0745,
+      "step": 21681
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.276834285714286e-06,
+      "loss": 1.0709,
+      "step": 21682
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.276565714285714e-06,
+      "loss": 1.0598,
+      "step": 21683
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.276297142857144e-06,
+      "loss": 0.9877,
+      "step": 21684
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.276028571428572e-06,
+      "loss": 1.1692,
+      "step": 21685
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.27576e-06,
+      "loss": 1.3549,
+      "step": 21686
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.275491428571428e-06,
+      "loss": 1.0855,
+      "step": 21687
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.275222857142858e-06,
+      "loss": 1.4927,
+      "step": 21688
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.274954285714285e-06,
+      "loss": 1.3608,
+      "step": 21689
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.274685714285715e-06,
+      "loss": 1.552,
+      "step": 21690
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.274417142857143e-06,
+      "loss": 1.3884,
+      "step": 21691
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.274148571428571e-06,
+      "loss": 0.9814,
+      "step": 21692
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.27388e-06,
+      "loss": 1.166,
+      "step": 21693
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.273611428571429e-06,
+      "loss": 1.0754,
+      "step": 21694
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.273342857142857e-06,
+      "loss": 1.334,
+      "step": 21695
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.273074285714285e-06,
+      "loss": 1.0447,
+      "step": 21696
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.272805714285715e-06,
+      "loss": 1.0349,
+      "step": 21697
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.272537142857143e-06,
+      "loss": 1.1361,
+      "step": 21698
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.272268571428572e-06,
+      "loss": 1.0968,
+      "step": 21699
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.272e-06,
+      "loss": 1.3345,
+      "step": 21700
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.27173142857143e-06,
+      "loss": 1.4711,
+      "step": 21701
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.271462857142856e-06,
+      "loss": 1.1243,
+      "step": 21702
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.271194285714286e-06,
+      "loss": 1.6231,
+      "step": 21703
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.270925714285714e-06,
+      "loss": 1.5598,
+      "step": 21704
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.270657142857142e-06,
+      "loss": 1.1235,
+      "step": 21705
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.270388571428572e-06,
+      "loss": 1.6209,
+      "step": 21706
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.27012e-06,
+      "loss": 1.0575,
+      "step": 21707
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.26985142857143e-06,
+      "loss": 1.0317,
+      "step": 21708
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.269582857142857e-06,
+      "loss": 1.7058,
+      "step": 21709
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.269314285714287e-06,
+      "loss": 1.419,
+      "step": 21710
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.269045714285715e-06,
+      "loss": 0.9012,
+      "step": 21711
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.268777142857143e-06,
+      "loss": 1.4101,
+      "step": 21712
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.268508571428571e-06,
+      "loss": 1.5116,
+      "step": 21713
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.268240000000001e-06,
+      "loss": 1.3336,
+      "step": 21714
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.267971428571427e-06,
+      "loss": 1.2773,
+      "step": 21715
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.267702857142857e-06,
+      "loss": 1.3461,
+      "step": 21716
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.267434285714286e-06,
+      "loss": 0.8802,
+      "step": 21717
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.267165714285714e-06,
+      "loss": 1.3258,
+      "step": 21718
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.266897142857144e-06,
+      "loss": 1.1345,
+      "step": 21719
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.266628571428572e-06,
+      "loss": 1.5055,
+      "step": 21720
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.266360000000002e-06,
+      "loss": 1.6185,
+      "step": 21721
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.266091428571428e-06,
+      "loss": 1.2281,
+      "step": 21722
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.265822857142858e-06,
+      "loss": 1.2897,
+      "step": 21723
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.265554285714286e-06,
+      "loss": 0.9989,
+      "step": 21724
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.265285714285714e-06,
+      "loss": 1.3647,
+      "step": 21725
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.265017142857142e-06,
+      "loss": 0.9574,
+      "step": 21726
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.264748571428572e-06,
+      "loss": 1.1491,
+      "step": 21727
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.264479999999999e-06,
+      "loss": 1.1358,
+      "step": 21728
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.264211428571429e-06,
+      "loss": 1.2795,
+      "step": 21729
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.263942857142857e-06,
+      "loss": 0.9175,
+      "step": 21730
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.263674285714287e-06,
+      "loss": 1.4302,
+      "step": 21731
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.263405714285715e-06,
+      "loss": 1.3032,
+      "step": 21732
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.263137142857143e-06,
+      "loss": 1.0967,
+      "step": 21733
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.262868571428573e-06,
+      "loss": 1.0127,
+      "step": 21734
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.2626e-06,
+      "loss": 1.088,
+      "step": 21735
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.26233142857143e-06,
+      "loss": 1.21,
+      "step": 21736
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.262062857142858e-06,
+      "loss": 1.1441,
+      "step": 21737
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.261794285714286e-06,
+      "loss": 1.0886,
+      "step": 21738
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.261525714285714e-06,
+      "loss": 0.6904,
+      "step": 21739
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.261257142857144e-06,
+      "loss": 1.4275,
+      "step": 21740
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.26098857142857e-06,
+      "loss": 1.4607,
+      "step": 21741
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.26072e-06,
+      "loss": 1.1462,
+      "step": 21742
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.260451428571428e-06,
+      "loss": 0.787,
+      "step": 21743
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.260182857142858e-06,
+      "loss": 1.0338,
+      "step": 21744
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.259914285714286e-06,
+      "loss": 1.5844,
+      "step": 21745
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.259645714285714e-06,
+      "loss": 1.1291,
+      "step": 21746
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.259377142857144e-06,
+      "loss": 1.0696,
+      "step": 21747
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.25910857142857e-06,
+      "loss": 0.8938,
+      "step": 21748
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.25884e-06,
+      "loss": 1.0666,
+      "step": 21749
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.258571428571429e-06,
+      "loss": 1.0402,
+      "step": 21750
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.258302857142857e-06,
+      "loss": 0.7064,
+      "step": 21751
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.258034285714285e-06,
+      "loss": 1.3212,
+      "step": 21752
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.257765714285715e-06,
+      "loss": 1.1117,
+      "step": 21753
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.257497142857143e-06,
+      "loss": 1.0365,
+      "step": 21754
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.257228571428571e-06,
+      "loss": 1.1119,
+      "step": 21755
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.25696e-06,
+      "loss": 1.1588,
+      "step": 21756
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.25669142857143e-06,
+      "loss": 1.1118,
+      "step": 21757
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.256422857142858e-06,
+      "loss": 1.1374,
+      "step": 21758
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.256154285714286e-06,
+      "loss": 1.3786,
+      "step": 21759
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.255885714285716e-06,
+      "loss": 0.8599,
+      "step": 21760
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.255617142857142e-06,
+      "loss": 0.8314,
+      "step": 21761
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.255348571428572e-06,
+      "loss": 0.9569,
+      "step": 21762
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.25508e-06,
+      "loss": 1.5992,
+      "step": 21763
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.254811428571428e-06,
+      "loss": 1.0983,
+      "step": 21764
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.254542857142857e-06,
+      "loss": 0.6865,
+      "step": 21765
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.254274285714286e-06,
+      "loss": 0.9467,
+      "step": 21766
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.254005714285715e-06,
+      "loss": 1.252,
+      "step": 21767
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.253737142857143e-06,
+      "loss": 1.0136,
+      "step": 21768
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.253468571428571e-06,
+      "loss": 1.3121,
+      "step": 21769
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.253200000000001e-06,
+      "loss": 1.602,
+      "step": 21770
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.252931428571429e-06,
+      "loss": 0.9185,
+      "step": 21771
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.252662857142857e-06,
+      "loss": 1.389,
+      "step": 21772
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.252394285714287e-06,
+      "loss": 1.1695,
+      "step": 21773
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.252125714285714e-06,
+      "loss": 1.1354,
+      "step": 21774
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.251857142857143e-06,
+      "loss": 0.9577,
+      "step": 21775
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.251588571428572e-06,
+      "loss": 1.125,
+      "step": 21776
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.251320000000001e-06,
+      "loss": 0.8304,
+      "step": 21777
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.251051428571428e-06,
+      "loss": 0.7515,
+      "step": 21778
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.250782857142858e-06,
+      "loss": 0.9849,
+      "step": 21779
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.250514285714286e-06,
+      "loss": 1.2086,
+      "step": 21780
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.250245714285714e-06,
+      "loss": 1.2851,
+      "step": 21781
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.249977142857142e-06,
+      "loss": 1.217,
+      "step": 21782
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.249708571428572e-06,
+      "loss": 1.3285,
+      "step": 21783
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.24944e-06,
+      "loss": 1.0358,
+      "step": 21784
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.249171428571429e-06,
+      "loss": 1.1045,
+      "step": 21785
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.248902857142858e-06,
+      "loss": 1.2281,
+      "step": 21786
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.248634285714285e-06,
+      "loss": 0.875,
+      "step": 21787
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.248365714285715e-06,
+      "loss": 1.279,
+      "step": 21788
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.248097142857143e-06,
+      "loss": 1.3582,
+      "step": 21789
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.247828571428573e-06,
+      "loss": 1.1596,
+      "step": 21790
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.24756e-06,
+      "loss": 1.1786,
+      "step": 21791
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.24729142857143e-06,
+      "loss": 1.3843,
+      "step": 21792
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.247022857142857e-06,
+      "loss": 1.0074,
+      "step": 21793
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.246754285714286e-06,
+      "loss": 0.9472,
+      "step": 21794
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.246485714285714e-06,
+      "loss": 0.9345,
+      "step": 21795
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.246217142857144e-06,
+      "loss": 0.957,
+      "step": 21796
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.245948571428572e-06,
+      "loss": 1.1387,
+      "step": 21797
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.24568e-06,
+      "loss": 1.6247,
+      "step": 21798
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.24541142857143e-06,
+      "loss": 1.0445,
+      "step": 21799
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.245142857142858e-06,
+      "loss": 1.0166,
+      "step": 21800
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.244874285714286e-06,
+      "loss": 1.1011,
+      "step": 21801
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.244605714285714e-06,
+      "loss": 1.2925,
+      "step": 21802
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.244337142857144e-06,
+      "loss": 1.1307,
+      "step": 21803
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.24406857142857e-06,
+      "loss": 1.1382,
+      "step": 21804
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.2438e-06,
+      "loss": 1.573,
+      "step": 21805
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.243531428571429e-06,
+      "loss": 1.3976,
+      "step": 21806
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.243262857142857e-06,
+      "loss": 0.9532,
+      "step": 21807
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.242994285714285e-06,
+      "loss": 0.7532,
+      "step": 21808
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.242725714285715e-06,
+      "loss": 1.1735,
+      "step": 21809
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.242457142857143e-06,
+      "loss": 1.0891,
+      "step": 21810
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.242188571428571e-06,
+      "loss": 1.4094,
+      "step": 21811
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.241920000000001e-06,
+      "loss": 1.919,
+      "step": 21812
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.24165142857143e-06,
+      "loss": 1.0377,
+      "step": 21813
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.241382857142858e-06,
+      "loss": 1.6579,
+      "step": 21814
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.241114285714286e-06,
+      "loss": 1.1821,
+      "step": 21815
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.240845714285716e-06,
+      "loss": 1.2627,
+      "step": 21816
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.240577142857142e-06,
+      "loss": 1.2327,
+      "step": 21817
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.240308571428572e-06,
+      "loss": 1.2678,
+      "step": 21818
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.24004e-06,
+      "loss": 1.3283,
+      "step": 21819
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.239771428571428e-06,
+      "loss": 1.1299,
+      "step": 21820
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.239502857142856e-06,
+      "loss": 1.2325,
+      "step": 21821
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.239234285714286e-06,
+      "loss": 1.1981,
+      "step": 21822
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.238965714285715e-06,
+      "loss": 1.2532,
+      "step": 21823
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.238697142857143e-06,
+      "loss": 1.0878,
+      "step": 21824
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.238428571428573e-06,
+      "loss": 1.3405,
+      "step": 21825
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.23816e-06,
+      "loss": 1.0826,
+      "step": 21826
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.237891428571429e-06,
+      "loss": 0.8781,
+      "step": 21827
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.237622857142857e-06,
+      "loss": 1.0406,
+      "step": 21828
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.237354285714287e-06,
+      "loss": 1.3216,
+      "step": 21829
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.237085714285713e-06,
+      "loss": 1.1056,
+      "step": 21830
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.236817142857143e-06,
+      "loss": 0.7662,
+      "step": 21831
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.236548571428572e-06,
+      "loss": 1.1617,
+      "step": 21832
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.23628e-06,
+      "loss": 1.0589,
+      "step": 21833
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.236011428571428e-06,
+      "loss": 0.9651,
+      "step": 21834
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.235742857142858e-06,
+      "loss": 1.3663,
+      "step": 21835
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.235474285714286e-06,
+      "loss": 1.5419,
+      "step": 21836
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.235205714285714e-06,
+      "loss": 0.7186,
+      "step": 21837
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.234937142857144e-06,
+      "loss": 1.0167,
+      "step": 21838
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.234668571428572e-06,
+      "loss": 1.0469,
+      "step": 21839
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.2344e-06,
+      "loss": 1.307,
+      "step": 21840
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.234131428571428e-06,
+      "loss": 1.2414,
+      "step": 21841
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.233862857142858e-06,
+      "loss": 1.4353,
+      "step": 21842
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.233594285714285e-06,
+      "loss": 0.9694,
+      "step": 21843
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.233325714285715e-06,
+      "loss": 1.2218,
+      "step": 21844
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.233057142857143e-06,
+      "loss": 1.306,
+      "step": 21845
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.232788571428573e-06,
+      "loss": 1.2914,
+      "step": 21846
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.23252e-06,
+      "loss": 1.6081,
+      "step": 21847
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.232251428571429e-06,
+      "loss": 0.8775,
+      "step": 21848
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.231982857142857e-06,
+      "loss": 0.9977,
+      "step": 21849
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.231714285714285e-06,
+      "loss": 1.1806,
+      "step": 21850
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.231445714285715e-06,
+      "loss": 1.2752,
+      "step": 21851
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.231177142857144e-06,
+      "loss": 0.9202,
+      "step": 21852
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.230908571428572e-06,
+      "loss": 1.2933,
+      "step": 21853
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.23064e-06,
+      "loss": 1.2513,
+      "step": 21854
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.23037142857143e-06,
+      "loss": 1.3449,
+      "step": 21855
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.230102857142856e-06,
+      "loss": 1.1868,
+      "step": 21856
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.229834285714286e-06,
+      "loss": 1.0035,
+      "step": 21857
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.229565714285714e-06,
+      "loss": 1.6537,
+      "step": 21858
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.229297142857144e-06,
+      "loss": 1.1381,
+      "step": 21859
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.22902857142857e-06,
+      "loss": 1.4618,
+      "step": 21860
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.22876e-06,
+      "loss": 1.3967,
+      "step": 21861
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.228491428571429e-06,
+      "loss": 1.6853,
+      "step": 21862
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.228222857142857e-06,
+      "loss": 1.5166,
+      "step": 21863
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.227954285714287e-06,
+      "loss": 1.3719,
+      "step": 21864
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.227685714285715e-06,
+      "loss": 1.0303,
+      "step": 21865
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.227417142857143e-06,
+      "loss": 0.9508,
+      "step": 21866
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.227148571428571e-06,
+      "loss": 1.241,
+      "step": 21867
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.226880000000001e-06,
+      "loss": 1.0606,
+      "step": 21868
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.226611428571428e-06,
+      "loss": 0.9859,
+      "step": 21869
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.226342857142857e-06,
+      "loss": 1.061,
+      "step": 21870
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.226074285714286e-06,
+      "loss": 1.3932,
+      "step": 21871
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.225805714285715e-06,
+      "loss": 1.4806,
+      "step": 21872
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.225537142857142e-06,
+      "loss": 1.0541,
+      "step": 21873
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.225268571428572e-06,
+      "loss": 1.0897,
+      "step": 21874
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.225e-06,
+      "loss": 1.0617,
+      "step": 21875
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.224731428571428e-06,
+      "loss": 0.7222,
+      "step": 21876
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.224462857142858e-06,
+      "loss": 1.3419,
+      "step": 21877
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.224194285714286e-06,
+      "loss": 1.2124,
+      "step": 21878
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.223925714285714e-06,
+      "loss": 1.6214,
+      "step": 21879
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.223657142857143e-06,
+      "loss": 1.3301,
+      "step": 21880
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.223388571428572e-06,
+      "loss": 1.1602,
+      "step": 21881
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.22312e-06,
+      "loss": 0.9833,
+      "step": 21882
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.222851428571429e-06,
+      "loss": 0.8203,
+      "step": 21883
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.222582857142857e-06,
+      "loss": 1.3697,
+      "step": 21884
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.222314285714287e-06,
+      "loss": 1.4878,
+      "step": 21885
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.222045714285713e-06,
+      "loss": 1.3189,
+      "step": 21886
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.221777142857143e-06,
+      "loss": 0.9506,
+      "step": 21887
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.221508571428571e-06,
+      "loss": 1.2041,
+      "step": 21888
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.22124e-06,
+      "loss": 0.9161,
+      "step": 21889
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.22097142857143e-06,
+      "loss": 0.9348,
+      "step": 21890
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.220702857142858e-06,
+      "loss": 1.096,
+      "step": 21891
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.220434285714286e-06,
+      "loss": 1.2428,
+      "step": 21892
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.220165714285714e-06,
+      "loss": 1.0747,
+      "step": 21893
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.219897142857144e-06,
+      "loss": 1.388,
+      "step": 21894
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.219628571428572e-06,
+      "loss": 1.5557,
+      "step": 21895
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.21936e-06,
+      "loss": 0.8682,
+      "step": 21896
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.219091428571428e-06,
+      "loss": 1.524,
+      "step": 21897
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.218822857142858e-06,
+      "loss": 1.1917,
+      "step": 21898
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.218554285714285e-06,
+      "loss": 1.39,
+      "step": 21899
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.218285714285715e-06,
+      "loss": 1.0169,
+      "step": 21900
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.218017142857143e-06,
+      "loss": 1.1778,
+      "step": 21901
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.217748571428571e-06,
+      "loss": 0.9396,
+      "step": 21902
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.21748e-06,
+      "loss": 1.032,
+      "step": 21903
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.217211428571429e-06,
+      "loss": 1.2251,
+      "step": 21904
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.216942857142859e-06,
+      "loss": 1.4146,
+      "step": 21905
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.216674285714285e-06,
+      "loss": 1.0479,
+      "step": 21906
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.216405714285715e-06,
+      "loss": 0.9626,
+      "step": 21907
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.216137142857143e-06,
+      "loss": 1.0282,
+      "step": 21908
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.215868571428572e-06,
+      "loss": 1.1291,
+      "step": 21909
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.2156e-06,
+      "loss": 0.7953,
+      "step": 21910
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.21533142857143e-06,
+      "loss": 0.9359,
+      "step": 21911
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.215062857142856e-06,
+      "loss": 0.874,
+      "step": 21912
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.214794285714286e-06,
+      "loss": 1.5797,
+      "step": 21913
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.214525714285714e-06,
+      "loss": 1.2861,
+      "step": 21914
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.214257142857142e-06,
+      "loss": 1.168,
+      "step": 21915
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.213988571428572e-06,
+      "loss": 0.9835,
+      "step": 21916
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.21372e-06,
+      "loss": 1.4923,
+      "step": 21917
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.21345142857143e-06,
+      "loss": 1.7194,
+      "step": 21918
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.213182857142857e-06,
+      "loss": 0.9693,
+      "step": 21919
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.212914285714287e-06,
+      "loss": 0.7502,
+      "step": 21920
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.212645714285715e-06,
+      "loss": 0.8635,
+      "step": 21921
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.212377142857143e-06,
+      "loss": 1.391,
+      "step": 21922
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.212108571428571e-06,
+      "loss": 1.2626,
+      "step": 21923
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.211840000000001e-06,
+      "loss": 1.3603,
+      "step": 21924
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.211571428571427e-06,
+      "loss": 1.2605,
+      "step": 21925
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.211302857142857e-06,
+      "loss": 1.4005,
+      "step": 21926
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.211034285714286e-06,
+      "loss": 1.2234,
+      "step": 21927
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.210765714285715e-06,
+      "loss": 1.0848,
+      "step": 21928
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.210497142857144e-06,
+      "loss": 1.5494,
+      "step": 21929
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.210228571428572e-06,
+      "loss": 0.9706,
+      "step": 21930
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.209960000000002e-06,
+      "loss": 1.2498,
+      "step": 21931
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.209691428571428e-06,
+      "loss": 1.0233,
+      "step": 21932
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.209422857142858e-06,
+      "loss": 1.069,
+      "step": 21933
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.209154285714286e-06,
+      "loss": 0.9805,
+      "step": 21934
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.208885714285714e-06,
+      "loss": 1.2238,
+      "step": 21935
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.208617142857142e-06,
+      "loss": 1.4762,
+      "step": 21936
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.208348571428572e-06,
+      "loss": 1.1793,
+      "step": 21937
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.208079999999999e-06,
+      "loss": 1.1158,
+      "step": 21938
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.207811428571429e-06,
+      "loss": 0.9851,
+      "step": 21939
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.207542857142857e-06,
+      "loss": 1.4643,
+      "step": 21940
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.207274285714287e-06,
+      "loss": 1.079,
+      "step": 21941
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.207005714285715e-06,
+      "loss": 1.2335,
+      "step": 21942
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.206737142857143e-06,
+      "loss": 1.0331,
+      "step": 21943
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.206468571428573e-06,
+      "loss": 1.329,
+      "step": 21944
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.2062e-06,
+      "loss": 1.0244,
+      "step": 21945
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.20593142857143e-06,
+      "loss": 1.2927,
+      "step": 21946
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.205662857142858e-06,
+      "loss": 1.1797,
+      "step": 21947
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.205394285714286e-06,
+      "loss": 0.8352,
+      "step": 21948
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.205125714285714e-06,
+      "loss": 1.2595,
+      "step": 21949
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.204857142857144e-06,
+      "loss": 1.2615,
+      "step": 21950
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.204588571428572e-06,
+      "loss": 1.0733,
+      "step": 21951
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.20432e-06,
+      "loss": 1.1206,
+      "step": 21952
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.204051428571428e-06,
+      "loss": 1.0541,
+      "step": 21953
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.203782857142858e-06,
+      "loss": 1.0846,
+      "step": 21954
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.203514285714286e-06,
+      "loss": 1.4951,
+      "step": 21955
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.203245714285714e-06,
+      "loss": 1.3653,
+      "step": 21956
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.202977142857144e-06,
+      "loss": 0.7679,
+      "step": 21957
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.20270857142857e-06,
+      "loss": 1.227,
+      "step": 21958
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.20244e-06,
+      "loss": 1.2516,
+      "step": 21959
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.202171428571429e-06,
+      "loss": 0.9854,
+      "step": 21960
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.201902857142857e-06,
+      "loss": 1.1626,
+      "step": 21961
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.201634285714285e-06,
+      "loss": 1.1947,
+      "step": 21962
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.201365714285715e-06,
+      "loss": 0.8408,
+      "step": 21963
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.201097142857143e-06,
+      "loss": 1.7889,
+      "step": 21964
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.200828571428571e-06,
+      "loss": 1.3027,
+      "step": 21965
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.20056e-06,
+      "loss": 1.4924,
+      "step": 21966
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.20029142857143e-06,
+      "loss": 0.5679,
+      "step": 21967
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.200022857142858e-06,
+      "loss": 1.1508,
+      "step": 21968
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.199754285714286e-06,
+      "loss": 0.9779,
+      "step": 21969
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.199485714285716e-06,
+      "loss": 1.4009,
+      "step": 21970
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.199217142857142e-06,
+      "loss": 0.7154,
+      "step": 21971
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.198948571428572e-06,
+      "loss": 1.0114,
+      "step": 21972
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.19868e-06,
+      "loss": 1.3687,
+      "step": 21973
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.19841142857143e-06,
+      "loss": 1.0408,
+      "step": 21974
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.198142857142857e-06,
+      "loss": 1.4023,
+      "step": 21975
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.197874285714286e-06,
+      "loss": 1.0251,
+      "step": 21976
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.197605714285715e-06,
+      "loss": 1.3775,
+      "step": 21977
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.197337142857143e-06,
+      "loss": 1.3623,
+      "step": 21978
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.197068571428571e-06,
+      "loss": 1.3687,
+      "step": 21979
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.196800000000001e-06,
+      "loss": 0.9066,
+      "step": 21980
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.196531428571429e-06,
+      "loss": 1.0099,
+      "step": 21981
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.196262857142857e-06,
+      "loss": 1.0008,
+      "step": 21982
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.195994285714287e-06,
+      "loss": 1.4323,
+      "step": 21983
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.195725714285714e-06,
+      "loss": 0.6846,
+      "step": 21984
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.195457142857143e-06,
+      "loss": 0.7659,
+      "step": 21985
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.195188571428572e-06,
+      "loss": 1.3242,
+      "step": 21986
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.194920000000001e-06,
+      "loss": 0.8054,
+      "step": 21987
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.194651428571428e-06,
+      "loss": 1.1675,
+      "step": 21988
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.194382857142858e-06,
+      "loss": 1.2389,
+      "step": 21989
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.194114285714286e-06,
+      "loss": 1.3459,
+      "step": 21990
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.193845714285714e-06,
+      "loss": 1.2603,
+      "step": 21991
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.193577142857142e-06,
+      "loss": 0.779,
+      "step": 21992
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.193308571428572e-06,
+      "loss": 1.0136,
+      "step": 21993
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.19304e-06,
+      "loss": 1.2493,
+      "step": 21994
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.192771428571429e-06,
+      "loss": 1.001,
+      "step": 21995
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.192502857142858e-06,
+      "loss": 1.4537,
+      "step": 21996
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.192234285714287e-06,
+      "loss": 1.2969,
+      "step": 21997
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.191965714285715e-06,
+      "loss": 1.0784,
+      "step": 21998
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.191697142857143e-06,
+      "loss": 1.5909,
+      "step": 21999
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.191428571428573e-06,
+      "loss": 1.2556,
+      "step": 22000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.19116e-06,
+      "loss": 1.1537,
+      "step": 22001
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.19089142857143e-06,
+      "loss": 1.0585,
+      "step": 22002
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.190622857142857e-06,
+      "loss": 1.4321,
+      "step": 22003
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.190354285714286e-06,
+      "loss": 1.011,
+      "step": 22004
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.190085714285714e-06,
+      "loss": 1.2907,
+      "step": 22005
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.189817142857144e-06,
+      "loss": 1.1359,
+      "step": 22006
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.189548571428572e-06,
+      "loss": 1.1632,
+      "step": 22007
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.18928e-06,
+      "loss": 1.0755,
+      "step": 22008
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.18901142857143e-06,
+      "loss": 1.5817,
+      "step": 22009
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.188742857142858e-06,
+      "loss": 0.9579,
+      "step": 22010
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.188474285714286e-06,
+      "loss": 1.172,
+      "step": 22011
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.188205714285714e-06,
+      "loss": 0.8399,
+      "step": 22012
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.187937142857144e-06,
+      "loss": 0.6682,
+      "step": 22013
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.18766857142857e-06,
+      "loss": 1.1563,
+      "step": 22014
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.1874e-06,
+      "loss": 1.1949,
+      "step": 22015
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.187131428571429e-06,
+      "loss": 1.4363,
+      "step": 22016
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.186862857142857e-06,
+      "loss": 1.0662,
+      "step": 22017
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.186594285714285e-06,
+      "loss": 1.0404,
+      "step": 22018
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.186325714285715e-06,
+      "loss": 1.1712,
+      "step": 22019
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.186057142857143e-06,
+      "loss": 1.2096,
+      "step": 22020
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.185788571428571e-06,
+      "loss": 1.3709,
+      "step": 22021
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.185520000000001e-06,
+      "loss": 1.0977,
+      "step": 22022
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.18525142857143e-06,
+      "loss": 1.6178,
+      "step": 22023
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.184982857142858e-06,
+      "loss": 1.2469,
+      "step": 22024
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.184714285714286e-06,
+      "loss": 0.8598,
+      "step": 22025
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.184445714285716e-06,
+      "loss": 1.1702,
+      "step": 22026
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.184177142857142e-06,
+      "loss": 1.0639,
+      "step": 22027
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.183908571428572e-06,
+      "loss": 1.1068,
+      "step": 22028
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.18364e-06,
+      "loss": 1.2174,
+      "step": 22029
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.183371428571428e-06,
+      "loss": 1.116,
+      "step": 22030
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.183102857142856e-06,
+      "loss": 1.3241,
+      "step": 22031
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.182834285714286e-06,
+      "loss": 0.9364,
+      "step": 22032
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.182565714285715e-06,
+      "loss": 1.4934,
+      "step": 22033
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.182297142857143e-06,
+      "loss": 1.0629,
+      "step": 22034
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.182028571428573e-06,
+      "loss": 0.9336,
+      "step": 22035
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.18176e-06,
+      "loss": 1.0769,
+      "step": 22036
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.181491428571429e-06,
+      "loss": 1.5214,
+      "step": 22037
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.181222857142857e-06,
+      "loss": 1.0602,
+      "step": 22038
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.180954285714287e-06,
+      "loss": 1.1571,
+      "step": 22039
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.180685714285713e-06,
+      "loss": 1.2057,
+      "step": 22040
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.180417142857143e-06,
+      "loss": 0.8788,
+      "step": 22041
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.180148571428572e-06,
+      "loss": 1.501,
+      "step": 22042
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.17988e-06,
+      "loss": 1.4151,
+      "step": 22043
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.179611428571428e-06,
+      "loss": 1.2433,
+      "step": 22044
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.179342857142858e-06,
+      "loss": 0.6,
+      "step": 22045
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.179074285714286e-06,
+      "loss": 1.0352,
+      "step": 22046
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.178805714285714e-06,
+      "loss": 0.9745,
+      "step": 22047
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.178537142857144e-06,
+      "loss": 1.5978,
+      "step": 22048
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.178268571428572e-06,
+      "loss": 0.9534,
+      "step": 22049
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.178e-06,
+      "loss": 1.1972,
+      "step": 22050
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.177731428571428e-06,
+      "loss": 1.1333,
+      "step": 22051
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.177462857142858e-06,
+      "loss": 0.8756,
+      "step": 22052
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.177194285714285e-06,
+      "loss": 1.1668,
+      "step": 22053
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.176925714285715e-06,
+      "loss": 1.0637,
+      "step": 22054
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.176657142857143e-06,
+      "loss": 1.2801,
+      "step": 22055
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.176388571428573e-06,
+      "loss": 1.4269,
+      "step": 22056
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.17612e-06,
+      "loss": 1.1095,
+      "step": 22057
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.175851428571429e-06,
+      "loss": 1.0575,
+      "step": 22058
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.175582857142857e-06,
+      "loss": 1.0607,
+      "step": 22059
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.175314285714285e-06,
+      "loss": 1.4936,
+      "step": 22060
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.175045714285715e-06,
+      "loss": 1.4931,
+      "step": 22061
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.174777142857144e-06,
+      "loss": 1.0385,
+      "step": 22062
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.174508571428572e-06,
+      "loss": 1.0657,
+      "step": 22063
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.17424e-06,
+      "loss": 1.2315,
+      "step": 22064
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.17397142857143e-06,
+      "loss": 0.936,
+      "step": 22065
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.173702857142856e-06,
+      "loss": 1.33,
+      "step": 22066
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.173434285714286e-06,
+      "loss": 1.3439,
+      "step": 22067
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.173165714285714e-06,
+      "loss": 1.415,
+      "step": 22068
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.172897142857144e-06,
+      "loss": 0.7262,
+      "step": 22069
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.17262857142857e-06,
+      "loss": 1.1768,
+      "step": 22070
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.17236e-06,
+      "loss": 1.2425,
+      "step": 22071
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.172091428571429e-06,
+      "loss": 1.2517,
+      "step": 22072
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.171822857142857e-06,
+      "loss": 1.5321,
+      "step": 22073
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.171554285714287e-06,
+      "loss": 0.8379,
+      "step": 22074
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.171285714285715e-06,
+      "loss": 1.1638,
+      "step": 22075
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.171017142857143e-06,
+      "loss": 1.4148,
+      "step": 22076
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.170748571428571e-06,
+      "loss": 0.9976,
+      "step": 22077
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.170480000000001e-06,
+      "loss": 1.3194,
+      "step": 22078
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.17021142857143e-06,
+      "loss": 1.4358,
+      "step": 22079
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.169942857142857e-06,
+      "loss": 1.0575,
+      "step": 22080
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.169674285714286e-06,
+      "loss": 0.9614,
+      "step": 22081
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.169405714285715e-06,
+      "loss": 1.0477,
+      "step": 22082
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.169137142857142e-06,
+      "loss": 1.0548,
+      "step": 22083
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.168868571428572e-06,
+      "loss": 0.8689,
+      "step": 22084
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.1686e-06,
+      "loss": 1.3805,
+      "step": 22085
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.168331428571428e-06,
+      "loss": 1.4584,
+      "step": 22086
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.168062857142858e-06,
+      "loss": 1.35,
+      "step": 22087
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.167794285714286e-06,
+      "loss": 1.2596,
+      "step": 22088
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.167525714285714e-06,
+      "loss": 0.9549,
+      "step": 22089
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.167257142857143e-06,
+      "loss": 1.0338,
+      "step": 22090
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.166988571428572e-06,
+      "loss": 1.1068,
+      "step": 22091
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.16672e-06,
+      "loss": 1.06,
+      "step": 22092
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.166451428571429e-06,
+      "loss": 1.3746,
+      "step": 22093
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.166182857142857e-06,
+      "loss": 1.2773,
+      "step": 22094
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.165914285714287e-06,
+      "loss": 1.5666,
+      "step": 22095
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.165645714285713e-06,
+      "loss": 0.9616,
+      "step": 22096
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.165377142857143e-06,
+      "loss": 1.1998,
+      "step": 22097
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.165108571428571e-06,
+      "loss": 0.8619,
+      "step": 22098
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.16484e-06,
+      "loss": 1.0094,
+      "step": 22099
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.16457142857143e-06,
+      "loss": 0.7353,
+      "step": 22100
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.164302857142858e-06,
+      "loss": 1.1605,
+      "step": 22101
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.164034285714287e-06,
+      "loss": 0.9423,
+      "step": 22102
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.163765714285714e-06,
+      "loss": 1.5992,
+      "step": 22103
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.163497142857144e-06,
+      "loss": 0.9591,
+      "step": 22104
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.163228571428572e-06,
+      "loss": 1.0336,
+      "step": 22105
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.16296e-06,
+      "loss": 0.9815,
+      "step": 22106
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.162691428571428e-06,
+      "loss": 0.9781,
+      "step": 22107
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.162422857142858e-06,
+      "loss": 1.2835,
+      "step": 22108
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.162154285714285e-06,
+      "loss": 0.9112,
+      "step": 22109
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.161885714285715e-06,
+      "loss": 1.1474,
+      "step": 22110
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.161617142857143e-06,
+      "loss": 1.1993,
+      "step": 22111
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.161348571428571e-06,
+      "loss": 1.0981,
+      "step": 22112
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.16108e-06,
+      "loss": 1.4101,
+      "step": 22113
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.160811428571429e-06,
+      "loss": 0.8523,
+      "step": 22114
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.160542857142859e-06,
+      "loss": 1.1785,
+      "step": 22115
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.160274285714285e-06,
+      "loss": 1.1952,
+      "step": 22116
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.160005714285715e-06,
+      "loss": 1.1562,
+      "step": 22117
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.159737142857143e-06,
+      "loss": 1.0068,
+      "step": 22118
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.159468571428572e-06,
+      "loss": 1.4446,
+      "step": 22119
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.1592e-06,
+      "loss": 0.996,
+      "step": 22120
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.15893142857143e-06,
+      "loss": 1.4338,
+      "step": 22121
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.158662857142856e-06,
+      "loss": 0.7865,
+      "step": 22122
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.158394285714286e-06,
+      "loss": 0.8893,
+      "step": 22123
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.158125714285714e-06,
+      "loss": 1.5278,
+      "step": 22124
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.157857142857144e-06,
+      "loss": 0.9209,
+      "step": 22125
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.157588571428572e-06,
+      "loss": 1.1489,
+      "step": 22126
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.15732e-06,
+      "loss": 1.492,
+      "step": 22127
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.15705142857143e-06,
+      "loss": 1.6301,
+      "step": 22128
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.156782857142857e-06,
+      "loss": 0.9529,
+      "step": 22129
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.156514285714287e-06,
+      "loss": 1.2148,
+      "step": 22130
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.156245714285715e-06,
+      "loss": 1.2163,
+      "step": 22131
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.155977142857143e-06,
+      "loss": 0.8198,
+      "step": 22132
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.155708571428571e-06,
+      "loss": 1.2123,
+      "step": 22133
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.155440000000001e-06,
+      "loss": 1.3461,
+      "step": 22134
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.155171428571427e-06,
+      "loss": 1.0473,
+      "step": 22135
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.154902857142857e-06,
+      "loss": 1.1819,
+      "step": 22136
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.154634285714286e-06,
+      "loss": 1.0533,
+      "step": 22137
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.154365714285715e-06,
+      "loss": 1.3215,
+      "step": 22138
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.154097142857144e-06,
+      "loss": 0.869,
+      "step": 22139
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.153828571428572e-06,
+      "loss": 1.2761,
+      "step": 22140
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.153560000000002e-06,
+      "loss": 1.3654,
+      "step": 22141
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.153291428571428e-06,
+      "loss": 1.4473,
+      "step": 22142
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.153022857142858e-06,
+      "loss": 1.1497,
+      "step": 22143
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.152754285714286e-06,
+      "loss": 0.8918,
+      "step": 22144
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.152485714285714e-06,
+      "loss": 1.0076,
+      "step": 22145
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.152217142857142e-06,
+      "loss": 1.2046,
+      "step": 22146
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.151948571428572e-06,
+      "loss": 1.6375,
+      "step": 22147
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.15168e-06,
+      "loss": 1.2644,
+      "step": 22148
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.151411428571429e-06,
+      "loss": 1.3743,
+      "step": 22149
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.151142857142857e-06,
+      "loss": 1.5807,
+      "step": 22150
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.150874285714287e-06,
+      "loss": 1.5018,
+      "step": 22151
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.150605714285715e-06,
+      "loss": 1.2291,
+      "step": 22152
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.150337142857143e-06,
+      "loss": 0.7501,
+      "step": 22153
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.150068571428573e-06,
+      "loss": 0.9503,
+      "step": 22154
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.1498e-06,
+      "loss": 1.216,
+      "step": 22155
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14953142857143e-06,
+      "loss": 1.16,
+      "step": 22156
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.149262857142858e-06,
+      "loss": 0.9883,
+      "step": 22157
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.148994285714286e-06,
+      "loss": 1.4443,
+      "step": 22158
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.148725714285714e-06,
+      "loss": 1.2949,
+      "step": 22159
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.148457142857144e-06,
+      "loss": 1.0143,
+      "step": 22160
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.148188571428572e-06,
+      "loss": 1.1013,
+      "step": 22161
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14792e-06,
+      "loss": 0.9856,
+      "step": 22162
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.147651428571428e-06,
+      "loss": 1.8674,
+      "step": 22163
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.147382857142858e-06,
+      "loss": 1.4174,
+      "step": 22164
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.147114285714286e-06,
+      "loss": 1.5245,
+      "step": 22165
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.146845714285714e-06,
+      "loss": 1.1636,
+      "step": 22166
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.146577142857144e-06,
+      "loss": 0.9998,
+      "step": 22167
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14630857142857e-06,
+      "loss": 1.2651,
+      "step": 22168
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14604e-06,
+      "loss": 0.8106,
+      "step": 22169
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.145771428571429e-06,
+      "loss": 1.3419,
+      "step": 22170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.145502857142859e-06,
+      "loss": 1.1765,
+      "step": 22171
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.145234285714285e-06,
+      "loss": 0.9317,
+      "step": 22172
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.144965714285715e-06,
+      "loss": 1.3509,
+      "step": 22173
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.144697142857143e-06,
+      "loss": 1.0366,
+      "step": 22174
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.144428571428571e-06,
+      "loss": 0.9545,
+      "step": 22175
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14416e-06,
+      "loss": 1.0315,
+      "step": 22176
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14389142857143e-06,
+      "loss": 0.9739,
+      "step": 22177
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.143622857142858e-06,
+      "loss": 0.8796,
+      "step": 22178
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.143354285714286e-06,
+      "loss": 1.1679,
+      "step": 22179
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.143085714285716e-06,
+      "loss": 1.0835,
+      "step": 22180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.142817142857142e-06,
+      "loss": 1.0562,
+      "step": 22181
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.142548571428572e-06,
+      "loss": 0.7777,
+      "step": 22182
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14228e-06,
+      "loss": 1.3011,
+      "step": 22183
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.14201142857143e-06,
+      "loss": 1.2587,
+      "step": 22184
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.141742857142857e-06,
+      "loss": 1.1628,
+      "step": 22185
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.141474285714286e-06,
+      "loss": 1.0978,
+      "step": 22186
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.141205714285715e-06,
+      "loss": 1.0853,
+      "step": 22187
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.140937142857143e-06,
+      "loss": 1.0751,
+      "step": 22188
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.140668571428571e-06,
+      "loss": 1.2244,
+      "step": 22189
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.140400000000001e-06,
+      "loss": 1.1093,
+      "step": 22190
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.140131428571429e-06,
+      "loss": 1.4007,
+      "step": 22191
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.139862857142857e-06,
+      "loss": 1.1129,
+      "step": 22192
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.139594285714287e-06,
+      "loss": 0.9438,
+      "step": 22193
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.139325714285714e-06,
+      "loss": 1.1533,
+      "step": 22194
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.139057142857143e-06,
+      "loss": 1.4966,
+      "step": 22195
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.138788571428572e-06,
+      "loss": 1.0502,
+      "step": 22196
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.138520000000001e-06,
+      "loss": 1.17,
+      "step": 22197
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.138251428571428e-06,
+      "loss": 1.3141,
+      "step": 22198
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.137982857142858e-06,
+      "loss": 1.4272,
+      "step": 22199
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.137714285714286e-06,
+      "loss": 0.8849,
+      "step": 22200
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.137445714285714e-06,
+      "loss": 0.7124,
+      "step": 22201
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.137177142857142e-06,
+      "loss": 1.454,
+      "step": 22202
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.136908571428572e-06,
+      "loss": 0.9748,
+      "step": 22203
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.13664e-06,
+      "loss": 0.8201,
+      "step": 22204
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.136371428571429e-06,
+      "loss": 0.7587,
+      "step": 22205
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.136102857142858e-06,
+      "loss": 1.0441,
+      "step": 22206
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.135834285714287e-06,
+      "loss": 1.0449,
+      "step": 22207
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.135565714285715e-06,
+      "loss": 1.2929,
+      "step": 22208
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.135297142857143e-06,
+      "loss": 1.1157,
+      "step": 22209
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.135028571428573e-06,
+      "loss": 0.8915,
+      "step": 22210
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.13476e-06,
+      "loss": 0.8753,
+      "step": 22211
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.13449142857143e-06,
+      "loss": 1.0368,
+      "step": 22212
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.134222857142857e-06,
+      "loss": 1.0716,
+      "step": 22213
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.133954285714286e-06,
+      "loss": 1.4645,
+      "step": 22214
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.133685714285714e-06,
+      "loss": 1.33,
+      "step": 22215
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.133417142857144e-06,
+      "loss": 1.3253,
+      "step": 22216
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.133148571428572e-06,
+      "loss": 0.8697,
+      "step": 22217
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.13288e-06,
+      "loss": 0.9382,
+      "step": 22218
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.13261142857143e-06,
+      "loss": 1.1684,
+      "step": 22219
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.132342857142858e-06,
+      "loss": 1.0663,
+      "step": 22220
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.132074285714286e-06,
+      "loss": 0.7976,
+      "step": 22221
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.131805714285714e-06,
+      "loss": 1.7375,
+      "step": 22222
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.131537142857144e-06,
+      "loss": 0.8794,
+      "step": 22223
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.13126857142857e-06,
+      "loss": 1.142,
+      "step": 22224
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.131e-06,
+      "loss": 1.5361,
+      "step": 22225
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.130731428571429e-06,
+      "loss": 1.083,
+      "step": 22226
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.130462857142857e-06,
+      "loss": 1.219,
+      "step": 22227
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.130194285714285e-06,
+      "loss": 0.9902,
+      "step": 22228
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.129925714285715e-06,
+      "loss": 1.1063,
+      "step": 22229
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.129657142857143e-06,
+      "loss": 1.3744,
+      "step": 22230
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.129388571428571e-06,
+      "loss": 1.233,
+      "step": 22231
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.129120000000001e-06,
+      "loss": 0.9433,
+      "step": 22232
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.12885142857143e-06,
+      "loss": 1.2345,
+      "step": 22233
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.128582857142858e-06,
+      "loss": 1.6339,
+      "step": 22234
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.128314285714286e-06,
+      "loss": 1.3028,
+      "step": 22235
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.128045714285716e-06,
+      "loss": 1.6619,
+      "step": 22236
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.127777142857142e-06,
+      "loss": 1.1458,
+      "step": 22237
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.127508571428572e-06,
+      "loss": 1.1792,
+      "step": 22238
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.12724e-06,
+      "loss": 1.3946,
+      "step": 22239
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.126971428571428e-06,
+      "loss": 0.8782,
+      "step": 22240
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.126702857142856e-06,
+      "loss": 1.5328,
+      "step": 22241
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.126434285714286e-06,
+      "loss": 1.2184,
+      "step": 22242
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.126165714285715e-06,
+      "loss": 1.4019,
+      "step": 22243
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.125897142857143e-06,
+      "loss": 1.2206,
+      "step": 22244
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.125628571428573e-06,
+      "loss": 1.034,
+      "step": 22245
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.12536e-06,
+      "loss": 1.1997,
+      "step": 22246
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.125091428571429e-06,
+      "loss": 1.0761,
+      "step": 22247
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.124822857142857e-06,
+      "loss": 1.5636,
+      "step": 22248
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.124554285714287e-06,
+      "loss": 1.469,
+      "step": 22249
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.124285714285713e-06,
+      "loss": 1.4357,
+      "step": 22250
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.124017142857143e-06,
+      "loss": 1.0607,
+      "step": 22251
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.123748571428572e-06,
+      "loss": 0.9481,
+      "step": 22252
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.123480000000001e-06,
+      "loss": 1.0051,
+      "step": 22253
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.123211428571428e-06,
+      "loss": 1.3923,
+      "step": 22254
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.122942857142858e-06,
+      "loss": 1.4477,
+      "step": 22255
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.122674285714286e-06,
+      "loss": 1.0565,
+      "step": 22256
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.122405714285714e-06,
+      "loss": 1.1627,
+      "step": 22257
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.122137142857144e-06,
+      "loss": 0.9684,
+      "step": 22258
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.121868571428572e-06,
+      "loss": 1.1454,
+      "step": 22259
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.1216e-06,
+      "loss": 0.7169,
+      "step": 22260
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.121331428571428e-06,
+      "loss": 1.1332,
+      "step": 22261
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.121062857142858e-06,
+      "loss": 1.538,
+      "step": 22262
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.120794285714285e-06,
+      "loss": 0.8511,
+      "step": 22263
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.120525714285715e-06,
+      "loss": 1.0969,
+      "step": 22264
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.120257142857143e-06,
+      "loss": 1.2849,
+      "step": 22265
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.119988571428573e-06,
+      "loss": 0.9925,
+      "step": 22266
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.11972e-06,
+      "loss": 1.1056,
+      "step": 22267
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.119451428571429e-06,
+      "loss": 1.2493,
+      "step": 22268
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.119182857142857e-06,
+      "loss": 1.58,
+      "step": 22269
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.118914285714285e-06,
+      "loss": 0.9868,
+      "step": 22270
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.118645714285715e-06,
+      "loss": 0.8364,
+      "step": 22271
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.118377142857143e-06,
+      "loss": 1.0858,
+      "step": 22272
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.118108571428572e-06,
+      "loss": 1.1681,
+      "step": 22273
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.11784e-06,
+      "loss": 1.2832,
+      "step": 22274
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.11757142857143e-06,
+      "loss": 1.2409,
+      "step": 22275
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.117302857142858e-06,
+      "loss": 0.9168,
+      "step": 22276
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.117034285714286e-06,
+      "loss": 1.0159,
+      "step": 22277
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.116765714285714e-06,
+      "loss": 0.8172,
+      "step": 22278
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.116497142857144e-06,
+      "loss": 0.9046,
+      "step": 22279
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.11622857142857e-06,
+      "loss": 1.1946,
+      "step": 22280
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.11596e-06,
+      "loss": 1.0892,
+      "step": 22281
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.115691428571429e-06,
+      "loss": 1.4172,
+      "step": 22282
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.115422857142857e-06,
+      "loss": 1.2883,
+      "step": 22283
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.115154285714287e-06,
+      "loss": 1.5763,
+      "step": 22284
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.114885714285715e-06,
+      "loss": 1.2782,
+      "step": 22285
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.114617142857143e-06,
+      "loss": 1.0974,
+      "step": 22286
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.114348571428571e-06,
+      "loss": 0.8937,
+      "step": 22287
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.114080000000001e-06,
+      "loss": 1.3285,
+      "step": 22288
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.11381142857143e-06,
+      "loss": 1.6619,
+      "step": 22289
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.113542857142857e-06,
+      "loss": 1.5005,
+      "step": 22290
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.113274285714286e-06,
+      "loss": 1.1325,
+      "step": 22291
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.113005714285715e-06,
+      "loss": 1.3234,
+      "step": 22292
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.112737142857142e-06,
+      "loss": 0.915,
+      "step": 22293
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.112468571428572e-06,
+      "loss": 1.0206,
+      "step": 22294
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.1122e-06,
+      "loss": 1.0087,
+      "step": 22295
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.111931428571428e-06,
+      "loss": 1.0045,
+      "step": 22296
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.111662857142858e-06,
+      "loss": 1.3899,
+      "step": 22297
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.111394285714286e-06,
+      "loss": 0.7404,
+      "step": 22298
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.111125714285716e-06,
+      "loss": 1.403,
+      "step": 22299
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.110857142857143e-06,
+      "loss": 1.1463,
+      "step": 22300
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.110588571428572e-06,
+      "loss": 1.3698,
+      "step": 22301
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.11032e-06,
+      "loss": 0.9702,
+      "step": 22302
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.110051428571429e-06,
+      "loss": 1.3532,
+      "step": 22303
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.109782857142857e-06,
+      "loss": 1.4562,
+      "step": 22304
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.109514285714287e-06,
+      "loss": 1.3598,
+      "step": 22305
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.109245714285713e-06,
+      "loss": 1.246,
+      "step": 22306
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.108977142857143e-06,
+      "loss": 1.2897,
+      "step": 22307
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.108708571428571e-06,
+      "loss": 0.8105,
+      "step": 22308
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.10844e-06,
+      "loss": 1.1532,
+      "step": 22309
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.10817142857143e-06,
+      "loss": 1.1567,
+      "step": 22310
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.107902857142858e-06,
+      "loss": 1.0681,
+      "step": 22311
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.107634285714287e-06,
+      "loss": 0.8793,
+      "step": 22312
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.107365714285714e-06,
+      "loss": 1.0534,
+      "step": 22313
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.107097142857144e-06,
+      "loss": 1.2963,
+      "step": 22314
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.106828571428572e-06,
+      "loss": 1.5159,
+      "step": 22315
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.10656e-06,
+      "loss": 1.2818,
+      "step": 22316
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.106291428571428e-06,
+      "loss": 1.3173,
+      "step": 22317
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.106022857142858e-06,
+      "loss": 1.0981,
+      "step": 22318
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.105754285714285e-06,
+      "loss": 1.0904,
+      "step": 22319
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.105485714285715e-06,
+      "loss": 1.1984,
+      "step": 22320
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.105217142857143e-06,
+      "loss": 1.0153,
+      "step": 22321
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.104948571428573e-06,
+      "loss": 1.4999,
+      "step": 22322
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.10468e-06,
+      "loss": 1.4249,
+      "step": 22323
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.104411428571429e-06,
+      "loss": 1.1099,
+      "step": 22324
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.104142857142859e-06,
+      "loss": 1.4281,
+      "step": 22325
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.103874285714285e-06,
+      "loss": 1.3465,
+      "step": 22326
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.103605714285715e-06,
+      "loss": 1.7271,
+      "step": 22327
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.103337142857143e-06,
+      "loss": 1.171,
+      "step": 22328
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.103068571428572e-06,
+      "loss": 1.2749,
+      "step": 22329
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.1028e-06,
+      "loss": 1.1856,
+      "step": 22330
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.10253142857143e-06,
+      "loss": 1.3021,
+      "step": 22331
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.102262857142856e-06,
+      "loss": 1.4233,
+      "step": 22332
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.101994285714286e-06,
+      "loss": 1.2341,
+      "step": 22333
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.101725714285714e-06,
+      "loss": 1.591,
+      "step": 22334
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.101457142857144e-06,
+      "loss": 1.6139,
+      "step": 22335
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.101188571428572e-06,
+      "loss": 1.1681,
+      "step": 22336
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.10092e-06,
+      "loss": 1.0184,
+      "step": 22337
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.10065142857143e-06,
+      "loss": 1.1869,
+      "step": 22338
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.100382857142857e-06,
+      "loss": 1.026,
+      "step": 22339
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.100114285714287e-06,
+      "loss": 1.3008,
+      "step": 22340
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.099845714285715e-06,
+      "loss": 0.8409,
+      "step": 22341
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.099577142857143e-06,
+      "loss": 1.1191,
+      "step": 22342
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.099308571428571e-06,
+      "loss": 1.4121,
+      "step": 22343
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.099040000000001e-06,
+      "loss": 1.3953,
+      "step": 22344
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.09877142857143e-06,
+      "loss": 0.7111,
+      "step": 22345
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.098502857142857e-06,
+      "loss": 0.933,
+      "step": 22346
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.098234285714286e-06,
+      "loss": 0.8502,
+      "step": 22347
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.097965714285715e-06,
+      "loss": 1.1026,
+      "step": 22348
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.097697142857144e-06,
+      "loss": 1.1047,
+      "step": 22349
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.097428571428572e-06,
+      "loss": 0.8969,
+      "step": 22350
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.097160000000002e-06,
+      "loss": 1.3808,
+      "step": 22351
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.096891428571428e-06,
+      "loss": 0.8797,
+      "step": 22352
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.096622857142858e-06,
+      "loss": 1.3343,
+      "step": 22353
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.096354285714286e-06,
+      "loss": 1.5095,
+      "step": 22354
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.096085714285714e-06,
+      "loss": 0.7717,
+      "step": 22355
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.095817142857142e-06,
+      "loss": 0.9387,
+      "step": 22356
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.095548571428572e-06,
+      "loss": 0.9037,
+      "step": 22357
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.09528e-06,
+      "loss": 1.2702,
+      "step": 22358
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.095011428571429e-06,
+      "loss": 0.9447,
+      "step": 22359
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.094742857142857e-06,
+      "loss": 1.1342,
+      "step": 22360
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.094474285714287e-06,
+      "loss": 0.803,
+      "step": 22361
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.094205714285715e-06,
+      "loss": 1.2198,
+      "step": 22362
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.093937142857143e-06,
+      "loss": 1.3312,
+      "step": 22363
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.093668571428573e-06,
+      "loss": 1.0476,
+      "step": 22364
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.0934e-06,
+      "loss": 0.945,
+      "step": 22365
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.09313142857143e-06,
+      "loss": 1.3096,
+      "step": 22366
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.092862857142858e-06,
+      "loss": 1.1838,
+      "step": 22367
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.092594285714286e-06,
+      "loss": 1.2942,
+      "step": 22368
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.092325714285714e-06,
+      "loss": 1.3207,
+      "step": 22369
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.092057142857144e-06,
+      "loss": 1.7625,
+      "step": 22370
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.091788571428572e-06,
+      "loss": 0.819,
+      "step": 22371
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.09152e-06,
+      "loss": 1.1387,
+      "step": 22372
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.091251428571428e-06,
+      "loss": 1.2497,
+      "step": 22373
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.090982857142858e-06,
+      "loss": 1.3894,
+      "step": 22374
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.090714285714285e-06,
+      "loss": 1.0254,
+      "step": 22375
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.090445714285714e-06,
+      "loss": 1.1353,
+      "step": 22376
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.090177142857144e-06,
+      "loss": 0.7163,
+      "step": 22377
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.08990857142857e-06,
+      "loss": 1.2086,
+      "step": 22378
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.08964e-06,
+      "loss": 1.4304,
+      "step": 22379
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.089371428571429e-06,
+      "loss": 1.1727,
+      "step": 22380
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.089102857142859e-06,
+      "loss": 0.9666,
+      "step": 22381
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.088834285714285e-06,
+      "loss": 1.4091,
+      "step": 22382
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.088565714285715e-06,
+      "loss": 1.2648,
+      "step": 22383
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.088297142857143e-06,
+      "loss": 1.194,
+      "step": 22384
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.088028571428571e-06,
+      "loss": 0.9164,
+      "step": 22385
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.08776e-06,
+      "loss": 1.5268,
+      "step": 22386
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.08749142857143e-06,
+      "loss": 1.1136,
+      "step": 22387
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.087222857142856e-06,
+      "loss": 1.2924,
+      "step": 22388
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.086954285714286e-06,
+      "loss": 1.1617,
+      "step": 22389
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.086685714285716e-06,
+      "loss": 1.3825,
+      "step": 22390
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.086417142857142e-06,
+      "loss": 1.2258,
+      "step": 22391
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.086148571428572e-06,
+      "loss": 1.2679,
+      "step": 22392
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.08588e-06,
+      "loss": 1.1742,
+      "step": 22393
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.08561142857143e-06,
+      "loss": 1.1177,
+      "step": 22394
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.085342857142857e-06,
+      "loss": 0.659,
+      "step": 22395
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.085074285714286e-06,
+      "loss": 1.475,
+      "step": 22396
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.084805714285715e-06,
+      "loss": 0.9087,
+      "step": 22397
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.084537142857143e-06,
+      "loss": 1.368,
+      "step": 22398
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.084268571428571e-06,
+      "loss": 1.4512,
+      "step": 22399
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.084000000000001e-06,
+      "loss": 1.173,
+      "step": 22400
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.083731428571427e-06,
+      "loss": 1.0779,
+      "step": 22401
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.083462857142857e-06,
+      "loss": 1.3241,
+      "step": 22402
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.083194285714287e-06,
+      "loss": 1.1699,
+      "step": 22403
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.082925714285715e-06,
+      "loss": 1.3494,
+      "step": 22404
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.082657142857143e-06,
+      "loss": 1.0134,
+      "step": 22405
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.082388571428572e-06,
+      "loss": 0.9837,
+      "step": 22406
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.082120000000001e-06,
+      "loss": 1.307,
+      "step": 22407
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.081851428571428e-06,
+      "loss": 0.9869,
+      "step": 22408
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.081582857142858e-06,
+      "loss": 1.0604,
+      "step": 22409
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.081314285714286e-06,
+      "loss": 1.3811,
+      "step": 22410
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.081045714285714e-06,
+      "loss": 1.2309,
+      "step": 22411
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.080777142857142e-06,
+      "loss": 1.4143,
+      "step": 22412
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.080508571428572e-06,
+      "loss": 0.9483,
+      "step": 22413
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.080239999999999e-06,
+      "loss": 1.2765,
+      "step": 22414
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.079971428571429e-06,
+      "loss": 0.9569,
+      "step": 22415
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.079702857142858e-06,
+      "loss": 1.1121,
+      "step": 22416
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.079434285714287e-06,
+      "loss": 1.0474,
+      "step": 22417
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.079165714285715e-06,
+      "loss": 0.925,
+      "step": 22418
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.078897142857143e-06,
+      "loss": 1.5292,
+      "step": 22419
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.078628571428573e-06,
+      "loss": 1.4421,
+      "step": 22420
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07836e-06,
+      "loss": 0.8476,
+      "step": 22421
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07809142857143e-06,
+      "loss": 0.9024,
+      "step": 22422
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.077822857142857e-06,
+      "loss": 0.9337,
+      "step": 22423
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.077554285714286e-06,
+      "loss": 1.1352,
+      "step": 22424
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.077285714285714e-06,
+      "loss": 1.0391,
+      "step": 22425
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.077017142857144e-06,
+      "loss": 1.445,
+      "step": 22426
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.076748571428572e-06,
+      "loss": 1.2468,
+      "step": 22427
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07648e-06,
+      "loss": 1.1629,
+      "step": 22428
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07621142857143e-06,
+      "loss": 1.1386,
+      "step": 22429
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.075942857142858e-06,
+      "loss": 1.1867,
+      "step": 22430
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.075674285714286e-06,
+      "loss": 1.0994,
+      "step": 22431
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.075405714285714e-06,
+      "loss": 1.3632,
+      "step": 22432
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.075137142857144e-06,
+      "loss": 1.0186,
+      "step": 22433
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07486857142857e-06,
+      "loss": 1.4069,
+      "step": 22434
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.0746e-06,
+      "loss": 1.2591,
+      "step": 22435
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.074331428571429e-06,
+      "loss": 1.0699,
+      "step": 22436
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.074062857142857e-06,
+      "loss": 1.0147,
+      "step": 22437
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.073794285714285e-06,
+      "loss": 1.6729,
+      "step": 22438
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.073525714285715e-06,
+      "loss": 1.2001,
+      "step": 22439
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.073257142857143e-06,
+      "loss": 0.7875,
+      "step": 22440
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.072988571428571e-06,
+      "loss": 0.9766,
+      "step": 22441
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.072720000000001e-06,
+      "loss": 1.2593,
+      "step": 22442
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07245142857143e-06,
+      "loss": 0.9256,
+      "step": 22443
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.072182857142858e-06,
+      "loss": 1.3025,
+      "step": 22444
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.071914285714286e-06,
+      "loss": 1.5457,
+      "step": 22445
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.071645714285716e-06,
+      "loss": 1.0865,
+      "step": 22446
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.071377142857142e-06,
+      "loss": 1.04,
+      "step": 22447
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.071108571428572e-06,
+      "loss": 1.2229,
+      "step": 22448
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07084e-06,
+      "loss": 0.9936,
+      "step": 22449
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.07057142857143e-06,
+      "loss": 1.0711,
+      "step": 22450
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.070302857142856e-06,
+      "loss": 0.8533,
+      "step": 22451
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.070034285714286e-06,
+      "loss": 1.1503,
+      "step": 22452
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.069765714285715e-06,
+      "loss": 1.0848,
+      "step": 22453
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.069497142857143e-06,
+      "loss": 1.2956,
+      "step": 22454
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.069228571428573e-06,
+      "loss": 1.3653,
+      "step": 22455
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.06896e-06,
+      "loss": 1.0284,
+      "step": 22456
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.068691428571429e-06,
+      "loss": 0.8697,
+      "step": 22457
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.068422857142857e-06,
+      "loss": 0.8167,
+      "step": 22458
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.068154285714287e-06,
+      "loss": 1.0666,
+      "step": 22459
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.067885714285713e-06,
+      "loss": 1.1797,
+      "step": 22460
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.067617142857143e-06,
+      "loss": 1.0586,
+      "step": 22461
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.067348571428572e-06,
+      "loss": 1.2907,
+      "step": 22462
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.067080000000001e-06,
+      "loss": 1.0858,
+      "step": 22463
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.066811428571428e-06,
+      "loss": 1.458,
+      "step": 22464
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.066542857142858e-06,
+      "loss": 1.1289,
+      "step": 22465
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.066274285714286e-06,
+      "loss": 1.3913,
+      "step": 22466
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.066005714285714e-06,
+      "loss": 1.5875,
+      "step": 22467
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.065737142857144e-06,
+      "loss": 1.1519,
+      "step": 22468
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.065468571428572e-06,
+      "loss": 0.8521,
+      "step": 22469
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.0652e-06,
+      "loss": 1.4436,
+      "step": 22470
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.064931428571428e-06,
+      "loss": 1.3983,
+      "step": 22471
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.064662857142858e-06,
+      "loss": 1.0182,
+      "step": 22472
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.064394285714287e-06,
+      "loss": 1.2979,
+      "step": 22473
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.064125714285715e-06,
+      "loss": 1.1242,
+      "step": 22474
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.063857142857143e-06,
+      "loss": 1.053,
+      "step": 22475
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.063588571428573e-06,
+      "loss": 1.3718,
+      "step": 22476
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.06332e-06,
+      "loss": 1.4431,
+      "step": 22477
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.063051428571429e-06,
+      "loss": 1.0315,
+      "step": 22478
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.062782857142857e-06,
+      "loss": 1.3275,
+      "step": 22479
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.062514285714285e-06,
+      "loss": 1.3522,
+      "step": 22480
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.062245714285714e-06,
+      "loss": 0.9183,
+      "step": 22481
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.061977142857143e-06,
+      "loss": 1.1561,
+      "step": 22482
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.061708571428572e-06,
+      "loss": 1.1557,
+      "step": 22483
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.06144e-06,
+      "loss": 1.77,
+      "step": 22484
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.06117142857143e-06,
+      "loss": 1.5709,
+      "step": 22485
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.060902857142858e-06,
+      "loss": 0.8545,
+      "step": 22486
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.060634285714286e-06,
+      "loss": 1.3114,
+      "step": 22487
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.060365714285714e-06,
+      "loss": 1.1119,
+      "step": 22488
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.060097142857144e-06,
+      "loss": 1.311,
+      "step": 22489
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.05982857142857e-06,
+      "loss": 1.4577,
+      "step": 22490
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.05956e-06,
+      "loss": 1.2973,
+      "step": 22491
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.059291428571429e-06,
+      "loss": 1.537,
+      "step": 22492
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.059022857142857e-06,
+      "loss": 0.9052,
+      "step": 22493
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.058754285714285e-06,
+      "loss": 1.0081,
+      "step": 22494
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.058485714285715e-06,
+      "loss": 1.2938,
+      "step": 22495
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.058217142857145e-06,
+      "loss": 1.0997,
+      "step": 22496
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.057948571428571e-06,
+      "loss": 0.9239,
+      "step": 22497
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.057680000000001e-06,
+      "loss": 1.0451,
+      "step": 22498
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.05741142857143e-06,
+      "loss": 0.9005,
+      "step": 22499
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.057142857142857e-06,
+      "loss": 1.1889,
+      "step": 22500
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.056874285714286e-06,
+      "loss": 0.9853,
+      "step": 22501
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.056605714285715e-06,
+      "loss": 1.308,
+      "step": 22502
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.056337142857142e-06,
+      "loss": 0.8421,
+      "step": 22503
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.056068571428572e-06,
+      "loss": 0.8854,
+      "step": 22504
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.0558e-06,
+      "loss": 1.3639,
+      "step": 22505
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.055531428571428e-06,
+      "loss": 1.3329,
+      "step": 22506
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.055262857142856e-06,
+      "loss": 0.9487,
+      "step": 22507
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.054994285714286e-06,
+      "loss": 1.3645,
+      "step": 22508
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.054725714285716e-06,
+      "loss": 1.466,
+      "step": 22509
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.054457142857143e-06,
+      "loss": 0.969,
+      "step": 22510
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.054188571428572e-06,
+      "loss": 1.1453,
+      "step": 22511
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.05392e-06,
+      "loss": 1.1685,
+      "step": 22512
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.053651428571429e-06,
+      "loss": 0.9221,
+      "step": 22513
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.053382857142857e-06,
+      "loss": 0.9622,
+      "step": 22514
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.053114285714287e-06,
+      "loss": 1.2866,
+      "step": 22515
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.052845714285713e-06,
+      "loss": 0.8448,
+      "step": 22516
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.052577142857143e-06,
+      "loss": 1.5281,
+      "step": 22517
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.052308571428571e-06,
+      "loss": 0.8751,
+      "step": 22518
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.05204e-06,
+      "loss": 1.5064,
+      "step": 22519
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.051771428571428e-06,
+      "loss": 1.0367,
+      "step": 22520
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.051502857142858e-06,
+      "loss": 1.0455,
+      "step": 22521
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.051234285714287e-06,
+      "loss": 1.3462,
+      "step": 22522
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.050965714285714e-06,
+      "loss": 1.1768,
+      "step": 22523
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.050697142857144e-06,
+      "loss": 1.1252,
+      "step": 22524
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.050428571428572e-06,
+      "loss": 0.9157,
+      "step": 22525
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.05016e-06,
+      "loss": 1.2627,
+      "step": 22526
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.049891428571428e-06,
+      "loss": 1.1678,
+      "step": 22527
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.049622857142858e-06,
+      "loss": 0.9473,
+      "step": 22528
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.049354285714285e-06,
+      "loss": 1.392,
+      "step": 22529
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.049085714285715e-06,
+      "loss": 1.0489,
+      "step": 22530
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.048817142857143e-06,
+      "loss": 1.107,
+      "step": 22531
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.048548571428573e-06,
+      "loss": 0.9025,
+      "step": 22532
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.048279999999999e-06,
+      "loss": 1.0895,
+      "step": 22533
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.048011428571429e-06,
+      "loss": 0.9399,
+      "step": 22534
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.047742857142859e-06,
+      "loss": 1.2327,
+      "step": 22535
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.047474285714285e-06,
+      "loss": 1.0705,
+      "step": 22536
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.047205714285715e-06,
+      "loss": 1.0968,
+      "step": 22537
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.046937142857143e-06,
+      "loss": 0.8535,
+      "step": 22538
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.046668571428572e-06,
+      "loss": 0.9601,
+      "step": 22539
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.0464e-06,
+      "loss": 1.0084,
+      "step": 22540
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.04613142857143e-06,
+      "loss": 1.0747,
+      "step": 22541
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.045862857142856e-06,
+      "loss": 1.4354,
+      "step": 22542
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.045594285714286e-06,
+      "loss": 1.2363,
+      "step": 22543
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.045325714285714e-06,
+      "loss": 1.183,
+      "step": 22544
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.045057142857144e-06,
+      "loss": 0.9506,
+      "step": 22545
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.04478857142857e-06,
+      "loss": 1.1724,
+      "step": 22546
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.04452e-06,
+      "loss": 1.1619,
+      "step": 22547
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.04425142857143e-06,
+      "loss": 1.4718,
+      "step": 22548
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.043982857142857e-06,
+      "loss": 1.5408,
+      "step": 22549
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.043714285714287e-06,
+      "loss": 0.9149,
+      "step": 22550
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.043445714285715e-06,
+      "loss": 1.3445,
+      "step": 22551
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.043177142857143e-06,
+      "loss": 1.197,
+      "step": 22552
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.042908571428571e-06,
+      "loss": 1.5067,
+      "step": 22553
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.042640000000001e-06,
+      "loss": 1.2158,
+      "step": 22554
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.04237142857143e-06,
+      "loss": 1.783,
+      "step": 22555
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.042102857142857e-06,
+      "loss": 0.9869,
+      "step": 22556
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.041834285714286e-06,
+      "loss": 1.0579,
+      "step": 22557
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.041565714285715e-06,
+      "loss": 1.1077,
+      "step": 22558
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.041297142857142e-06,
+      "loss": 1.3257,
+      "step": 22559
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.041028571428572e-06,
+      "loss": 0.9766,
+      "step": 22560
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.040760000000002e-06,
+      "loss": 1.3684,
+      "step": 22561
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.040491428571428e-06,
+      "loss": 1.417,
+      "step": 22562
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.040222857142858e-06,
+      "loss": 1.1966,
+      "step": 22563
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.039954285714286e-06,
+      "loss": 1.2867,
+      "step": 22564
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.039685714285714e-06,
+      "loss": 0.8459,
+      "step": 22565
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.039417142857142e-06,
+      "loss": 0.6816,
+      "step": 22566
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.039148571428572e-06,
+      "loss": 0.8768,
+      "step": 22567
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.03888e-06,
+      "loss": 1.1645,
+      "step": 22568
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.038611428571429e-06,
+      "loss": 1.6653,
+      "step": 22569
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.038342857142857e-06,
+      "loss": 1.3299,
+      "step": 22570
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.038074285714287e-06,
+      "loss": 1.1831,
+      "step": 22571
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.037805714285713e-06,
+      "loss": 1.1661,
+      "step": 22572
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.037537142857143e-06,
+      "loss": 1.2246,
+      "step": 22573
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.037268571428573e-06,
+      "loss": 1.0481,
+      "step": 22574
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.037e-06,
+      "loss": 0.8041,
+      "step": 22575
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.03673142857143e-06,
+      "loss": 1.2937,
+      "step": 22576
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.036462857142857e-06,
+      "loss": 1.4374,
+      "step": 22577
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.036194285714287e-06,
+      "loss": 0.9609,
+      "step": 22578
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.035925714285714e-06,
+      "loss": 0.903,
+      "step": 22579
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.035657142857144e-06,
+      "loss": 1.0503,
+      "step": 22580
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.035388571428572e-06,
+      "loss": 0.9797,
+      "step": 22581
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.03512e-06,
+      "loss": 1.1092,
+      "step": 22582
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.034851428571428e-06,
+      "loss": 0.7757,
+      "step": 22583
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.034582857142858e-06,
+      "loss": 0.9336,
+      "step": 22584
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.034314285714285e-06,
+      "loss": 1.1036,
+      "step": 22585
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.034045714285714e-06,
+      "loss": 1.3168,
+      "step": 22586
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.033777142857143e-06,
+      "loss": 1.2669,
+      "step": 22587
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.03350857142857e-06,
+      "loss": 1.0169,
+      "step": 22588
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.03324e-06,
+      "loss": 0.8479,
+      "step": 22589
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.032971428571429e-06,
+      "loss": 1.4256,
+      "step": 22590
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.032702857142859e-06,
+      "loss": 1.3416,
+      "step": 22591
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.032434285714285e-06,
+      "loss": 1.2116,
+      "step": 22592
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.032165714285715e-06,
+      "loss": 1.3438,
+      "step": 22593
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.031897142857143e-06,
+      "loss": 1.411,
+      "step": 22594
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.031628571428571e-06,
+      "loss": 0.9187,
+      "step": 22595
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.03136e-06,
+      "loss": 0.9796,
+      "step": 22596
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.03109142857143e-06,
+      "loss": 0.8604,
+      "step": 22597
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.030822857142856e-06,
+      "loss": 1.1133,
+      "step": 22598
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.030554285714286e-06,
+      "loss": 1.0232,
+      "step": 22599
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.030285714285714e-06,
+      "loss": 1.1917,
+      "step": 22600
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.030017142857144e-06,
+      "loss": 0.8996,
+      "step": 22601
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.029748571428572e-06,
+      "loss": 1.1563,
+      "step": 22602
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.02948e-06,
+      "loss": 0.9854,
+      "step": 22603
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.02921142857143e-06,
+      "loss": 0.8205,
+      "step": 22604
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.028942857142857e-06,
+      "loss": 1.0843,
+      "step": 22605
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.028674285714286e-06,
+      "loss": 1.1609,
+      "step": 22606
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.028405714285715e-06,
+      "loss": 1.405,
+      "step": 22607
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.028137142857143e-06,
+      "loss": 1.1032,
+      "step": 22608
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.027868571428571e-06,
+      "loss": 1.5568,
+      "step": 22609
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.027600000000001e-06,
+      "loss": 1.3023,
+      "step": 22610
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.027331428571427e-06,
+      "loss": 1.5358,
+      "step": 22611
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.027062857142857e-06,
+      "loss": 0.9084,
+      "step": 22612
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.026794285714285e-06,
+      "loss": 0.9702,
+      "step": 22613
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.026525714285715e-06,
+      "loss": 1.2126,
+      "step": 22614
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.026257142857143e-06,
+      "loss": 1.3178,
+      "step": 22615
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.025988571428572e-06,
+      "loss": 0.9249,
+      "step": 22616
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.025720000000001e-06,
+      "loss": 1.2384,
+      "step": 22617
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.025451428571428e-06,
+      "loss": 0.7597,
+      "step": 22618
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.025182857142858e-06,
+      "loss": 0.7597,
+      "step": 22619
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.024914285714286e-06,
+      "loss": 1.2086,
+      "step": 22620
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.024645714285714e-06,
+      "loss": 1.2338,
+      "step": 22621
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.024377142857142e-06,
+      "loss": 1.6085,
+      "step": 22622
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.024108571428572e-06,
+      "loss": 1.7395,
+      "step": 22623
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.02384e-06,
+      "loss": 1.0114,
+      "step": 22624
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.023571428571429e-06,
+      "loss": 1.2047,
+      "step": 22625
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.023302857142857e-06,
+      "loss": 0.7144,
+      "step": 22626
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.023034285714287e-06,
+      "loss": 1.3244,
+      "step": 22627
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.022765714285715e-06,
+      "loss": 1.4042,
+      "step": 22628
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.022497142857143e-06,
+      "loss": 1.3891,
+      "step": 22629
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.022228571428573e-06,
+      "loss": 1.0447,
+      "step": 22630
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.02196e-06,
+      "loss": 1.1843,
+      "step": 22631
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.02169142857143e-06,
+      "loss": 0.7573,
+      "step": 22632
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.021422857142857e-06,
+      "loss": 1.7252,
+      "step": 22633
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.021154285714286e-06,
+      "loss": 1.6592,
+      "step": 22634
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.020885714285714e-06,
+      "loss": 1.1316,
+      "step": 22635
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.020617142857144e-06,
+      "loss": 1.5167,
+      "step": 22636
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.020348571428572e-06,
+      "loss": 1.6049,
+      "step": 22637
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.02008e-06,
+      "loss": 1.2708,
+      "step": 22638
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.019811428571428e-06,
+      "loss": 1.7161,
+      "step": 22639
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.019542857142858e-06,
+      "loss": 1.0605,
+      "step": 22640
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.019274285714286e-06,
+      "loss": 1.1401,
+      "step": 22641
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.019005714285714e-06,
+      "loss": 0.7053,
+      "step": 22642
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.018737142857144e-06,
+      "loss": 0.9993,
+      "step": 22643
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.01846857142857e-06,
+      "loss": 1.2104,
+      "step": 22644
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.0182e-06,
+      "loss": 0.9279,
+      "step": 22645
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.017931428571429e-06,
+      "loss": 1.0607,
+      "step": 22646
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.017662857142859e-06,
+      "loss": 1.2509,
+      "step": 22647
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.017394285714285e-06,
+      "loss": 1.1435,
+      "step": 22648
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.017125714285715e-06,
+      "loss": 0.9474,
+      "step": 22649
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.016857142857143e-06,
+      "loss": 0.8296,
+      "step": 22650
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.016588571428571e-06,
+      "loss": 1.2452,
+      "step": 22651
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.01632e-06,
+      "loss": 1.3347,
+      "step": 22652
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.01605142857143e-06,
+      "loss": 1.2693,
+      "step": 22653
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.015782857142858e-06,
+      "loss": 1.3985,
+      "step": 22654
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.015514285714286e-06,
+      "loss": 1.2892,
+      "step": 22655
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.015245714285716e-06,
+      "loss": 0.9233,
+      "step": 22656
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.014977142857142e-06,
+      "loss": 1.4608,
+      "step": 22657
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.014708571428572e-06,
+      "loss": 0.7939,
+      "step": 22658
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.01444e-06,
+      "loss": 1.118,
+      "step": 22659
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.01417142857143e-06,
+      "loss": 1.5378,
+      "step": 22660
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.013902857142856e-06,
+      "loss": 1.1474,
+      "step": 22661
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.013634285714286e-06,
+      "loss": 1.25,
+      "step": 22662
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.013365714285715e-06,
+      "loss": 1.1274,
+      "step": 22663
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.013097142857143e-06,
+      "loss": 1.6166,
+      "step": 22664
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.012828571428571e-06,
+      "loss": 1.5387,
+      "step": 22665
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.01256e-06,
+      "loss": 0.9247,
+      "step": 22666
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.012291428571429e-06,
+      "loss": 1.1374,
+      "step": 22667
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.012022857142857e-06,
+      "loss": 1.2662,
+      "step": 22668
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.011754285714287e-06,
+      "loss": 1.4827,
+      "step": 22669
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.011485714285715e-06,
+      "loss": 0.9924,
+      "step": 22670
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.011217142857143e-06,
+      "loss": 0.9573,
+      "step": 22671
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.010948571428572e-06,
+      "loss": 0.9337,
+      "step": 22672
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.010680000000001e-06,
+      "loss": 0.9743,
+      "step": 22673
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.010411428571428e-06,
+      "loss": 0.9559,
+      "step": 22674
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.010142857142858e-06,
+      "loss": 0.8731,
+      "step": 22675
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.009874285714286e-06,
+      "loss": 1.0886,
+      "step": 22676
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.009605714285714e-06,
+      "loss": 0.9476,
+      "step": 22677
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.009337142857142e-06,
+      "loss": 1.0716,
+      "step": 22678
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.009068571428572e-06,
+      "loss": 1.0685,
+      "step": 22679
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.0088e-06,
+      "loss": 0.8638,
+      "step": 22680
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.008531428571428e-06,
+      "loss": 1.0315,
+      "step": 22681
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.008262857142858e-06,
+      "loss": 1.1888,
+      "step": 22682
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.007994285714287e-06,
+      "loss": 1.5123,
+      "step": 22683
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.007725714285715e-06,
+      "loss": 1.0552,
+      "step": 22684
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.007457142857143e-06,
+      "loss": 1.1542,
+      "step": 22685
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.007188571428573e-06,
+      "loss": 1.247,
+      "step": 22686
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.00692e-06,
+      "loss": 0.9138,
+      "step": 22687
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.006651428571429e-06,
+      "loss": 1.2282,
+      "step": 22688
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.006382857142857e-06,
+      "loss": 0.9497,
+      "step": 22689
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.006114285714285e-06,
+      "loss": 1.1586,
+      "step": 22690
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.005845714285714e-06,
+      "loss": 0.9426,
+      "step": 22691
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.005577142857143e-06,
+      "loss": 0.9517,
+      "step": 22692
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.005308571428572e-06,
+      "loss": 1.2489,
+      "step": 22693
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.00504e-06,
+      "loss": 1.1221,
+      "step": 22694
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.00477142857143e-06,
+      "loss": 1.2238,
+      "step": 22695
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.004502857142858e-06,
+      "loss": 1.1935,
+      "step": 22696
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.004234285714286e-06,
+      "loss": 1.3273,
+      "step": 22697
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.003965714285714e-06,
+      "loss": 1.4879,
+      "step": 22698
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.003697142857144e-06,
+      "loss": 1.0913,
+      "step": 22699
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.00342857142857e-06,
+      "loss": 1.0083,
+      "step": 22700
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.00316e-06,
+      "loss": 1.2886,
+      "step": 22701
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.002891428571429e-06,
+      "loss": 1.1758,
+      "step": 22702
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.002622857142857e-06,
+      "loss": 1.416,
+      "step": 22703
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.002354285714285e-06,
+      "loss": 1.3474,
+      "step": 22704
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.002085714285715e-06,
+      "loss": 1.2412,
+      "step": 22705
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.001817142857143e-06,
+      "loss": 0.9655,
+      "step": 22706
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.001548571428571e-06,
+      "loss": 1.4403,
+      "step": 22707
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.001280000000001e-06,
+      "loss": 0.9929,
+      "step": 22708
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.00101142857143e-06,
+      "loss": 1.1637,
+      "step": 22709
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.000742857142857e-06,
+      "loss": 1.0423,
+      "step": 22710
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.000474285714286e-06,
+      "loss": 1.4368,
+      "step": 22711
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.000205714285715e-06,
+      "loss": 1.1087,
+      "step": 22712
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.999937142857142e-06,
+      "loss": 1.2759,
+      "step": 22713
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.999668571428572e-06,
+      "loss": 1.0006,
+      "step": 22714
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.9994e-06,
+      "loss": 1.0093,
+      "step": 22715
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.999131428571428e-06,
+      "loss": 1.2195,
+      "step": 22716
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.998862857142856e-06,
+      "loss": 1.3775,
+      "step": 22717
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.998594285714286e-06,
+      "loss": 1.0959,
+      "step": 22718
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.998325714285714e-06,
+      "loss": 1.502,
+      "step": 22719
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.998057142857143e-06,
+      "loss": 1.2147,
+      "step": 22720
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.997788571428572e-06,
+      "loss": 0.7732,
+      "step": 22721
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.99752e-06,
+      "loss": 0.7656,
+      "step": 22722
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.997251428571429e-06,
+      "loss": 1.2749,
+      "step": 22723
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.996982857142857e-06,
+      "loss": 1.3751,
+      "step": 22724
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.996714285714287e-06,
+      "loss": 1.2429,
+      "step": 22725
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.996445714285713e-06,
+      "loss": 1.5593,
+      "step": 22726
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.996177142857143e-06,
+      "loss": 1.2012,
+      "step": 22727
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.995908571428571e-06,
+      "loss": 1.3364,
+      "step": 22728
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.995640000000001e-06,
+      "loss": 1.0356,
+      "step": 22729
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.995371428571428e-06,
+      "loss": 1.5001,
+      "step": 22730
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.995102857142858e-06,
+      "loss": 1.2068,
+      "step": 22731
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.994834285714286e-06,
+      "loss": 1.0214,
+      "step": 22732
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.994565714285714e-06,
+      "loss": 1.355,
+      "step": 22733
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.994297142857144e-06,
+      "loss": 1.1489,
+      "step": 22734
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.994028571428572e-06,
+      "loss": 0.8663,
+      "step": 22735
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.99376e-06,
+      "loss": 0.8638,
+      "step": 22736
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.993491428571428e-06,
+      "loss": 1.1246,
+      "step": 22737
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.993222857142858e-06,
+      "loss": 0.8039,
+      "step": 22738
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.992954285714285e-06,
+      "loss": 1.0363,
+      "step": 22739
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.992685714285715e-06,
+      "loss": 1.4025,
+      "step": 22740
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.992417142857143e-06,
+      "loss": 1.2517,
+      "step": 22741
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.992148571428573e-06,
+      "loss": 1.1294,
+      "step": 22742
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.991879999999999e-06,
+      "loss": 0.8164,
+      "step": 22743
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.991611428571429e-06,
+      "loss": 1.3121,
+      "step": 22744
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.991342857142857e-06,
+      "loss": 1.0811,
+      "step": 22745
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.991074285714285e-06,
+      "loss": 1.2254,
+      "step": 22746
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.990805714285715e-06,
+      "loss": 1.1448,
+      "step": 22747
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.990537142857143e-06,
+      "loss": 1.2154,
+      "step": 22748
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.990268571428572e-06,
+      "loss": 1.5529,
+      "step": 22749
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.99e-06,
+      "loss": 1.0253,
+      "step": 22750
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.98973142857143e-06,
+      "loss": 1.2026,
+      "step": 22751
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.989462857142858e-06,
+      "loss": 1.029,
+      "step": 22752
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.989194285714286e-06,
+      "loss": 1.2002,
+      "step": 22753
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.988925714285714e-06,
+      "loss": 1.6435,
+      "step": 22754
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.988657142857144e-06,
+      "loss": 1.4023,
+      "step": 22755
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.98838857142857e-06,
+      "loss": 0.7806,
+      "step": 22756
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.98812e-06,
+      "loss": 1.2456,
+      "step": 22757
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.987851428571429e-06,
+      "loss": 0.8194,
+      "step": 22758
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.987582857142857e-06,
+      "loss": 0.8043,
+      "step": 22759
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.987314285714287e-06,
+      "loss": 0.9047,
+      "step": 22760
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.987045714285715e-06,
+      "loss": 0.9032,
+      "step": 22761
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.986777142857143e-06,
+      "loss": 1.2828,
+      "step": 22762
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.986508571428571e-06,
+      "loss": 1.4825,
+      "step": 22763
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.986240000000001e-06,
+      "loss": 1.3675,
+      "step": 22764
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.985971428571429e-06,
+      "loss": 1.5777,
+      "step": 22765
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.985702857142857e-06,
+      "loss": 1.6302,
+      "step": 22766
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.985434285714286e-06,
+      "loss": 0.963,
+      "step": 22767
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.985165714285715e-06,
+      "loss": 1.0115,
+      "step": 22768
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.984897142857142e-06,
+      "loss": 1.1577,
+      "step": 22769
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.984628571428572e-06,
+      "loss": 1.2268,
+      "step": 22770
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.98436e-06,
+      "loss": 1.1002,
+      "step": 22771
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.984091428571428e-06,
+      "loss": 1.0592,
+      "step": 22772
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.983822857142858e-06,
+      "loss": 1.0242,
+      "step": 22773
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.983554285714286e-06,
+      "loss": 1.1917,
+      "step": 22774
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.983285714285716e-06,
+      "loss": 0.9999,
+      "step": 22775
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.983017142857142e-06,
+      "loss": 1.2737,
+      "step": 22776
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.982748571428572e-06,
+      "loss": 0.9239,
+      "step": 22777
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.98248e-06,
+      "loss": 1.0715,
+      "step": 22778
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.982211428571429e-06,
+      "loss": 1.2992,
+      "step": 22779
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.981942857142857e-06,
+      "loss": 1.3484,
+      "step": 22780
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.981674285714287e-06,
+      "loss": 1.1969,
+      "step": 22781
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.981405714285713e-06,
+      "loss": 0.9694,
+      "step": 22782
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.981137142857143e-06,
+      "loss": 1.5155,
+      "step": 22783
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.980868571428571e-06,
+      "loss": 1.0039,
+      "step": 22784
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.9806e-06,
+      "loss": 1.1296,
+      "step": 22785
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.98033142857143e-06,
+      "loss": 1.0997,
+      "step": 22786
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.980062857142857e-06,
+      "loss": 1.4056,
+      "step": 22787
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.979794285714287e-06,
+      "loss": 1.1905,
+      "step": 22788
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.979525714285714e-06,
+      "loss": 1.1595,
+      "step": 22789
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.979257142857144e-06,
+      "loss": 0.8617,
+      "step": 22790
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.978988571428572e-06,
+      "loss": 1.1591,
+      "step": 22791
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.97872e-06,
+      "loss": 1.2189,
+      "step": 22792
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.978451428571428e-06,
+      "loss": 1.2264,
+      "step": 22793
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.978182857142858e-06,
+      "loss": 0.9532,
+      "step": 22794
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.977914285714285e-06,
+      "loss": 1.6776,
+      "step": 22795
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.977645714285714e-06,
+      "loss": 1.1993,
+      "step": 22796
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.977377142857143e-06,
+      "loss": 1.1748,
+      "step": 22797
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.977108571428573e-06,
+      "loss": 0.9173,
+      "step": 22798
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.97684e-06,
+      "loss": 1.3929,
+      "step": 22799
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.976571428571429e-06,
+      "loss": 0.7663,
+      "step": 22800
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.976302857142859e-06,
+      "loss": 1.3182,
+      "step": 22801
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.976034285714285e-06,
+      "loss": 1.1939,
+      "step": 22802
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.975765714285715e-06,
+      "loss": 1.0812,
+      "step": 22803
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.975497142857143e-06,
+      "loss": 0.9674,
+      "step": 22804
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.975228571428571e-06,
+      "loss": 1.2189,
+      "step": 22805
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.97496e-06,
+      "loss": 0.8943,
+      "step": 22806
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.97469142857143e-06,
+      "loss": 1.2932,
+      "step": 22807
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.974422857142856e-06,
+      "loss": 1.4529,
+      "step": 22808
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.974154285714286e-06,
+      "loss": 1.0803,
+      "step": 22809
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.973885714285714e-06,
+      "loss": 1.3094,
+      "step": 22810
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.973617142857144e-06,
+      "loss": 1.1846,
+      "step": 22811
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.973348571428572e-06,
+      "loss": 1.2986,
+      "step": 22812
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.97308e-06,
+      "loss": 1.4254,
+      "step": 22813
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.97281142857143e-06,
+      "loss": 0.9649,
+      "step": 22814
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.972542857142857e-06,
+      "loss": 1.0956,
+      "step": 22815
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.972274285714286e-06,
+      "loss": 0.8456,
+      "step": 22816
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.972005714285715e-06,
+      "loss": 0.6614,
+      "step": 22817
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.971737142857143e-06,
+      "loss": 1.1216,
+      "step": 22818
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.971468571428571e-06,
+      "loss": 0.7833,
+      "step": 22819
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.971200000000001e-06,
+      "loss": 1.1201,
+      "step": 22820
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.970931428571429e-06,
+      "loss": 1.2572,
+      "step": 22821
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.970662857142857e-06,
+      "loss": 0.7682,
+      "step": 22822
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.970394285714285e-06,
+      "loss": 0.7703,
+      "step": 22823
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.970125714285715e-06,
+      "loss": 0.7479,
+      "step": 22824
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.969857142857143e-06,
+      "loss": 1.1576,
+      "step": 22825
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.969588571428572e-06,
+      "loss": 0.8838,
+      "step": 22826
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.969320000000001e-06,
+      "loss": 1.3913,
+      "step": 22827
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.969051428571428e-06,
+      "loss": 0.9362,
+      "step": 22828
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.968782857142858e-06,
+      "loss": 1.2302,
+      "step": 22829
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.968514285714286e-06,
+      "loss": 0.7698,
+      "step": 22830
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.968245714285714e-06,
+      "loss": 0.9265,
+      "step": 22831
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.967977142857142e-06,
+      "loss": 1.4114,
+      "step": 22832
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.967708571428572e-06,
+      "loss": 1.6071,
+      "step": 22833
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.96744e-06,
+      "loss": 1.2013,
+      "step": 22834
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.967171428571429e-06,
+      "loss": 0.827,
+      "step": 22835
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.966902857142857e-06,
+      "loss": 1.0515,
+      "step": 22836
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.966634285714287e-06,
+      "loss": 1.2559,
+      "step": 22837
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.966365714285715e-06,
+      "loss": 1.3774,
+      "step": 22838
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.966097142857143e-06,
+      "loss": 1.1031,
+      "step": 22839
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.965828571428573e-06,
+      "loss": 1.281,
+      "step": 22840
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.96556e-06,
+      "loss": 1.4912,
+      "step": 22841
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.96529142857143e-06,
+      "loss": 1.3106,
+      "step": 22842
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.965022857142857e-06,
+      "loss": 0.9609,
+      "step": 22843
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.964754285714287e-06,
+      "loss": 0.9295,
+      "step": 22844
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.964485714285714e-06,
+      "loss": 1.5387,
+      "step": 22845
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.964217142857144e-06,
+      "loss": 1.3287,
+      "step": 22846
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.963948571428572e-06,
+      "loss": 0.576,
+      "step": 22847
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.96368e-06,
+      "loss": 1.1322,
+      "step": 22848
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.963411428571428e-06,
+      "loss": 0.6945,
+      "step": 22849
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.963142857142858e-06,
+      "loss": 1.7682,
+      "step": 22850
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.962874285714286e-06,
+      "loss": 1.6546,
+      "step": 22851
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.962605714285714e-06,
+      "loss": 1.1741,
+      "step": 22852
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.962337142857144e-06,
+      "loss": 1.4107,
+      "step": 22853
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.96206857142857e-06,
+      "loss": 1.6055,
+      "step": 22854
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.9618e-06,
+      "loss": 0.9611,
+      "step": 22855
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.961531428571429e-06,
+      "loss": 1.1816,
+      "step": 22856
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.961262857142859e-06,
+      "loss": 1.1693,
+      "step": 22857
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.960994285714285e-06,
+      "loss": 0.857,
+      "step": 22858
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.960725714285715e-06,
+      "loss": 1.2327,
+      "step": 22859
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.960457142857143e-06,
+      "loss": 1.2931,
+      "step": 22860
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.960188571428571e-06,
+      "loss": 1.4413,
+      "step": 22861
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.95992e-06,
+      "loss": 1.348,
+      "step": 22862
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.95965142857143e-06,
+      "loss": 0.9894,
+      "step": 22863
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.959382857142858e-06,
+      "loss": 1.5181,
+      "step": 22864
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.959114285714286e-06,
+      "loss": 1.2448,
+      "step": 22865
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.958845714285716e-06,
+      "loss": 1.1948,
+      "step": 22866
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.958577142857142e-06,
+      "loss": 1.3578,
+      "step": 22867
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.958308571428572e-06,
+      "loss": 1.6895,
+      "step": 22868
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.95804e-06,
+      "loss": 1.2062,
+      "step": 22869
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.95777142857143e-06,
+      "loss": 0.9386,
+      "step": 22870
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.957502857142856e-06,
+      "loss": 0.9596,
+      "step": 22871
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.957234285714286e-06,
+      "loss": 1.2376,
+      "step": 22872
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.956965714285715e-06,
+      "loss": 1.1749,
+      "step": 22873
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.956697142857143e-06,
+      "loss": 1.1147,
+      "step": 22874
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.956428571428571e-06,
+      "loss": 0.8348,
+      "step": 22875
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.95616e-06,
+      "loss": 0.9117,
+      "step": 22876
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.955891428571429e-06,
+      "loss": 1.2316,
+      "step": 22877
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.955622857142857e-06,
+      "loss": 0.7059,
+      "step": 22878
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.955354285714287e-06,
+      "loss": 1.3371,
+      "step": 22879
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.955085714285715e-06,
+      "loss": 1.2228,
+      "step": 22880
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.954817142857143e-06,
+      "loss": 0.9084,
+      "step": 22881
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.954548571428571e-06,
+      "loss": 1.4563,
+      "step": 22882
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.954280000000001e-06,
+      "loss": 1.1751,
+      "step": 22883
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.954011428571428e-06,
+      "loss": 0.9297,
+      "step": 22884
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.953742857142858e-06,
+      "loss": 1.2388,
+      "step": 22885
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.953474285714286e-06,
+      "loss": 1.0231,
+      "step": 22886
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.953205714285714e-06,
+      "loss": 0.9521,
+      "step": 22887
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.952937142857142e-06,
+      "loss": 1.2854,
+      "step": 22888
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.952668571428572e-06,
+      "loss": 1.1267,
+      "step": 22889
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.9524e-06,
+      "loss": 1.2402,
+      "step": 22890
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.952131428571428e-06,
+      "loss": 1.1284,
+      "step": 22891
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.951862857142858e-06,
+      "loss": 1.262,
+      "step": 22892
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.951594285714287e-06,
+      "loss": 0.9816,
+      "step": 22893
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.951325714285715e-06,
+      "loss": 1.1592,
+      "step": 22894
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.951057142857143e-06,
+      "loss": 0.7323,
+      "step": 22895
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.950788571428573e-06,
+      "loss": 1.0636,
+      "step": 22896
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.95052e-06,
+      "loss": 1.2932,
+      "step": 22897
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.950251428571429e-06,
+      "loss": 1.3679,
+      "step": 22898
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.949982857142857e-06,
+      "loss": 0.7965,
+      "step": 22899
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.949714285714285e-06,
+      "loss": 1.1588,
+      "step": 22900
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.949445714285714e-06,
+      "loss": 1.05,
+      "step": 22901
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.949177142857143e-06,
+      "loss": 1.1347,
+      "step": 22902
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.948908571428572e-06,
+      "loss": 1.0179,
+      "step": 22903
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.94864e-06,
+      "loss": 1.1609,
+      "step": 22904
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.94837142857143e-06,
+      "loss": 1.222,
+      "step": 22905
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.948102857142858e-06,
+      "loss": 1.0495,
+      "step": 22906
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.947834285714286e-06,
+      "loss": 0.8534,
+      "step": 22907
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.947565714285714e-06,
+      "loss": 1.3707,
+      "step": 22908
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.947297142857144e-06,
+      "loss": 1.074,
+      "step": 22909
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.94702857142857e-06,
+      "loss": 0.9608,
+      "step": 22910
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.94676e-06,
+      "loss": 1.2364,
+      "step": 22911
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.946491428571429e-06,
+      "loss": 1.1491,
+      "step": 22912
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.946222857142857e-06,
+      "loss": 1.4046,
+      "step": 22913
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.945954285714285e-06,
+      "loss": 0.6444,
+      "step": 22914
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.945685714285715e-06,
+      "loss": 1.2037,
+      "step": 22915
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.945417142857143e-06,
+      "loss": 1.0975,
+      "step": 22916
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.945148571428571e-06,
+      "loss": 0.8587,
+      "step": 22917
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.944880000000001e-06,
+      "loss": 0.8991,
+      "step": 22918
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.94461142857143e-06,
+      "loss": 1.2737,
+      "step": 22919
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.944342857142857e-06,
+      "loss": 0.956,
+      "step": 22920
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.944074285714286e-06,
+      "loss": 1.1533,
+      "step": 22921
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.943805714285715e-06,
+      "loss": 1.2051,
+      "step": 22922
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.943537142857142e-06,
+      "loss": 1.031,
+      "step": 22923
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.943268571428572e-06,
+      "loss": 1.3947,
+      "step": 22924
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.943e-06,
+      "loss": 0.7755,
+      "step": 22925
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.94273142857143e-06,
+      "loss": 1.2257,
+      "step": 22926
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.942462857142856e-06,
+      "loss": 0.8774,
+      "step": 22927
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.942194285714286e-06,
+      "loss": 1.3618,
+      "step": 22928
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.941925714285714e-06,
+      "loss": 1.1554,
+      "step": 22929
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.941657142857143e-06,
+      "loss": 1.4104,
+      "step": 22930
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.941388571428572e-06,
+      "loss": 1.0078,
+      "step": 22931
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.94112e-06,
+      "loss": 0.9957,
+      "step": 22932
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.940851428571429e-06,
+      "loss": 1.1536,
+      "step": 22933
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.940582857142857e-06,
+      "loss": 0.7887,
+      "step": 22934
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.940314285714287e-06,
+      "loss": 0.7355,
+      "step": 22935
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.940045714285713e-06,
+      "loss": 1.1231,
+      "step": 22936
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.939777142857143e-06,
+      "loss": 1.2947,
+      "step": 22937
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.939508571428571e-06,
+      "loss": 0.8621,
+      "step": 22938
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.939240000000001e-06,
+      "loss": 0.6554,
+      "step": 22939
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.938971428571428e-06,
+      "loss": 1.0072,
+      "step": 22940
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.938702857142858e-06,
+      "loss": 0.8458,
+      "step": 22941
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.938434285714286e-06,
+      "loss": 1.1831,
+      "step": 22942
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.938165714285714e-06,
+      "loss": 1.2168,
+      "step": 22943
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.937897142857144e-06,
+      "loss": 0.9023,
+      "step": 22944
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.937628571428572e-06,
+      "loss": 1.3098,
+      "step": 22945
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.93736e-06,
+      "loss": 1.0365,
+      "step": 22946
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.937091428571428e-06,
+      "loss": 0.9031,
+      "step": 22947
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.936822857142858e-06,
+      "loss": 1.4455,
+      "step": 22948
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.936554285714286e-06,
+      "loss": 0.9839,
+      "step": 22949
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.936285714285715e-06,
+      "loss": 1.2639,
+      "step": 22950
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.936017142857143e-06,
+      "loss": 0.7423,
+      "step": 22951
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.935748571428573e-06,
+      "loss": 0.9787,
+      "step": 22952
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.935479999999999e-06,
+      "loss": 1.1451,
+      "step": 22953
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.935211428571429e-06,
+      "loss": 0.8326,
+      "step": 22954
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.934942857142857e-06,
+      "loss": 1.1751,
+      "step": 22955
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.934674285714285e-06,
+      "loss": 0.6534,
+      "step": 22956
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.934405714285715e-06,
+      "loss": 1.4015,
+      "step": 22957
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.934137142857143e-06,
+      "loss": 1.3513,
+      "step": 22958
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.933868571428572e-06,
+      "loss": 1.302,
+      "step": 22959
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.9336e-06,
+      "loss": 1.0608,
+      "step": 22960
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.93333142857143e-06,
+      "loss": 1.1959,
+      "step": 22961
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.933062857142858e-06,
+      "loss": 1.0334,
+      "step": 22962
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.932794285714286e-06,
+      "loss": 1.1787,
+      "step": 22963
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.932525714285714e-06,
+      "loss": 1.1351,
+      "step": 22964
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.932257142857144e-06,
+      "loss": 1.2412,
+      "step": 22965
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.93198857142857e-06,
+      "loss": 1.1313,
+      "step": 22966
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.93172e-06,
+      "loss": 0.9042,
+      "step": 22967
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.931451428571429e-06,
+      "loss": 1.4155,
+      "step": 22968
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.931182857142857e-06,
+      "loss": 1.3107,
+      "step": 22969
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.930914285714287e-06,
+      "loss": 1.3297,
+      "step": 22970
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.930645714285715e-06,
+      "loss": 0.8888,
+      "step": 22971
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.930377142857145e-06,
+      "loss": 1.3016,
+      "step": 22972
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.930108571428571e-06,
+      "loss": 1.3233,
+      "step": 22973
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.929840000000001e-06,
+      "loss": 1.6859,
+      "step": 22974
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.929571428571429e-06,
+      "loss": 0.9233,
+      "step": 22975
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.929302857142857e-06,
+      "loss": 0.9001,
+      "step": 22976
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.929034285714286e-06,
+      "loss": 1.6064,
+      "step": 22977
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.928765714285715e-06,
+      "loss": 1.2929,
+      "step": 22978
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.928497142857142e-06,
+      "loss": 1.0527,
+      "step": 22979
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.928228571428572e-06,
+      "loss": 1.3059,
+      "step": 22980
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.92796e-06,
+      "loss": 0.9184,
+      "step": 22981
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.927691428571428e-06,
+      "loss": 1.5935,
+      "step": 22982
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.927422857142858e-06,
+      "loss": 0.9861,
+      "step": 22983
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.927154285714286e-06,
+      "loss": 1.0532,
+      "step": 22984
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.926885714285716e-06,
+      "loss": 1.1247,
+      "step": 22985
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.926617142857142e-06,
+      "loss": 1.1331,
+      "step": 22986
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.926348571428572e-06,
+      "loss": 1.0599,
+      "step": 22987
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.92608e-06,
+      "loss": 1.2519,
+      "step": 22988
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.925811428571429e-06,
+      "loss": 1.1248,
+      "step": 22989
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.925542857142857e-06,
+      "loss": 1.1998,
+      "step": 22990
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.925274285714287e-06,
+      "loss": 1.1959,
+      "step": 22991
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.925005714285713e-06,
+      "loss": 1.746,
+      "step": 22992
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.924737142857143e-06,
+      "loss": 1.0744,
+      "step": 22993
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.924468571428571e-06,
+      "loss": 1.5411,
+      "step": 22994
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.924200000000001e-06,
+      "loss": 1.2277,
+      "step": 22995
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.92393142857143e-06,
+      "loss": 1.0405,
+      "step": 22996
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.923662857142857e-06,
+      "loss": 1.3097,
+      "step": 22997
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.923394285714287e-06,
+      "loss": 1.0927,
+      "step": 22998
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.923125714285714e-06,
+      "loss": 1.2083,
+      "step": 22999
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.922857142857144e-06,
+      "loss": 1.2285,
+      "step": 23000
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.922588571428572e-06,
+      "loss": 0.7142,
+      "step": 23001
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.92232e-06,
+      "loss": 1.2463,
+      "step": 23002
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.922051428571428e-06,
+      "loss": 1.375,
+      "step": 23003
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.921782857142858e-06,
+      "loss": 1.1043,
+      "step": 23004
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.921514285714285e-06,
+      "loss": 1.345,
+      "step": 23005
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.921245714285714e-06,
+      "loss": 1.4875,
+      "step": 23006
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.920977142857143e-06,
+      "loss": 1.0221,
+      "step": 23007
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.920708571428573e-06,
+      "loss": 1.3646,
+      "step": 23008
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.92044e-06,
+      "loss": 0.8451,
+      "step": 23009
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.920171428571429e-06,
+      "loss": 1.4827,
+      "step": 23010
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.919902857142859e-06,
+      "loss": 1.7058,
+      "step": 23011
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.919634285714285e-06,
+      "loss": 1.4391,
+      "step": 23012
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.919365714285715e-06,
+      "loss": 1.1556,
+      "step": 23013
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.919097142857143e-06,
+      "loss": 1.0464,
+      "step": 23014
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.918828571428571e-06,
+      "loss": 1.1828,
+      "step": 23015
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.91856e-06,
+      "loss": 1.4449,
+      "step": 23016
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.91829142857143e-06,
+      "loss": 1.345,
+      "step": 23017
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.918022857142856e-06,
+      "loss": 0.8856,
+      "step": 23018
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.917754285714286e-06,
+      "loss": 1.1083,
+      "step": 23019
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.917485714285714e-06,
+      "loss": 1.3005,
+      "step": 23020
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.917217142857144e-06,
+      "loss": 1.2181,
+      "step": 23021
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.916948571428572e-06,
+      "loss": 0.7095,
+      "step": 23022
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.91668e-06,
+      "loss": 1.0761,
+      "step": 23023
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.91641142857143e-06,
+      "loss": 0.9535,
+      "step": 23024
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.916142857142857e-06,
+      "loss": 1.5416,
+      "step": 23025
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.915874285714286e-06,
+      "loss": 1.0561,
+      "step": 23026
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.915605714285715e-06,
+      "loss": 1.1451,
+      "step": 23027
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.915337142857143e-06,
+      "loss": 1.3536,
+      "step": 23028
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.915068571428571e-06,
+      "loss": 0.6697,
+      "step": 23029
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.914800000000001e-06,
+      "loss": 1.4107,
+      "step": 23030
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.914531428571429e-06,
+      "loss": 1.2293,
+      "step": 23031
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.914262857142857e-06,
+      "loss": 1.08,
+      "step": 23032
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.913994285714285e-06,
+      "loss": 1.5126,
+      "step": 23033
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.913725714285715e-06,
+      "loss": 1.1907,
+      "step": 23034
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.913457142857143e-06,
+      "loss": 0.9641,
+      "step": 23035
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.913188571428572e-06,
+      "loss": 0.8264,
+      "step": 23036
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.912920000000001e-06,
+      "loss": 1.0898,
+      "step": 23037
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.912651428571428e-06,
+      "loss": 0.8603,
+      "step": 23038
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.912382857142858e-06,
+      "loss": 0.7602,
+      "step": 23039
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.912114285714286e-06,
+      "loss": 1.4706,
+      "step": 23040
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.911845714285714e-06,
+      "loss": 1.0324,
+      "step": 23041
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.911577142857142e-06,
+      "loss": 1.1369,
+      "step": 23042
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.911308571428572e-06,
+      "loss": 1.1228,
+      "step": 23043
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.91104e-06,
+      "loss": 1.0982,
+      "step": 23044
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.910771428571429e-06,
+      "loss": 1.3326,
+      "step": 23045
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.910502857142857e-06,
+      "loss": 1.1881,
+      "step": 23046
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.910234285714287e-06,
+      "loss": 0.8886,
+      "step": 23047
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.909965714285715e-06,
+      "loss": 1.2242,
+      "step": 23048
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.909697142857143e-06,
+      "loss": 1.0785,
+      "step": 23049
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.909428571428573e-06,
+      "loss": 1.3345,
+      "step": 23050
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90916e-06,
+      "loss": 0.9894,
+      "step": 23051
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90889142857143e-06,
+      "loss": 1.184,
+      "step": 23052
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.908622857142857e-06,
+      "loss": 1.0353,
+      "step": 23053
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.908354285714287e-06,
+      "loss": 1.0397,
+      "step": 23054
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.908085714285714e-06,
+      "loss": 1.0015,
+      "step": 23055
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.907817142857144e-06,
+      "loss": 1.4654,
+      "step": 23056
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.907548571428572e-06,
+      "loss": 1.0726,
+      "step": 23057
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90728e-06,
+      "loss": 1.0712,
+      "step": 23058
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.907011428571428e-06,
+      "loss": 0.8861,
+      "step": 23059
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.906742857142858e-06,
+      "loss": 1.1823,
+      "step": 23060
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.906474285714286e-06,
+      "loss": 1.2563,
+      "step": 23061
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.906205714285714e-06,
+      "loss": 1.6057,
+      "step": 23062
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.905937142857144e-06,
+      "loss": 1.0798,
+      "step": 23063
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90566857142857e-06,
+      "loss": 1.0732,
+      "step": 23064
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.9054e-06,
+      "loss": 1.1667,
+      "step": 23065
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.905131428571429e-06,
+      "loss": 1.0802,
+      "step": 23066
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.904862857142859e-06,
+      "loss": 1.1071,
+      "step": 23067
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.904594285714285e-06,
+      "loss": 1.2234,
+      "step": 23068
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.904325714285715e-06,
+      "loss": 1.0622,
+      "step": 23069
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.904057142857143e-06,
+      "loss": 1.3016,
+      "step": 23070
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.903788571428571e-06,
+      "loss": 1.1882,
+      "step": 23071
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90352e-06,
+      "loss": 1.0571,
+      "step": 23072
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90325142857143e-06,
+      "loss": 0.7145,
+      "step": 23073
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.902982857142858e-06,
+      "loss": 0.9847,
+      "step": 23074
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.902714285714286e-06,
+      "loss": 0.8086,
+      "step": 23075
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.902445714285716e-06,
+      "loss": 1.2681,
+      "step": 23076
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.902177142857144e-06,
+      "loss": 1.1405,
+      "step": 23077
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.901908571428572e-06,
+      "loss": 1.0108,
+      "step": 23078
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90164e-06,
+      "loss": 1.4894,
+      "step": 23079
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.90137142857143e-06,
+      "loss": 1.3616,
+      "step": 23080
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.901102857142856e-06,
+      "loss": 1.183,
+      "step": 23081
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.900834285714286e-06,
+      "loss": 1.2013,
+      "step": 23082
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.900565714285715e-06,
+      "loss": 0.6211,
+      "step": 23083
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.900297142857143e-06,
+      "loss": 1.0507,
+      "step": 23084
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.900028571428571e-06,
+      "loss": 1.2723,
+      "step": 23085
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.89976e-06,
+      "loss": 0.9864,
+      "step": 23086
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.899491428571429e-06,
+      "loss": 1.1932,
+      "step": 23087
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.899222857142857e-06,
+      "loss": 1.1965,
+      "step": 23088
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.898954285714287e-06,
+      "loss": 0.7262,
+      "step": 23089
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.898685714285715e-06,
+      "loss": 1.2265,
+      "step": 23090
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.898417142857143e-06,
+      "loss": 1.2207,
+      "step": 23091
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.898148571428571e-06,
+      "loss": 1.3208,
+      "step": 23092
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.897880000000001e-06,
+      "loss": 1.0328,
+      "step": 23093
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.897611428571428e-06,
+      "loss": 1.3934,
+      "step": 23094
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.897342857142858e-06,
+      "loss": 1.1575,
+      "step": 23095
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.897074285714286e-06,
+      "loss": 0.7538,
+      "step": 23096
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.896805714285714e-06,
+      "loss": 1.4044,
+      "step": 23097
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.896537142857142e-06,
+      "loss": 0.944,
+      "step": 23098
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.896268571428572e-06,
+      "loss": 1.4583,
+      "step": 23099
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.896e-06,
+      "loss": 1.4717,
+      "step": 23100
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.895731428571428e-06,
+      "loss": 1.0157,
+      "step": 23101
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.895462857142858e-06,
+      "loss": 1.1064,
+      "step": 23102
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.895194285714287e-06,
+      "loss": 1.2661,
+      "step": 23103
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.894925714285715e-06,
+      "loss": 0.8329,
+      "step": 23104
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.894657142857143e-06,
+      "loss": 1.1844,
+      "step": 23105
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.894388571428573e-06,
+      "loss": 0.9321,
+      "step": 23106
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.89412e-06,
+      "loss": 1.2296,
+      "step": 23107
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.893851428571429e-06,
+      "loss": 1.141,
+      "step": 23108
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.893582857142857e-06,
+      "loss": 1.0687,
+      "step": 23109
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.893314285714285e-06,
+      "loss": 1.1742,
+      "step": 23110
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.893045714285714e-06,
+      "loss": 1.3686,
+      "step": 23111
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.892777142857143e-06,
+      "loss": 1.3523,
+      "step": 23112
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.892508571428572e-06,
+      "loss": 1.3319,
+      "step": 23113
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.89224e-06,
+      "loss": 1.0782,
+      "step": 23114
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.89197142857143e-06,
+      "loss": 0.9019,
+      "step": 23115
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.891702857142858e-06,
+      "loss": 0.9337,
+      "step": 23116
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.891434285714286e-06,
+      "loss": 1.3079,
+      "step": 23117
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.891165714285714e-06,
+      "loss": 1.1564,
+      "step": 23118
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.890897142857144e-06,
+      "loss": 0.7469,
+      "step": 23119
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.89062857142857e-06,
+      "loss": 1.1676,
+      "step": 23120
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.89036e-06,
+      "loss": 1.0999,
+      "step": 23121
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.890091428571429e-06,
+      "loss": 1.3496,
+      "step": 23122
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.889822857142859e-06,
+      "loss": 1.2501,
+      "step": 23123
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.889554285714285e-06,
+      "loss": 1.1874,
+      "step": 23124
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.889285714285715e-06,
+      "loss": 1.0486,
+      "step": 23125
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.889017142857143e-06,
+      "loss": 1.5526,
+      "step": 23126
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.888748571428571e-06,
+      "loss": 1.1189,
+      "step": 23127
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.888480000000001e-06,
+      "loss": 1.1695,
+      "step": 23128
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.88821142857143e-06,
+      "loss": 0.939,
+      "step": 23129
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.887942857142857e-06,
+      "loss": 1.1387,
+      "step": 23130
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.887674285714286e-06,
+      "loss": 1.4082,
+      "step": 23131
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.887405714285715e-06,
+      "loss": 0.8716,
+      "step": 23132
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.887137142857142e-06,
+      "loss": 1.0817,
+      "step": 23133
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.886868571428572e-06,
+      "loss": 1.3288,
+      "step": 23134
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.8866e-06,
+      "loss": 1.3941,
+      "step": 23135
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.88633142857143e-06,
+      "loss": 0.9865,
+      "step": 23136
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.886062857142856e-06,
+      "loss": 0.7884,
+      "step": 23137
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.885794285714286e-06,
+      "loss": 0.8699,
+      "step": 23138
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.885525714285714e-06,
+      "loss": 1.2408,
+      "step": 23139
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.885257142857143e-06,
+      "loss": 1.4265,
+      "step": 23140
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.884988571428572e-06,
+      "loss": 1.2049,
+      "step": 23141
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.88472e-06,
+      "loss": 1.311,
+      "step": 23142
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.884451428571429e-06,
+      "loss": 1.493,
+      "step": 23143
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.884182857142857e-06,
+      "loss": 1.0533,
+      "step": 23144
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.883914285714287e-06,
+      "loss": 1.1795,
+      "step": 23145
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.883645714285715e-06,
+      "loss": 1.2745,
+      "step": 23146
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.883377142857143e-06,
+      "loss": 1.1121,
+      "step": 23147
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.883108571428571e-06,
+      "loss": 1.0998,
+      "step": 23148
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.882840000000001e-06,
+      "loss": 0.8538,
+      "step": 23149
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.882571428571428e-06,
+      "loss": 1.2926,
+      "step": 23150
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.882302857142858e-06,
+      "loss": 0.8232,
+      "step": 23151
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.882034285714286e-06,
+      "loss": 1.2519,
+      "step": 23152
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.881765714285714e-06,
+      "loss": 0.9619,
+      "step": 23153
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.881497142857144e-06,
+      "loss": 0.765,
+      "step": 23154
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.881228571428572e-06,
+      "loss": 1.2076,
+      "step": 23155
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.88096e-06,
+      "loss": 1.0176,
+      "step": 23156
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.880691428571428e-06,
+      "loss": 0.9403,
+      "step": 23157
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.880422857142858e-06,
+      "loss": 1.3754,
+      "step": 23158
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.880154285714286e-06,
+      "loss": 1.1641,
+      "step": 23159
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.879885714285715e-06,
+      "loss": 1.1245,
+      "step": 23160
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.879617142857143e-06,
+      "loss": 0.9509,
+      "step": 23161
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.879348571428573e-06,
+      "loss": 1.1228,
+      "step": 23162
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.879079999999999e-06,
+      "loss": 1.0163,
+      "step": 23163
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.878811428571429e-06,
+      "loss": 1.0061,
+      "step": 23164
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.878542857142857e-06,
+      "loss": 1.4343,
+      "step": 23165
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.878274285714285e-06,
+      "loss": 1.4328,
+      "step": 23166
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.878005714285715e-06,
+      "loss": 1.5913,
+      "step": 23167
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.877737142857143e-06,
+      "loss": 1.2579,
+      "step": 23168
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.877468571428573e-06,
+      "loss": 1.2412,
+      "step": 23169
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.8772e-06,
+      "loss": 1.0609,
+      "step": 23170
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.87693142857143e-06,
+      "loss": 1.0789,
+      "step": 23171
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.876662857142858e-06,
+      "loss": 1.195,
+      "step": 23172
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.876394285714286e-06,
+      "loss": 1.3597,
+      "step": 23173
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.876125714285714e-06,
+      "loss": 1.3407,
+      "step": 23174
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.875857142857144e-06,
+      "loss": 1.5144,
+      "step": 23175
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.87558857142857e-06,
+      "loss": 0.8381,
+      "step": 23176
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.87532e-06,
+      "loss": 1.2382,
+      "step": 23177
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.875051428571429e-06,
+      "loss": 1.0254,
+      "step": 23178
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.874782857142857e-06,
+      "loss": 1.527,
+      "step": 23179
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.874514285714287e-06,
+      "loss": 1.1313,
+      "step": 23180
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.874245714285715e-06,
+      "loss": 1.1038,
+      "step": 23181
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.873977142857145e-06,
+      "loss": 0.7952,
+      "step": 23182
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.873708571428571e-06,
+      "loss": 1.3541,
+      "step": 23183
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.873440000000001e-06,
+      "loss": 1.1953,
+      "step": 23184
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.873171428571429e-06,
+      "loss": 1.2737,
+      "step": 23185
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.872902857142857e-06,
+      "loss": 1.5047,
+      "step": 23186
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.872634285714286e-06,
+      "loss": 1.4647,
+      "step": 23187
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.872365714285715e-06,
+      "loss": 1.069,
+      "step": 23188
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.872097142857142e-06,
+      "loss": 1.4753,
+      "step": 23189
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.871828571428572e-06,
+      "loss": 1.2942,
+      "step": 23190
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.87156e-06,
+      "loss": 0.8968,
+      "step": 23191
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.871291428571428e-06,
+      "loss": 1.246,
+      "step": 23192
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.871022857142858e-06,
+      "loss": 1.4242,
+      "step": 23193
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.870754285714286e-06,
+      "loss": 1.4681,
+      "step": 23194
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.870485714285716e-06,
+      "loss": 1.0906,
+      "step": 23195
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.870217142857142e-06,
+      "loss": 1.2878,
+      "step": 23196
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.869948571428572e-06,
+      "loss": 1.1753,
+      "step": 23197
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86968e-06,
+      "loss": 1.3756,
+      "step": 23198
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.869411428571429e-06,
+      "loss": 0.7394,
+      "step": 23199
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.869142857142857e-06,
+      "loss": 0.709,
+      "step": 23200
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.868874285714287e-06,
+      "loss": 1.2338,
+      "step": 23201
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.868605714285713e-06,
+      "loss": 1.3638,
+      "step": 23202
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.868337142857143e-06,
+      "loss": 1.5663,
+      "step": 23203
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.868068571428571e-06,
+      "loss": 1.0663,
+      "step": 23204
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.867800000000001e-06,
+      "loss": 1.4564,
+      "step": 23205
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86753142857143e-06,
+      "loss": 0.9047,
+      "step": 23206
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.867262857142857e-06,
+      "loss": 0.6174,
+      "step": 23207
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.866994285714287e-06,
+      "loss": 1.2133,
+      "step": 23208
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.866725714285714e-06,
+      "loss": 1.2881,
+      "step": 23209
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.866457142857144e-06,
+      "loss": 1.1092,
+      "step": 23210
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.866188571428572e-06,
+      "loss": 1.1609,
+      "step": 23211
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86592e-06,
+      "loss": 0.7722,
+      "step": 23212
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.865651428571428e-06,
+      "loss": 1.119,
+      "step": 23213
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.865382857142858e-06,
+      "loss": 1.1754,
+      "step": 23214
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.865114285714285e-06,
+      "loss": 1.0962,
+      "step": 23215
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.864845714285714e-06,
+      "loss": 1.4011,
+      "step": 23216
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.864577142857143e-06,
+      "loss": 1.0703,
+      "step": 23217
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.864308571428573e-06,
+      "loss": 1.0446,
+      "step": 23218
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86404e-06,
+      "loss": 1.4339,
+      "step": 23219
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.863771428571429e-06,
+      "loss": 1.3128,
+      "step": 23220
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.863502857142859e-06,
+      "loss": 0.9273,
+      "step": 23221
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.863234285714285e-06,
+      "loss": 1.1256,
+      "step": 23222
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.862965714285715e-06,
+      "loss": 1.3968,
+      "step": 23223
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.862697142857143e-06,
+      "loss": 1.0694,
+      "step": 23224
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.862428571428571e-06,
+      "loss": 1.1144,
+      "step": 23225
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86216e-06,
+      "loss": 0.955,
+      "step": 23226
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86189142857143e-06,
+      "loss": 1.2375,
+      "step": 23227
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.861622857142858e-06,
+      "loss": 1.4069,
+      "step": 23228
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.861354285714286e-06,
+      "loss": 1.4977,
+      "step": 23229
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.861085714285714e-06,
+      "loss": 1.094,
+      "step": 23230
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.860817142857144e-06,
+      "loss": 0.9719,
+      "step": 23231
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.860548571428572e-06,
+      "loss": 1.1844,
+      "step": 23232
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86028e-06,
+      "loss": 1.0617,
+      "step": 23233
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.86001142857143e-06,
+      "loss": 0.9965,
+      "step": 23234
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.859742857142857e-06,
+      "loss": 1.1516,
+      "step": 23235
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.859474285714286e-06,
+      "loss": 1.2512,
+      "step": 23236
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.859205714285715e-06,
+      "loss": 0.8946,
+      "step": 23237
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.858937142857143e-06,
+      "loss": 1.2252,
+      "step": 23238
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.858668571428571e-06,
+      "loss": 0.7896,
+      "step": 23239
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.858400000000001e-06,
+      "loss": 1.0082,
+      "step": 23240
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.858131428571429e-06,
+      "loss": 1.0917,
+      "step": 23241
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.857862857142857e-06,
+      "loss": 1.1939,
+      "step": 23242
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.857594285714285e-06,
+      "loss": 1.0913,
+      "step": 23243
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.857325714285715e-06,
+      "loss": 1.4993,
+      "step": 23244
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.857057142857143e-06,
+      "loss": 1.3456,
+      "step": 23245
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.856788571428572e-06,
+      "loss": 0.9701,
+      "step": 23246
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.856520000000001e-06,
+      "loss": 1.3387,
+      "step": 23247
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.856251428571428e-06,
+      "loss": 1.1883,
+      "step": 23248
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.855982857142858e-06,
+      "loss": 1.3952,
+      "step": 23249
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.855714285714286e-06,
+      "loss": 1.2193,
+      "step": 23250
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.855445714285716e-06,
+      "loss": 1.3754,
+      "step": 23251
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.855177142857142e-06,
+      "loss": 1.0335,
+      "step": 23252
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.854908571428572e-06,
+      "loss": 1.0298,
+      "step": 23253
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.85464e-06,
+      "loss": 1.2987,
+      "step": 23254
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.854371428571429e-06,
+      "loss": 0.9374,
+      "step": 23255
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.854102857142857e-06,
+      "loss": 1.2987,
+      "step": 23256
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.853834285714287e-06,
+      "loss": 1.7377,
+      "step": 23257
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.853565714285715e-06,
+      "loss": 1.4748,
+      "step": 23258
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.853297142857143e-06,
+      "loss": 1.0427,
+      "step": 23259
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.853028571428573e-06,
+      "loss": 1.1824,
+      "step": 23260
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.85276e-06,
+      "loss": 1.256,
+      "step": 23261
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.85249142857143e-06,
+      "loss": 1.2284,
+      "step": 23262
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.852222857142857e-06,
+      "loss": 0.9261,
+      "step": 23263
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.851954285714287e-06,
+      "loss": 1.1642,
+      "step": 23264
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.851685714285714e-06,
+      "loss": 1.0453,
+      "step": 23265
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.851417142857144e-06,
+      "loss": 1.1107,
+      "step": 23266
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.851148571428572e-06,
+      "loss": 1.2087,
+      "step": 23267
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.85088e-06,
+      "loss": 1.2358,
+      "step": 23268
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.850611428571428e-06,
+      "loss": 1.2329,
+      "step": 23269
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.850342857142858e-06,
+      "loss": 1.1867,
+      "step": 23270
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.850074285714286e-06,
+      "loss": 0.8335,
+      "step": 23271
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.849805714285714e-06,
+      "loss": 1.2069,
+      "step": 23272
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.849537142857144e-06,
+      "loss": 1.2416,
+      "step": 23273
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.849268571428572e-06,
+      "loss": 0.8729,
+      "step": 23274
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.849e-06,
+      "loss": 0.8238,
+      "step": 23275
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.848731428571429e-06,
+      "loss": 1.2026,
+      "step": 23276
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.848462857142859e-06,
+      "loss": 1.2366,
+      "step": 23277
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.848194285714285e-06,
+      "loss": 0.9942,
+      "step": 23278
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.847925714285715e-06,
+      "loss": 1.5329,
+      "step": 23279
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.847657142857143e-06,
+      "loss": 1.2972,
+      "step": 23280
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.847388571428571e-06,
+      "loss": 1.1878,
+      "step": 23281
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.84712e-06,
+      "loss": 1.3814,
+      "step": 23282
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.84685142857143e-06,
+      "loss": 1.4828,
+      "step": 23283
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.846582857142856e-06,
+      "loss": 1.2601,
+      "step": 23284
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.846314285714286e-06,
+      "loss": 1.592,
+      "step": 23285
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.846045714285716e-06,
+      "loss": 1.5264,
+      "step": 23286
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.845777142857144e-06,
+      "loss": 1.1085,
+      "step": 23287
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.845508571428572e-06,
+      "loss": 1.0806,
+      "step": 23288
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.84524e-06,
+      "loss": 1.4919,
+      "step": 23289
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.84497142857143e-06,
+      "loss": 1.2876,
+      "step": 23290
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.844702857142856e-06,
+      "loss": 1.1272,
+      "step": 23291
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.844434285714286e-06,
+      "loss": 1.3649,
+      "step": 23292
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.844165714285715e-06,
+      "loss": 1.2183,
+      "step": 23293
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.843897142857143e-06,
+      "loss": 0.9771,
+      "step": 23294
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.843628571428571e-06,
+      "loss": 1.3688,
+      "step": 23295
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.84336e-06,
+      "loss": 1.0238,
+      "step": 23296
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.843091428571429e-06,
+      "loss": 1.0112,
+      "step": 23297
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.842822857142857e-06,
+      "loss": 0.9994,
+      "step": 23298
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.842554285714287e-06,
+      "loss": 0.89,
+      "step": 23299
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.842285714285715e-06,
+      "loss": 1.6477,
+      "step": 23300
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.842017142857143e-06,
+      "loss": 1.3906,
+      "step": 23301
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.841748571428571e-06,
+      "loss": 1.0161,
+      "step": 23302
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.841480000000001e-06,
+      "loss": 1.0152,
+      "step": 23303
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.841211428571428e-06,
+      "loss": 1.0374,
+      "step": 23304
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.840942857142858e-06,
+      "loss": 1.231,
+      "step": 23305
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.840674285714286e-06,
+      "loss": 1.014,
+      "step": 23306
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.840405714285714e-06,
+      "loss": 0.7914,
+      "step": 23307
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.840137142857142e-06,
+      "loss": 0.7281,
+      "step": 23308
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.839868571428572e-06,
+      "loss": 1.0701,
+      "step": 23309
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.8396e-06,
+      "loss": 0.9468,
+      "step": 23310
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.839331428571428e-06,
+      "loss": 1.0332,
+      "step": 23311
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.839062857142858e-06,
+      "loss": 0.8523,
+      "step": 23312
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.838794285714287e-06,
+      "loss": 0.9493,
+      "step": 23313
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.838525714285715e-06,
+      "loss": 1.4218,
+      "step": 23314
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.838257142857143e-06,
+      "loss": 1.0568,
+      "step": 23315
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.837988571428573e-06,
+      "loss": 1.0151,
+      "step": 23316
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.83772e-06,
+      "loss": 1.0618,
+      "step": 23317
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.837451428571429e-06,
+      "loss": 1.1125,
+      "step": 23318
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.837182857142857e-06,
+      "loss": 1.0383,
+      "step": 23319
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.836914285714287e-06,
+      "loss": 0.9112,
+      "step": 23320
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.836645714285714e-06,
+      "loss": 1.2556,
+      "step": 23321
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.836377142857143e-06,
+      "loss": 1.2805,
+      "step": 23322
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.836108571428572e-06,
+      "loss": 1.2219,
+      "step": 23323
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.83584e-06,
+      "loss": 0.9275,
+      "step": 23324
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.83557142857143e-06,
+      "loss": 1.022,
+      "step": 23325
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.835302857142858e-06,
+      "loss": 0.9109,
+      "step": 23326
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.835034285714286e-06,
+      "loss": 1.0455,
+      "step": 23327
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.834765714285714e-06,
+      "loss": 1.107,
+      "step": 23328
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.834497142857144e-06,
+      "loss": 1.0812,
+      "step": 23329
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.83422857142857e-06,
+      "loss": 1.3615,
+      "step": 23330
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.83396e-06,
+      "loss": 1.1157,
+      "step": 23331
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.833691428571429e-06,
+      "loss": 1.1522,
+      "step": 23332
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.833422857142859e-06,
+      "loss": 1.3158,
+      "step": 23333
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.833154285714285e-06,
+      "loss": 0.9547,
+      "step": 23334
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.832885714285715e-06,
+      "loss": 1.2258,
+      "step": 23335
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.832617142857143e-06,
+      "loss": 1.0132,
+      "step": 23336
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.832348571428571e-06,
+      "loss": 1.428,
+      "step": 23337
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.832080000000001e-06,
+      "loss": 1.1604,
+      "step": 23338
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.83181142857143e-06,
+      "loss": 1.2812,
+      "step": 23339
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.831542857142857e-06,
+      "loss": 1.4065,
+      "step": 23340
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.831274285714286e-06,
+      "loss": 1.513,
+      "step": 23341
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.831005714285715e-06,
+      "loss": 1.3864,
+      "step": 23342
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.830737142857142e-06,
+      "loss": 1.461,
+      "step": 23343
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.830468571428572e-06,
+      "loss": 1.0329,
+      "step": 23344
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.8302e-06,
+      "loss": 0.8391,
+      "step": 23345
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.82993142857143e-06,
+      "loss": 0.9138,
+      "step": 23346
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.829662857142856e-06,
+      "loss": 1.1923,
+      "step": 23347
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.829394285714286e-06,
+      "loss": 1.1387,
+      "step": 23348
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.829125714285714e-06,
+      "loss": 1.1513,
+      "step": 23349
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.828857142857143e-06,
+      "loss": 0.8916,
+      "step": 23350
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.828588571428572e-06,
+      "loss": 0.9771,
+      "step": 23351
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.82832e-06,
+      "loss": 1.215,
+      "step": 23352
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.828051428571429e-06,
+      "loss": 0.9879,
+      "step": 23353
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.827782857142857e-06,
+      "loss": 0.935,
+      "step": 23354
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.827514285714287e-06,
+      "loss": 1.1384,
+      "step": 23355
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.827245714285715e-06,
+      "loss": 1.1726,
+      "step": 23356
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.826977142857143e-06,
+      "loss": 1.284,
+      "step": 23357
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.826708571428571e-06,
+      "loss": 1.4431,
+      "step": 23358
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.826440000000001e-06,
+      "loss": 0.9004,
+      "step": 23359
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.826171428571428e-06,
+      "loss": 1.3097,
+      "step": 23360
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.825902857142858e-06,
+      "loss": 1.104,
+      "step": 23361
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 7.825634285714286e-06,
+      "loss": 1.3995,
+      "step": 23362
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.825365714285714e-06,
+      "loss": 1.1677,
+      "step": 23363
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.825097142857144e-06,
+      "loss": 0.9551,
+      "step": 23364
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.824828571428572e-06,
+      "loss": 1.3963,
+      "step": 23365
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.82456e-06,
+      "loss": 1.1467,
+      "step": 23366
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.824291428571428e-06,
+      "loss": 1.2688,
+      "step": 23367
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.824022857142858e-06,
+      "loss": 1.312,
+      "step": 23368
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.823754285714286e-06,
+      "loss": 0.843,
+      "step": 23369
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.823485714285715e-06,
+      "loss": 0.986,
+      "step": 23370
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.823217142857143e-06,
+      "loss": 0.9263,
+      "step": 23371
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.822948571428573e-06,
+      "loss": 1.2806,
+      "step": 23372
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.822679999999999e-06,
+      "loss": 0.9357,
+      "step": 23373
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.822411428571429e-06,
+      "loss": 1.1539,
+      "step": 23374
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.822142857142857e-06,
+      "loss": 0.9145,
+      "step": 23375
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.821874285714285e-06,
+      "loss": 1.1876,
+      "step": 23376
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.821605714285715e-06,
+      "loss": 0.9327,
+      "step": 23377
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.821337142857143e-06,
+      "loss": 1.4142,
+      "step": 23378
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.821068571428573e-06,
+      "loss": 1.2183,
+      "step": 23379
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.8208e-06,
+      "loss": 0.9015,
+      "step": 23380
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.82053142857143e-06,
+      "loss": 0.8554,
+      "step": 23381
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.820262857142858e-06,
+      "loss": 1.7525,
+      "step": 23382
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.819994285714286e-06,
+      "loss": 1.373,
+      "step": 23383
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.819725714285714e-06,
+      "loss": 1.1144,
+      "step": 23384
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.819457142857144e-06,
+      "loss": 1.0781,
+      "step": 23385
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.81918857142857e-06,
+      "loss": 0.7701,
+      "step": 23386
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.81892e-06,
+      "loss": 1.1662,
+      "step": 23387
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.818651428571429e-06,
+      "loss": 1.3226,
+      "step": 23388
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.818382857142857e-06,
+      "loss": 1.2364,
+      "step": 23389
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.818114285714285e-06,
+      "loss": 1.5109,
+      "step": 23390
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.817845714285715e-06,
+      "loss": 0.9671,
+      "step": 23391
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.817577142857145e-06,
+      "loss": 0.678,
+      "step": 23392
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.817308571428571e-06,
+      "loss": 1.2602,
+      "step": 23393
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.817040000000001e-06,
+      "loss": 1.1861,
+      "step": 23394
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.816771428571429e-06,
+      "loss": 1.0854,
+      "step": 23395
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.816502857142857e-06,
+      "loss": 1.4154,
+      "step": 23396
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.816234285714285e-06,
+      "loss": 1.6842,
+      "step": 23397
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.815965714285715e-06,
+      "loss": 0.9912,
+      "step": 23398
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.815697142857142e-06,
+      "loss": 1.3688,
+      "step": 23399
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.815428571428572e-06,
+      "loss": 0.9445,
+      "step": 23400
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.81516e-06,
+      "loss": 1.1063,
+      "step": 23401
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.81489142857143e-06,
+      "loss": 1.2583,
+      "step": 23402
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.814622857142856e-06,
+      "loss": 1.5176,
+      "step": 23403
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.814354285714286e-06,
+      "loss": 1.4259,
+      "step": 23404
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.814085714285716e-06,
+      "loss": 1.1166,
+      "step": 23405
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.813817142857142e-06,
+      "loss": 0.9145,
+      "step": 23406
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.813548571428572e-06,
+      "loss": 1.0417,
+      "step": 23407
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.81328e-06,
+      "loss": 1.6199,
+      "step": 23408
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.813011428571429e-06,
+      "loss": 1.2869,
+      "step": 23409
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.812742857142857e-06,
+      "loss": 1.1232,
+      "step": 23410
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.812474285714287e-06,
+      "loss": 1.1097,
+      "step": 23411
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.812205714285713e-06,
+      "loss": 1.2514,
+      "step": 23412
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.811937142857143e-06,
+      "loss": 1.1748,
+      "step": 23413
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.811668571428571e-06,
+      "loss": 1.2869,
+      "step": 23414
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.811400000000001e-06,
+      "loss": 1.1842,
+      "step": 23415
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.811131428571428e-06,
+      "loss": 1.0848,
+      "step": 23416
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.810862857142857e-06,
+      "loss": 1.2684,
+      "step": 23417
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.810594285714287e-06,
+      "loss": 1.491,
+      "step": 23418
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.810325714285714e-06,
+      "loss": 1.7211,
+      "step": 23419
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.810057142857144e-06,
+      "loss": 1.234,
+      "step": 23420
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.809788571428572e-06,
+      "loss": 0.8833,
+      "step": 23421
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.80952e-06,
+      "loss": 1.1175,
+      "step": 23422
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.809251428571428e-06,
+      "loss": 1.0486,
+      "step": 23423
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.808982857142858e-06,
+      "loss": 0.8472,
+      "step": 23424
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.808714285714286e-06,
+      "loss": 0.9853,
+      "step": 23425
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.808445714285714e-06,
+      "loss": 0.996,
+      "step": 23426
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.808177142857143e-06,
+      "loss": 1.4494,
+      "step": 23427
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.807908571428573e-06,
+      "loss": 1.1937,
+      "step": 23428
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.807639999999999e-06,
+      "loss": 0.917,
+      "step": 23429
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.807371428571429e-06,
+      "loss": 0.9159,
+      "step": 23430
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.807102857142859e-06,
+      "loss": 1.2432,
+      "step": 23431
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.806834285714285e-06,
+      "loss": 1.031,
+      "step": 23432
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.806565714285715e-06,
+      "loss": 1.2123,
+      "step": 23433
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.806297142857143e-06,
+      "loss": 1.0238,
+      "step": 23434
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.806028571428571e-06,
+      "loss": 1.0336,
+      "step": 23435
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.80576e-06,
+      "loss": 0.8544,
+      "step": 23436
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.80549142857143e-06,
+      "loss": 1.2429,
+      "step": 23437
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.805222857142858e-06,
+      "loss": 1.2433,
+      "step": 23438
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.804954285714286e-06,
+      "loss": 1.4016,
+      "step": 23439
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.804685714285714e-06,
+      "loss": 0.4904,
+      "step": 23440
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.804417142857144e-06,
+      "loss": 0.8372,
+      "step": 23441
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.80414857142857e-06,
+      "loss": 1.656,
+      "step": 23442
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.80388e-06,
+      "loss": 1.0462,
+      "step": 23443
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.80361142857143e-06,
+      "loss": 1.2677,
+      "step": 23444
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.803342857142857e-06,
+      "loss": 1.2962,
+      "step": 23445
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.803074285714286e-06,
+      "loss": 1.271,
+      "step": 23446
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.802805714285715e-06,
+      "loss": 1.0448,
+      "step": 23447
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.802537142857145e-06,
+      "loss": 1.2708,
+      "step": 23448
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.802268571428571e-06,
+      "loss": 0.6677,
+      "step": 23449
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.802000000000001e-06,
+      "loss": 0.945,
+      "step": 23450
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.801731428571429e-06,
+      "loss": 1.0612,
+      "step": 23451
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.801462857142857e-06,
+      "loss": 1.2421,
+      "step": 23452
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.801194285714285e-06,
+      "loss": 1.231,
+      "step": 23453
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.800925714285715e-06,
+      "loss": 1.7932,
+      "step": 23454
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.800657142857142e-06,
+      "loss": 1.7159,
+      "step": 23455
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.800388571428572e-06,
+      "loss": 1.082,
+      "step": 23456
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.800120000000001e-06,
+      "loss": 1.1656,
+      "step": 23457
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.799851428571428e-06,
+      "loss": 1.2034,
+      "step": 23458
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.799582857142858e-06,
+      "loss": 1.2535,
+      "step": 23459
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.799314285714286e-06,
+      "loss": 1.0302,
+      "step": 23460
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.799045714285716e-06,
+      "loss": 1.1147,
+      "step": 23461
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.798777142857142e-06,
+      "loss": 1.2774,
+      "step": 23462
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.798508571428572e-06,
+      "loss": 0.8797,
+      "step": 23463
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.79824e-06,
+      "loss": 1.3382,
+      "step": 23464
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.797971428571429e-06,
+      "loss": 1.267,
+      "step": 23465
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.797702857142857e-06,
+      "loss": 0.8616,
+      "step": 23466
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.797434285714287e-06,
+      "loss": 1.3355,
+      "step": 23467
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.797165714285713e-06,
+      "loss": 0.844,
+      "step": 23468
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.796897142857143e-06,
+      "loss": 1.3706,
+      "step": 23469
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.796628571428573e-06,
+      "loss": 1.3057,
+      "step": 23470
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.796360000000001e-06,
+      "loss": 0.8185,
+      "step": 23471
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.79609142857143e-06,
+      "loss": 0.7331,
+      "step": 23472
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.795822857142857e-06,
+      "loss": 1.1849,
+      "step": 23473
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.795554285714287e-06,
+      "loss": 1.0619,
+      "step": 23474
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.795285714285714e-06,
+      "loss": 1.1743,
+      "step": 23475
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.795017142857144e-06,
+      "loss": 0.754,
+      "step": 23476
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.794748571428572e-06,
+      "loss": 0.9228,
+      "step": 23477
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.79448e-06,
+      "loss": 0.8911,
+      "step": 23478
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.794211428571428e-06,
+      "loss": 1.3078,
+      "step": 23479
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.793942857142858e-06,
+      "loss": 0.9793,
+      "step": 23480
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.793674285714284e-06,
+      "loss": 1.0901,
+      "step": 23481
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.793405714285714e-06,
+      "loss": 0.9363,
+      "step": 23482
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.793137142857144e-06,
+      "loss": 1.5572,
+      "step": 23483
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.792868571428572e-06,
+      "loss": 1.3872,
+      "step": 23484
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.7926e-06,
+      "loss": 1.2534,
+      "step": 23485
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.792331428571429e-06,
+      "loss": 1.3551,
+      "step": 23486
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.792062857142859e-06,
+      "loss": 1.2746,
+      "step": 23487
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.791794285714285e-06,
+      "loss": 1.4491,
+      "step": 23488
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.791525714285715e-06,
+      "loss": 1.0281,
+      "step": 23489
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.791257142857143e-06,
+      "loss": 0.8521,
+      "step": 23490
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.790988571428571e-06,
+      "loss": 1.1305,
+      "step": 23491
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.79072e-06,
+      "loss": 1.3459,
+      "step": 23492
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.79045142857143e-06,
+      "loss": 0.947,
+      "step": 23493
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.790182857142858e-06,
+      "loss": 1.5729,
+      "step": 23494
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.789914285714286e-06,
+      "loss": 1.0532,
+      "step": 23495
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.789645714285714e-06,
+      "loss": 0.9791,
+      "step": 23496
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.789377142857144e-06,
+      "loss": 0.8989,
+      "step": 23497
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.789108571428572e-06,
+      "loss": 1.057,
+      "step": 23498
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.78884e-06,
+      "loss": 1.1946,
+      "step": 23499
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.78857142857143e-06,
+      "loss": 0.8737,
+      "step": 23500
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.788302857142856e-06,
+      "loss": 1.7456,
+      "step": 23501
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.788034285714286e-06,
+      "loss": 1.4465,
+      "step": 23502
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.787765714285715e-06,
+      "loss": 1.436,
+      "step": 23503
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.787497142857143e-06,
+      "loss": 1.2124,
+      "step": 23504
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.787228571428571e-06,
+      "loss": 1.2409,
+      "step": 23505
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.78696e-06,
+      "loss": 1.4632,
+      "step": 23506
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.786691428571429e-06,
+      "loss": 0.8787,
+      "step": 23507
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.786422857142857e-06,
+      "loss": 1.1025,
+      "step": 23508
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.786154285714285e-06,
+      "loss": 1.2988,
+      "step": 23509
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.785885714285715e-06,
+      "loss": 1.3037,
+      "step": 23510
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.785617142857143e-06,
+      "loss": 0.8732,
+      "step": 23511
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.785348571428571e-06,
+      "loss": 1.021,
+      "step": 23512
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.785080000000001e-06,
+      "loss": 0.83,
+      "step": 23513
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.784811428571428e-06,
+      "loss": 1.1506,
+      "step": 23514
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.784542857142858e-06,
+      "loss": 1.2559,
+      "step": 23515
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.784274285714286e-06,
+      "loss": 0.9941,
+      "step": 23516
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.784005714285714e-06,
+      "loss": 1.1109,
+      "step": 23517
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.783737142857142e-06,
+      "loss": 1.1228,
+      "step": 23518
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.783468571428572e-06,
+      "loss": 1.0212,
+      "step": 23519
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.7832e-06,
+      "loss": 1.1088,
+      "step": 23520
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.782931428571428e-06,
+      "loss": 0.9198,
+      "step": 23521
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.782662857142857e-06,
+      "loss": 1.3828,
+      "step": 23522
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.782394285714287e-06,
+      "loss": 1.4228,
+      "step": 23523
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.782125714285715e-06,
+      "loss": 0.8892,
+      "step": 23524
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.781857142857143e-06,
+      "loss": 1.1346,
+      "step": 23525
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.781588571428573e-06,
+      "loss": 1.6225,
+      "step": 23526
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.78132e-06,
+      "loss": 0.9352,
+      "step": 23527
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.781051428571429e-06,
+      "loss": 1.1396,
+      "step": 23528
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.780782857142857e-06,
+      "loss": 1.2374,
+      "step": 23529
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.780514285714287e-06,
+      "loss": 1.036,
+      "step": 23530
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.780245714285714e-06,
+      "loss": 1.2001,
+      "step": 23531
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.779977142857143e-06,
+      "loss": 1.2255,
+      "step": 23532
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.779708571428572e-06,
+      "loss": 1.0427,
+      "step": 23533
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.77944e-06,
+      "loss": 1.172,
+      "step": 23534
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.779171428571428e-06,
+      "loss": 1.2719,
+      "step": 23535
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.778902857142858e-06,
+      "loss": 0.8884,
+      "step": 23536
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.778634285714286e-06,
+      "loss": 1.2561,
+      "step": 23537
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.778365714285714e-06,
+      "loss": 1.0384,
+      "step": 23538
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.778097142857144e-06,
+      "loss": 1.0319,
+      "step": 23539
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.77782857142857e-06,
+      "loss": 1.0057,
+      "step": 23540
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.77756e-06,
+      "loss": 1.3447,
+      "step": 23541
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.777291428571429e-06,
+      "loss": 1.0714,
+      "step": 23542
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.777022857142859e-06,
+      "loss": 0.8676,
+      "step": 23543
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.776754285714285e-06,
+      "loss": 1.0584,
+      "step": 23544
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.776485714285715e-06,
+      "loss": 1.0481,
+      "step": 23545
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.776217142857143e-06,
+      "loss": 1.0392,
+      "step": 23546
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.775948571428571e-06,
+      "loss": 1.0869,
+      "step": 23547
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.77568e-06,
+      "loss": 1.0285,
+      "step": 23548
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.77541142857143e-06,
+      "loss": 0.7351,
+      "step": 23549
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.775142857142857e-06,
+      "loss": 1.2333,
+      "step": 23550
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.774874285714286e-06,
+      "loss": 1.2195,
+      "step": 23551
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.774605714285715e-06,
+      "loss": 1.4877,
+      "step": 23552
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.774337142857144e-06,
+      "loss": 1.3781,
+      "step": 23553
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.774068571428572e-06,
+      "loss": 0.9885,
+      "step": 23554
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.7738e-06,
+      "loss": 1.0303,
+      "step": 23555
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.77353142857143e-06,
+      "loss": 1.0694,
+      "step": 23556
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.773262857142856e-06,
+      "loss": 1.3865,
+      "step": 23557
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.772994285714286e-06,
+      "loss": 0.9803,
+      "step": 23558
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.772725714285714e-06,
+      "loss": 1.0615,
+      "step": 23559
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.772457142857143e-06,
+      "loss": 1.357,
+      "step": 23560
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.77218857142857e-06,
+      "loss": 1.3305,
+      "step": 23561
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.77192e-06,
+      "loss": 0.9988,
+      "step": 23562
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.771651428571429e-06,
+      "loss": 1.4705,
+      "step": 23563
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.771382857142857e-06,
+      "loss": 1.2696,
+      "step": 23564
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.771114285714287e-06,
+      "loss": 0.8485,
+      "step": 23565
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.770845714285715e-06,
+      "loss": 1.3787,
+      "step": 23566
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.770577142857143e-06,
+      "loss": 1.0397,
+      "step": 23567
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.770308571428571e-06,
+      "loss": 0.9827,
+      "step": 23568
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.770040000000001e-06,
+      "loss": 0.9426,
+      "step": 23569
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.769771428571428e-06,
+      "loss": 1.1689,
+      "step": 23570
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.769502857142858e-06,
+      "loss": 0.9404,
+      "step": 23571
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.769234285714286e-06,
+      "loss": 1.1858,
+      "step": 23572
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.768965714285714e-06,
+      "loss": 1.2346,
+      "step": 23573
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.768697142857142e-06,
+      "loss": 0.9528,
+      "step": 23574
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.768428571428572e-06,
+      "loss": 1.1253,
+      "step": 23575
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.768160000000002e-06,
+      "loss": 1.0511,
+      "step": 23576
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.767891428571428e-06,
+      "loss": 1.0371,
+      "step": 23577
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.767622857142858e-06,
+      "loss": 0.8009,
+      "step": 23578
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.767354285714286e-06,
+      "loss": 1.3242,
+      "step": 23579
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.767085714285715e-06,
+      "loss": 1.4995,
+      "step": 23580
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.766817142857143e-06,
+      "loss": 0.9462,
+      "step": 23581
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.766548571428573e-06,
+      "loss": 1.1145,
+      "step": 23582
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.766279999999999e-06,
+      "loss": 1.4755,
+      "step": 23583
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.766011428571429e-06,
+      "loss": 1.48,
+      "step": 23584
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.765742857142857e-06,
+      "loss": 1.1918,
+      "step": 23585
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.765474285714285e-06,
+      "loss": 1.1251,
+      "step": 23586
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.765205714285714e-06,
+      "loss": 1.3626,
+      "step": 23587
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.764937142857143e-06,
+      "loss": 1.3118,
+      "step": 23588
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.764668571428573e-06,
+      "loss": 1.3193,
+      "step": 23589
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.7644e-06,
+      "loss": 1.2981,
+      "step": 23590
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.76413142857143e-06,
+      "loss": 1.4927,
+      "step": 23591
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.763862857142858e-06,
+      "loss": 1.3499,
+      "step": 23592
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.763594285714286e-06,
+      "loss": 1.1731,
+      "step": 23593
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.763325714285714e-06,
+      "loss": 1.4301,
+      "step": 23594
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.763057142857144e-06,
+      "loss": 0.998,
+      "step": 23595
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.76278857142857e-06,
+      "loss": 1.3172,
+      "step": 23596
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.76252e-06,
+      "loss": 1.1005,
+      "step": 23597
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.762251428571429e-06,
+      "loss": 0.9614,
+      "step": 23598
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.761982857142858e-06,
+      "loss": 0.9519,
+      "step": 23599
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.761714285714285e-06,
+      "loss": 1.1472,
+      "step": 23600
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.761445714285715e-06,
+      "loss": 1.3368,
+      "step": 23601
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.761177142857143e-06,
+      "loss": 1.0633,
+      "step": 23602
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.760908571428571e-06,
+      "loss": 1.1063,
+      "step": 23603
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.760640000000001e-06,
+      "loss": 0.9949,
+      "step": 23604
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.760371428571429e-06,
+      "loss": 0.9035,
+      "step": 23605
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.760102857142857e-06,
+      "loss": 1.0466,
+      "step": 23606
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.759834285714285e-06,
+      "loss": 1.0692,
+      "step": 23607
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.759565714285715e-06,
+      "loss": 1.1465,
+      "step": 23608
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.759297142857142e-06,
+      "loss": 1.2502,
+      "step": 23609
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.759028571428572e-06,
+      "loss": 1.2154,
+      "step": 23610
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.75876e-06,
+      "loss": 1.0443,
+      "step": 23611
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.75849142857143e-06,
+      "loss": 1.034,
+      "step": 23612
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.758222857142856e-06,
+      "loss": 1.2978,
+      "step": 23613
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.757954285714286e-06,
+      "loss": 1.0384,
+      "step": 23614
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.757685714285714e-06,
+      "loss": 1.441,
+      "step": 23615
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.757417142857142e-06,
+      "loss": 1.1528,
+      "step": 23616
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.757148571428572e-06,
+      "loss": 1.1179,
+      "step": 23617
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.75688e-06,
+      "loss": 0.871,
+      "step": 23618
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.756611428571429e-06,
+      "loss": 1.261,
+      "step": 23619
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.756342857142857e-06,
+      "loss": 1.0442,
+      "step": 23620
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.756074285714287e-06,
+      "loss": 1.0917,
+      "step": 23621
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.755805714285715e-06,
+      "loss": 1.0432,
+      "step": 23622
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.755537142857143e-06,
+      "loss": 1.2229,
+      "step": 23623
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.755268571428571e-06,
+      "loss": 0.8039,
+      "step": 23624
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.755000000000001e-06,
+      "loss": 0.7312,
+      "step": 23625
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.754731428571428e-06,
+      "loss": 1.1649,
+      "step": 23626
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.754462857142857e-06,
+      "loss": 1.2802,
+      "step": 23627
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.754194285714286e-06,
+      "loss": 1.6459,
+      "step": 23628
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.753925714285714e-06,
+      "loss": 1.2697,
+      "step": 23629
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.753657142857144e-06,
+      "loss": 0.8728,
+      "step": 23630
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.753388571428572e-06,
+      "loss": 0.8513,
+      "step": 23631
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.75312e-06,
+      "loss": 0.7466,
+      "step": 23632
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.752851428571428e-06,
+      "loss": 1.3104,
+      "step": 23633
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.752582857142858e-06,
+      "loss": 0.8635,
+      "step": 23634
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.752314285714286e-06,
+      "loss": 1.341,
+      "step": 23635
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.752045714285714e-06,
+      "loss": 0.9826,
+      "step": 23636
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.751777142857143e-06,
+      "loss": 0.7351,
+      "step": 23637
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.751508571428573e-06,
+      "loss": 1.2898,
+      "step": 23638
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.751239999999999e-06,
+      "loss": 1.4885,
+      "step": 23639
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.750971428571429e-06,
+      "loss": 0.895,
+      "step": 23640
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.750702857142857e-06,
+      "loss": 0.7004,
+      "step": 23641
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.750434285714285e-06,
+      "loss": 1.1405,
+      "step": 23642
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.750165714285715e-06,
+      "loss": 1.022,
+      "step": 23643
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.749897142857143e-06,
+      "loss": 1.4942,
+      "step": 23644
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.749628571428573e-06,
+      "loss": 1.1218,
+      "step": 23645
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.74936e-06,
+      "loss": 1.2037,
+      "step": 23646
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.74909142857143e-06,
+      "loss": 0.919,
+      "step": 23647
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.748822857142858e-06,
+      "loss": 1.2995,
+      "step": 23648
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.748554285714286e-06,
+      "loss": 1.3046,
+      "step": 23649
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.748285714285714e-06,
+      "loss": 1.0289,
+      "step": 23650
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.748017142857144e-06,
+      "loss": 0.946,
+      "step": 23651
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.74774857142857e-06,
+      "loss": 1.7627,
+      "step": 23652
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.74748e-06,
+      "loss": 1.3012,
+      "step": 23653
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.747211428571428e-06,
+      "loss": 1.0555,
+      "step": 23654
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.746942857142857e-06,
+      "loss": 1.3744,
+      "step": 23655
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.746674285714286e-06,
+      "loss": 1.2581,
+      "step": 23656
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.746405714285715e-06,
+      "loss": 1.1568,
+      "step": 23657
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.746137142857145e-06,
+      "loss": 1.4936,
+      "step": 23658
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.745868571428571e-06,
+      "loss": 0.9121,
+      "step": 23659
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.745600000000001e-06,
+      "loss": 0.8012,
+      "step": 23660
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.745331428571429e-06,
+      "loss": 1.3043,
+      "step": 23661
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.745062857142857e-06,
+      "loss": 1.0496,
+      "step": 23662
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.744794285714285e-06,
+      "loss": 0.8788,
+      "step": 23663
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.744525714285715e-06,
+      "loss": 0.9303,
+      "step": 23664
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.744257142857142e-06,
+      "loss": 1.0932,
+      "step": 23665
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.743988571428572e-06,
+      "loss": 0.6634,
+      "step": 23666
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.74372e-06,
+      "loss": 1.2529,
+      "step": 23667
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.74345142857143e-06,
+      "loss": 1.2386,
+      "step": 23668
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.743182857142858e-06,
+      "loss": 0.9547,
+      "step": 23669
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.742914285714286e-06,
+      "loss": 0.9657,
+      "step": 23670
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.742645714285716e-06,
+      "loss": 1.0077,
+      "step": 23671
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.742377142857142e-06,
+      "loss": 1.8056,
+      "step": 23672
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.742108571428572e-06,
+      "loss": 1.1352,
+      "step": 23673
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.74184e-06,
+      "loss": 1.4655,
+      "step": 23674
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.741571428571429e-06,
+      "loss": 1.1579,
+      "step": 23675
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.741302857142857e-06,
+      "loss": 1.1231,
+      "step": 23676
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.741034285714287e-06,
+      "loss": 1.4628,
+      "step": 23677
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.740765714285713e-06,
+      "loss": 1.323,
+      "step": 23678
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.740497142857143e-06,
+      "loss": 1.3679,
+      "step": 23679
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.740228571428571e-06,
+      "loss": 1.2083,
+      "step": 23680
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.739960000000001e-06,
+      "loss": 1.2179,
+      "step": 23681
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.73969142857143e-06,
+      "loss": 1.1337,
+      "step": 23682
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.739422857142857e-06,
+      "loss": 0.8506,
+      "step": 23683
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.739154285714287e-06,
+      "loss": 1.0194,
+      "step": 23684
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.738885714285714e-06,
+      "loss": 0.9855,
+      "step": 23685
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.738617142857144e-06,
+      "loss": 1.0683,
+      "step": 23686
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.738348571428572e-06,
+      "loss": 1.1085,
+      "step": 23687
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.73808e-06,
+      "loss": 1.2548,
+      "step": 23688
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.737811428571428e-06,
+      "loss": 0.8625,
+      "step": 23689
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.737542857142858e-06,
+      "loss": 0.9909,
+      "step": 23690
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.737274285714284e-06,
+      "loss": 1.1884,
+      "step": 23691
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.737005714285714e-06,
+      "loss": 1.3503,
+      "step": 23692
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.736737142857143e-06,
+      "loss": 1.5603,
+      "step": 23693
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.736468571428572e-06,
+      "loss": 1.1749,
+      "step": 23694
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.7362e-06,
+      "loss": 1.2894,
+      "step": 23695
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.735931428571429e-06,
+      "loss": 1.4579,
+      "step": 23696
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.735662857142859e-06,
+      "loss": 0.9479,
+      "step": 23697
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.735394285714285e-06,
+      "loss": 1.3819,
+      "step": 23698
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.735125714285715e-06,
+      "loss": 1.0174,
+      "step": 23699
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.734857142857143e-06,
+      "loss": 1.4319,
+      "step": 23700
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.734588571428571e-06,
+      "loss": 1.0171,
+      "step": 23701
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.73432e-06,
+      "loss": 0.8267,
+      "step": 23702
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.73405142857143e-06,
+      "loss": 0.8643,
+      "step": 23703
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.733782857142858e-06,
+      "loss": 1.449,
+      "step": 23704
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.733514285714286e-06,
+      "loss": 1.2683,
+      "step": 23705
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.733245714285714e-06,
+      "loss": 1.008,
+      "step": 23706
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.732977142857144e-06,
+      "loss": 1.4665,
+      "step": 23707
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.732708571428572e-06,
+      "loss": 1.3345,
+      "step": 23708
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.73244e-06,
+      "loss": 1.4754,
+      "step": 23709
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.73217142857143e-06,
+      "loss": 1.4192,
+      "step": 23710
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.731902857142856e-06,
+      "loss": 1.5522,
+      "step": 23711
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 7.731634285714286e-06,
+      "loss": 1.242,
+      "step": 23712
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.731365714285715e-06,
+      "loss": 1.1671,
+      "step": 23713
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.731097142857143e-06,
+      "loss": 1.3317,
+      "step": 23714
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.730828571428571e-06,
+      "loss": 1.089,
+      "step": 23715
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.73056e-06,
+      "loss": 1.0879,
+      "step": 23716
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.730291428571429e-06,
+      "loss": 1.1659,
+      "step": 23717
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.730022857142857e-06,
+      "loss": 1.155,
+      "step": 23718
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.729754285714285e-06,
+      "loss": 1.1209,
+      "step": 23719
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.729485714285715e-06,
+      "loss": 0.9815,
+      "step": 23720
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.729217142857143e-06,
+      "loss": 0.874,
+      "step": 23721
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.728948571428571e-06,
+      "loss": 0.9826,
+      "step": 23722
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.728680000000001e-06,
+      "loss": 0.9228,
+      "step": 23723
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.728411428571428e-06,
+      "loss": 0.8137,
+      "step": 23724
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.728142857142858e-06,
+      "loss": 1.5817,
+      "step": 23725
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.727874285714286e-06,
+      "loss": 0.6194,
+      "step": 23726
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.727605714285716e-06,
+      "loss": 1.169,
+      "step": 23727
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.727337142857142e-06,
+      "loss": 1.2057,
+      "step": 23728
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.727068571428572e-06,
+      "loss": 1.042,
+      "step": 23729
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.7268e-06,
+      "loss": 1.2817,
+      "step": 23730
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.726531428571428e-06,
+      "loss": 0.9603,
+      "step": 23731
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.726262857142857e-06,
+      "loss": 1.4042,
+      "step": 23732
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.725994285714287e-06,
+      "loss": 0.9276,
+      "step": 23733
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.725725714285715e-06,
+      "loss": 0.9321,
+      "step": 23734
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.725457142857143e-06,
+      "loss": 1.33,
+      "step": 23735
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.725188571428573e-06,
+      "loss": 1.2077,
+      "step": 23736
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.72492e-06,
+      "loss": 1.448,
+      "step": 23737
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.724651428571429e-06,
+      "loss": 0.8533,
+      "step": 23738
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.724382857142857e-06,
+      "loss": 1.0515,
+      "step": 23739
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.724114285714287e-06,
+      "loss": 1.3585,
+      "step": 23740
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.723845714285714e-06,
+      "loss": 1.5242,
+      "step": 23741
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.723577142857143e-06,
+      "loss": 1.4679,
+      "step": 23742
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.723308571428572e-06,
+      "loss": 1.5279,
+      "step": 23743
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.72304e-06,
+      "loss": 1.1932,
+      "step": 23744
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.722771428571428e-06,
+      "loss": 1.103,
+      "step": 23745
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.722502857142858e-06,
+      "loss": 1.3415,
+      "step": 23746
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.722234285714286e-06,
+      "loss": 0.9698,
+      "step": 23747
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.721965714285714e-06,
+      "loss": 1.4703,
+      "step": 23748
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.721697142857144e-06,
+      "loss": 1.2959,
+      "step": 23749
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.721428571428572e-06,
+      "loss": 1.1756,
+      "step": 23750
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.72116e-06,
+      "loss": 1.4364,
+      "step": 23751
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.720891428571429e-06,
+      "loss": 1.122,
+      "step": 23752
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.720622857142859e-06,
+      "loss": 1.1373,
+      "step": 23753
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.720354285714285e-06,
+      "loss": 1.0327,
+      "step": 23754
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.720085714285715e-06,
+      "loss": 0.8145,
+      "step": 23755
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.719817142857143e-06,
+      "loss": 1.2829,
+      "step": 23756
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.719548571428571e-06,
+      "loss": 1.283,
+      "step": 23757
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.71928e-06,
+      "loss": 1.4352,
+      "step": 23758
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.71901142857143e-06,
+      "loss": 1.2319,
+      "step": 23759
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.718742857142857e-06,
+      "loss": 1.4157,
+      "step": 23760
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.718474285714286e-06,
+      "loss": 1.2776,
+      "step": 23761
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.718205714285715e-06,
+      "loss": 0.8035,
+      "step": 23762
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.717937142857144e-06,
+      "loss": 1.0785,
+      "step": 23763
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.717668571428572e-06,
+      "loss": 1.1853,
+      "step": 23764
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.7174e-06,
+      "loss": 1.5995,
+      "step": 23765
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.71713142857143e-06,
+      "loss": 1.1466,
+      "step": 23766
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.716862857142856e-06,
+      "loss": 1.4162,
+      "step": 23767
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.716594285714286e-06,
+      "loss": 1.1557,
+      "step": 23768
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.716325714285714e-06,
+      "loss": 1.1089,
+      "step": 23769
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.716057142857143e-06,
+      "loss": 1.3606,
+      "step": 23770
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.71578857142857e-06,
+      "loss": 1.3102,
+      "step": 23771
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.71552e-06,
+      "loss": 1.4513,
+      "step": 23772
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.715251428571429e-06,
+      "loss": 1.2402,
+      "step": 23773
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.714982857142857e-06,
+      "loss": 1.7283,
+      "step": 23774
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.714714285714287e-06,
+      "loss": 1.2314,
+      "step": 23775
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.714445714285715e-06,
+      "loss": 1.423,
+      "step": 23776
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.714177142857143e-06,
+      "loss": 1.3832,
+      "step": 23777
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.713908571428571e-06,
+      "loss": 1.0845,
+      "step": 23778
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.713640000000001e-06,
+      "loss": 1.2668,
+      "step": 23779
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.713371428571428e-06,
+      "loss": 0.7313,
+      "step": 23780
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.713102857142858e-06,
+      "loss": 0.6131,
+      "step": 23781
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.712834285714286e-06,
+      "loss": 1.1866,
+      "step": 23782
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.712565714285714e-06,
+      "loss": 1.0463,
+      "step": 23783
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.712297142857142e-06,
+      "loss": 1.3285,
+      "step": 23784
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.712028571428572e-06,
+      "loss": 0.9562,
+      "step": 23785
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.71176e-06,
+      "loss": 1.1421,
+      "step": 23786
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.711491428571428e-06,
+      "loss": 1.2323,
+      "step": 23787
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.711222857142858e-06,
+      "loss": 1.095,
+      "step": 23788
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.710954285714286e-06,
+      "loss": 1.2142,
+      "step": 23789
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.710685714285715e-06,
+      "loss": 1.0478,
+      "step": 23790
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.710417142857143e-06,
+      "loss": 0.9787,
+      "step": 23791
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.710148571428573e-06,
+      "loss": 1.1437,
+      "step": 23792
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.709879999999999e-06,
+      "loss": 1.4701,
+      "step": 23793
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.709611428571429e-06,
+      "loss": 1.4237,
+      "step": 23794
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.709342857142857e-06,
+      "loss": 0.8485,
+      "step": 23795
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.709074285714287e-06,
+      "loss": 1.1154,
+      "step": 23796
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.708805714285714e-06,
+      "loss": 1.1226,
+      "step": 23797
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.708537142857143e-06,
+      "loss": 1.2564,
+      "step": 23798
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.708268571428572e-06,
+      "loss": 1.0207,
+      "step": 23799
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.708e-06,
+      "loss": 1.1552,
+      "step": 23800
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.70773142857143e-06,
+      "loss": 1.0061,
+      "step": 23801
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.707462857142858e-06,
+      "loss": 1.2053,
+      "step": 23802
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.707194285714286e-06,
+      "loss": 1.0015,
+      "step": 23803
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.706925714285714e-06,
+      "loss": 1.3079,
+      "step": 23804
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.706657142857144e-06,
+      "loss": 0.8273,
+      "step": 23805
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.70638857142857e-06,
+      "loss": 1.0736,
+      "step": 23806
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.70612e-06,
+      "loss": 1.0189,
+      "step": 23807
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.705851428571429e-06,
+      "loss": 1.4312,
+      "step": 23808
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.705582857142858e-06,
+      "loss": 0.8221,
+      "step": 23809
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.705314285714285e-06,
+      "loss": 1.0825,
+      "step": 23810
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.705045714285715e-06,
+      "loss": 1.5037,
+      "step": 23811
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.704777142857143e-06,
+      "loss": 1.53,
+      "step": 23812
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.704508571428571e-06,
+      "loss": 0.8942,
+      "step": 23813
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.704240000000001e-06,
+      "loss": 1.558,
+      "step": 23814
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.703971428571429e-06,
+      "loss": 1.1988,
+      "step": 23815
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.703702857142857e-06,
+      "loss": 1.4091,
+      "step": 23816
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.703434285714285e-06,
+      "loss": 1.0179,
+      "step": 23817
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.703165714285715e-06,
+      "loss": 1.0694,
+      "step": 23818
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.702897142857144e-06,
+      "loss": 1.0091,
+      "step": 23819
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.702628571428572e-06,
+      "loss": 0.9844,
+      "step": 23820
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.70236e-06,
+      "loss": 1.0756,
+      "step": 23821
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.70209142857143e-06,
+      "loss": 0.9692,
+      "step": 23822
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.701822857142856e-06,
+      "loss": 1.1423,
+      "step": 23823
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.701554285714286e-06,
+      "loss": 1.3103,
+      "step": 23824
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.701285714285714e-06,
+      "loss": 1.4658,
+      "step": 23825
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.701017142857142e-06,
+      "loss": 1.2725,
+      "step": 23826
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.700748571428572e-06,
+      "loss": 0.9684,
+      "step": 23827
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.70048e-06,
+      "loss": 1.1774,
+      "step": 23828
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.700211428571429e-06,
+      "loss": 0.8665,
+      "step": 23829
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.699942857142857e-06,
+      "loss": 0.963,
+      "step": 23830
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.699674285714287e-06,
+      "loss": 1.2127,
+      "step": 23831
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.699405714285715e-06,
+      "loss": 1.199,
+      "step": 23832
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.699137142857143e-06,
+      "loss": 0.9054,
+      "step": 23833
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.698868571428571e-06,
+      "loss": 1.3929,
+      "step": 23834
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.698600000000001e-06,
+      "loss": 1.5297,
+      "step": 23835
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.698331428571428e-06,
+      "loss": 0.9605,
+      "step": 23836
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.698062857142857e-06,
+      "loss": 1.2868,
+      "step": 23837
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.697794285714286e-06,
+      "loss": 1.1588,
+      "step": 23838
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.697525714285714e-06,
+      "loss": 1.4007,
+      "step": 23839
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.697257142857144e-06,
+      "loss": 0.8113,
+      "step": 23840
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.696988571428572e-06,
+      "loss": 1.0149,
+      "step": 23841
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.69672e-06,
+      "loss": 1.0727,
+      "step": 23842
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.696451428571428e-06,
+      "loss": 1.592,
+      "step": 23843
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.696182857142858e-06,
+      "loss": 0.9424,
+      "step": 23844
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.695914285714286e-06,
+      "loss": 1.2445,
+      "step": 23845
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.695645714285714e-06,
+      "loss": 0.9073,
+      "step": 23846
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.695377142857143e-06,
+      "loss": 1.0506,
+      "step": 23847
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.695108571428573e-06,
+      "loss": 1.1709,
+      "step": 23848
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.694839999999999e-06,
+      "loss": 1.2113,
+      "step": 23849
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.694571428571429e-06,
+      "loss": 1.2659,
+      "step": 23850
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.694302857142857e-06,
+      "loss": 1.2272,
+      "step": 23851
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.694034285714285e-06,
+      "loss": 1.1152,
+      "step": 23852
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.693765714285715e-06,
+      "loss": 1.4212,
+      "step": 23853
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.693497142857143e-06,
+      "loss": 1.2594,
+      "step": 23854
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.693228571428573e-06,
+      "loss": 1.3222,
+      "step": 23855
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.69296e-06,
+      "loss": 1.3667,
+      "step": 23856
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.69269142857143e-06,
+      "loss": 1.0348,
+      "step": 23857
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.692422857142858e-06,
+      "loss": 1.4145,
+      "step": 23858
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.692154285714286e-06,
+      "loss": 1.0105,
+      "step": 23859
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.691885714285714e-06,
+      "loss": 1.0375,
+      "step": 23860
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.691617142857144e-06,
+      "loss": 0.9456,
+      "step": 23861
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.69134857142857e-06,
+      "loss": 0.9375,
+      "step": 23862
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.69108e-06,
+      "loss": 1.0135,
+      "step": 23863
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.690811428571428e-06,
+      "loss": 1.1298,
+      "step": 23864
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.690542857142857e-06,
+      "loss": 1.3712,
+      "step": 23865
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.690274285714286e-06,
+      "loss": 1.1097,
+      "step": 23866
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.690005714285715e-06,
+      "loss": 0.9507,
+      "step": 23867
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.689737142857144e-06,
+      "loss": 1.5382,
+      "step": 23868
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.689468571428571e-06,
+      "loss": 1.2348,
+      "step": 23869
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.689200000000001e-06,
+      "loss": 1.3689,
+      "step": 23870
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.688931428571429e-06,
+      "loss": 0.9667,
+      "step": 23871
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.688662857142857e-06,
+      "loss": 1.0397,
+      "step": 23872
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.688394285714285e-06,
+      "loss": 1.3246,
+      "step": 23873
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.688125714285715e-06,
+      "loss": 1.6537,
+      "step": 23874
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.687857142857142e-06,
+      "loss": 1.0548,
+      "step": 23875
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.687588571428572e-06,
+      "loss": 0.9587,
+      "step": 23876
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.68732e-06,
+      "loss": 1.094,
+      "step": 23877
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.68705142857143e-06,
+      "loss": 1.0921,
+      "step": 23878
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.686782857142858e-06,
+      "loss": 1.2337,
+      "step": 23879
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.686514285714286e-06,
+      "loss": 1.1052,
+      "step": 23880
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.686245714285716e-06,
+      "loss": 1.3655,
+      "step": 23881
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.685977142857142e-06,
+      "loss": 0.7669,
+      "step": 23882
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.685708571428572e-06,
+      "loss": 0.952,
+      "step": 23883
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.68544e-06,
+      "loss": 1.1174,
+      "step": 23884
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.685171428571429e-06,
+      "loss": 0.71,
+      "step": 23885
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.684902857142857e-06,
+      "loss": 1.1639,
+      "step": 23886
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 7.684634285714287e-06,
+      "loss": 0.807,
+      "step": 23887
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.684365714285713e-06,
+      "loss": 1.0314,
+      "step": 23888
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.684097142857143e-06,
+      "loss": 1.233,
+      "step": 23889
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.683828571428571e-06,
+      "loss": 1.1236,
+      "step": 23890
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.683560000000001e-06,
+      "loss": 0.9258,
+      "step": 23891
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.68329142857143e-06,
+      "loss": 1.2821,
+      "step": 23892
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.683022857142857e-06,
+      "loss": 1.0479,
+      "step": 23893
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.682754285714287e-06,
+      "loss": 1.4252,
+      "step": 23894
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.682485714285714e-06,
+      "loss": 1.111,
+      "step": 23895
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.682217142857144e-06,
+      "loss": 1.065,
+      "step": 23896
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.681948571428572e-06,
+      "loss": 1.2508,
+      "step": 23897
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.68168e-06,
+      "loss": 1.3965,
+      "step": 23898
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.681411428571428e-06,
+      "loss": 0.9831,
+      "step": 23899
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.681142857142858e-06,
+      "loss": 1.2558,
+      "step": 23900
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.680874285714286e-06,
+      "loss": 1.3016,
+      "step": 23901
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.680605714285714e-06,
+      "loss": 1.1785,
+      "step": 23902
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.680337142857143e-06,
+      "loss": 1.1657,
+      "step": 23903
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.680068571428572e-06,
+      "loss": 1.372,
+      "step": 23904
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.6798e-06,
+      "loss": 0.9191,
+      "step": 23905
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.679531428571429e-06,
+      "loss": 0.9999,
+      "step": 23906
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.679262857142859e-06,
+      "loss": 0.8898,
+      "step": 23907
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.678994285714285e-06,
+      "loss": 1.4828,
+      "step": 23908
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.678725714285715e-06,
+      "loss": 1.2435,
+      "step": 23909
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.678457142857143e-06,
+      "loss": 1.0589,
+      "step": 23910
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.678188571428571e-06,
+      "loss": 0.9696,
+      "step": 23911
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.67792e-06,
+      "loss": 0.6806,
+      "step": 23912
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.67765142857143e-06,
+      "loss": 1.0771,
+      "step": 23913
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.677382857142858e-06,
+      "loss": 1.1195,
+      "step": 23914
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.677114285714286e-06,
+      "loss": 0.9672,
+      "step": 23915
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.676845714285714e-06,
+      "loss": 1.1864,
+      "step": 23916
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.676577142857144e-06,
+      "loss": 1.0852,
+      "step": 23917
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.676308571428572e-06,
+      "loss": 0.8957,
+      "step": 23918
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.67604e-06,
+      "loss": 1.324,
+      "step": 23919
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.67577142857143e-06,
+      "loss": 0.7339,
+      "step": 23920
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.675502857142856e-06,
+      "loss": 1.1117,
+      "step": 23921
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.675234285714286e-06,
+      "loss": 1.0811,
+      "step": 23922
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.674965714285715e-06,
+      "loss": 1.0816,
+      "step": 23923
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.674697142857144e-06,
+      "loss": 1.2358,
+      "step": 23924
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.674428571428571e-06,
+      "loss": 1.2648,
+      "step": 23925
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.67416e-06,
+      "loss": 1.2259,
+      "step": 23926
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.673891428571429e-06,
+      "loss": 1.4889,
+      "step": 23927
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.673622857142857e-06,
+      "loss": 1.3227,
+      "step": 23928
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.673354285714285e-06,
+      "loss": 1.4612,
+      "step": 23929
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.673085714285715e-06,
+      "loss": 1.0031,
+      "step": 23930
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.672817142857143e-06,
+      "loss": 1.5265,
+      "step": 23931
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.672548571428571e-06,
+      "loss": 1.4557,
+      "step": 23932
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.672280000000001e-06,
+      "loss": 1.1546,
+      "step": 23933
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.672011428571428e-06,
+      "loss": 1.0696,
+      "step": 23934
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.671742857142858e-06,
+      "loss": 1.2183,
+      "step": 23935
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.671474285714286e-06,
+      "loss": 0.9618,
+      "step": 23936
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.671205714285716e-06,
+      "loss": 1.0529,
+      "step": 23937
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.670937142857142e-06,
+      "loss": 1.0135,
+      "step": 23938
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.670668571428572e-06,
+      "loss": 0.9163,
+      "step": 23939
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.6704e-06,
+      "loss": 0.9101,
+      "step": 23940
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.670131428571428e-06,
+      "loss": 0.8977,
+      "step": 23941
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.669862857142857e-06,
+      "loss": 1.7817,
+      "step": 23942
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.669594285714287e-06,
+      "loss": 1.3697,
+      "step": 23943
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.669325714285715e-06,
+      "loss": 0.8939,
+      "step": 23944
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.669057142857143e-06,
+      "loss": 1.2726,
+      "step": 23945
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.668788571428573e-06,
+      "loss": 0.9208,
+      "step": 23946
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.668520000000001e-06,
+      "loss": 1.7494,
+      "step": 23947
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.668251428571429e-06,
+      "loss": 1.0556,
+      "step": 23948
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.667982857142857e-06,
+      "loss": 1.039,
+      "step": 23949
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.667714285714287e-06,
+      "loss": 1.2185,
+      "step": 23950
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.667445714285714e-06,
+      "loss": 0.7511,
+      "step": 23951
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.667177142857143e-06,
+      "loss": 1.126,
+      "step": 23952
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.666908571428572e-06,
+      "loss": 1.3185,
+      "step": 23953
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.66664e-06,
+      "loss": 1.1519,
+      "step": 23954
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.666371428571428e-06,
+      "loss": 1.3184,
+      "step": 23955
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.666102857142858e-06,
+      "loss": 0.8663,
+      "step": 23956
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.665834285714286e-06,
+      "loss": 1.2444,
+      "step": 23957
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.665565714285714e-06,
+      "loss": 1.1691,
+      "step": 23958
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.665297142857144e-06,
+      "loss": 0.9459,
+      "step": 23959
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.665028571428572e-06,
+      "loss": 1.0414,
+      "step": 23960
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.66476e-06,
+      "loss": 0.8826,
+      "step": 23961
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.664491428571429e-06,
+      "loss": 0.79,
+      "step": 23962
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.664222857142859e-06,
+      "loss": 1.2412,
+      "step": 23963
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.663954285714285e-06,
+      "loss": 1.41,
+      "step": 23964
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.663685714285715e-06,
+      "loss": 1.1529,
+      "step": 23965
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.663417142857143e-06,
+      "loss": 1.3464,
+      "step": 23966
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.663148571428571e-06,
+      "loss": 0.6369,
+      "step": 23967
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.66288e-06,
+      "loss": 1.0444,
+      "step": 23968
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.66261142857143e-06,
+      "loss": 0.9205,
+      "step": 23969
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.662342857142857e-06,
+      "loss": 1.2445,
+      "step": 23970
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.662074285714286e-06,
+      "loss": 0.8789,
+      "step": 23971
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.661805714285715e-06,
+      "loss": 1.3324,
+      "step": 23972
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.661537142857144e-06,
+      "loss": 1.3913,
+      "step": 23973
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.661268571428572e-06,
+      "loss": 1.0376,
+      "step": 23974
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.661e-06,
+      "loss": 0.6025,
+      "step": 23975
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.66073142857143e-06,
+      "loss": 1.4037,
+      "step": 23976
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.660462857142856e-06,
+      "loss": 1.1547,
+      "step": 23977
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.660194285714286e-06,
+      "loss": 0.8229,
+      "step": 23978
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.659925714285714e-06,
+      "loss": 1.1204,
+      "step": 23979
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.659657142857143e-06,
+      "loss": 1.1859,
+      "step": 23980
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.65938857142857e-06,
+      "loss": 1.2351,
+      "step": 23981
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.65912e-06,
+      "loss": 1.0723,
+      "step": 23982
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.658851428571429e-06,
+      "loss": 1.4943,
+      "step": 23983
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.658582857142857e-06,
+      "loss": 1.3298,
+      "step": 23984
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.658314285714287e-06,
+      "loss": 0.6175,
+      "step": 23985
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.658045714285715e-06,
+      "loss": 1.091,
+      "step": 23986
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.657777142857143e-06,
+      "loss": 1.0494,
+      "step": 23987
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.657508571428571e-06,
+      "loss": 1.4461,
+      "step": 23988
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.657240000000001e-06,
+      "loss": 1.1971,
+      "step": 23989
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.656971428571428e-06,
+      "loss": 1.4631,
+      "step": 23990
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.656702857142858e-06,
+      "loss": 0.8493,
+      "step": 23991
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.656434285714286e-06,
+      "loss": 1.1008,
+      "step": 23992
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.656165714285716e-06,
+      "loss": 0.9069,
+      "step": 23993
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.655897142857142e-06,
+      "loss": 1.1188,
+      "step": 23994
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.655628571428572e-06,
+      "loss": 1.0478,
+      "step": 23995
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.65536e-06,
+      "loss": 1.3353,
+      "step": 23996
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.655091428571428e-06,
+      "loss": 1.5435,
+      "step": 23997
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.654822857142858e-06,
+      "loss": 1.1909,
+      "step": 23998
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.654554285714286e-06,
+      "loss": 1.2561,
+      "step": 23999
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.654285714285715e-06,
+      "loss": 1.3436,
+      "step": 24000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.654017142857143e-06,
+      "loss": 0.7458,
+      "step": 24001
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.653748571428573e-06,
+      "loss": 1.6798,
+      "step": 24002
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.653479999999999e-06,
+      "loss": 1.1096,
+      "step": 24003
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.653211428571429e-06,
+      "loss": 0.8386,
+      "step": 24004
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.652942857142857e-06,
+      "loss": 1.0745,
+      "step": 24005
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.652674285714287e-06,
+      "loss": 1.3798,
+      "step": 24006
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.652405714285713e-06,
+      "loss": 1.038,
+      "step": 24007
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.652137142857143e-06,
+      "loss": 0.7253,
+      "step": 24008
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.651868571428572e-06,
+      "loss": 0.7802,
+      "step": 24009
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.6516e-06,
+      "loss": 1.1079,
+      "step": 24010
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.65133142857143e-06,
+      "loss": 1.2656,
+      "step": 24011
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.651062857142858e-06,
+      "loss": 1.2235,
+      "step": 24012
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.650794285714286e-06,
+      "loss": 1.0167,
+      "step": 24013
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.650525714285714e-06,
+      "loss": 0.9721,
+      "step": 24014
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.650257142857144e-06,
+      "loss": 1.2233,
+      "step": 24015
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.64998857142857e-06,
+      "loss": 1.2465,
+      "step": 24016
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.64972e-06,
+      "loss": 1.1989,
+      "step": 24017
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.649451428571429e-06,
+      "loss": 1.1529,
+      "step": 24018
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.649182857142858e-06,
+      "loss": 1.2016,
+      "step": 24019
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.648914285714285e-06,
+      "loss": 1.0588,
+      "step": 24020
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.648645714285715e-06,
+      "loss": 1.3654,
+      "step": 24021
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.648377142857143e-06,
+      "loss": 1.2278,
+      "step": 24022
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.648108571428571e-06,
+      "loss": 1.4791,
+      "step": 24023
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.647840000000001e-06,
+      "loss": 1.3998,
+      "step": 24024
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.647571428571429e-06,
+      "loss": 0.9596,
+      "step": 24025
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.647302857142857e-06,
+      "loss": 0.7695,
+      "step": 24026
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.647034285714285e-06,
+      "loss": 1.204,
+      "step": 24027
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.646765714285715e-06,
+      "loss": 1.1476,
+      "step": 24028
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.646497142857144e-06,
+      "loss": 1.1691,
+      "step": 24029
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.646228571428572e-06,
+      "loss": 1.3429,
+      "step": 24030
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.64596e-06,
+      "loss": 0.8662,
+      "step": 24031
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.64569142857143e-06,
+      "loss": 1.2975,
+      "step": 24032
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.645422857142856e-06,
+      "loss": 1.3378,
+      "step": 24033
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.645154285714286e-06,
+      "loss": 0.8858,
+      "step": 24034
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.644885714285714e-06,
+      "loss": 1.1459,
+      "step": 24035
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.644617142857142e-06,
+      "loss": 1.095,
+      "step": 24036
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.644348571428572e-06,
+      "loss": 1.6073,
+      "step": 24037
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.64408e-06,
+      "loss": 1.4829,
+      "step": 24038
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.643811428571429e-06,
+      "loss": 1.3006,
+      "step": 24039
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.643542857142857e-06,
+      "loss": 0.8783,
+      "step": 24040
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.643274285714287e-06,
+      "loss": 1.3839,
+      "step": 24041
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.643005714285715e-06,
+      "loss": 1.1715,
+      "step": 24042
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.642737142857143e-06,
+      "loss": 0.8112,
+      "step": 24043
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.642468571428571e-06,
+      "loss": 0.9823,
+      "step": 24044
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.642200000000001e-06,
+      "loss": 1.1141,
+      "step": 24045
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.641931428571428e-06,
+      "loss": 0.9983,
+      "step": 24046
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.641662857142857e-06,
+      "loss": 0.7116,
+      "step": 24047
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.641394285714286e-06,
+      "loss": 1.1595,
+      "step": 24048
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.641125714285714e-06,
+      "loss": 1.271,
+      "step": 24049
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.640857142857144e-06,
+      "loss": 1.0125,
+      "step": 24050
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.640588571428572e-06,
+      "loss": 1.2286,
+      "step": 24051
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.640320000000002e-06,
+      "loss": 1.0912,
+      "step": 24052
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.640051428571428e-06,
+      "loss": 1.4785,
+      "step": 24053
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.639782857142858e-06,
+      "loss": 1.0437,
+      "step": 24054
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.639514285714286e-06,
+      "loss": 1.323,
+      "step": 24055
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.639245714285714e-06,
+      "loss": 0.8316,
+      "step": 24056
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.638977142857143e-06,
+      "loss": 1.1994,
+      "step": 24057
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.638708571428573e-06,
+      "loss": 1.0661,
+      "step": 24058
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.638439999999999e-06,
+      "loss": 0.9383,
+      "step": 24059
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.638171428571429e-06,
+      "loss": 0.9826,
+      "step": 24060
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.637902857142857e-06,
+      "loss": 1.0535,
+      "step": 24061
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 7.637634285714285e-06,
+      "loss": 0.8461,
+      "step": 24062
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.637365714285715e-06,
+      "loss": 1.5073,
+      "step": 24063
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.637097142857143e-06,
+      "loss": 1.3364,
+      "step": 24064
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.636828571428573e-06,
+      "loss": 0.8356,
+      "step": 24065
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.63656e-06,
+      "loss": 1.2364,
+      "step": 24066
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.63629142857143e-06,
+      "loss": 1.3356,
+      "step": 24067
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.636022857142858e-06,
+      "loss": 1.5459,
+      "step": 24068
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.635754285714286e-06,
+      "loss": 1.2455,
+      "step": 24069
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.635485714285714e-06,
+      "loss": 1.0472,
+      "step": 24070
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.635217142857144e-06,
+      "loss": 0.9257,
+      "step": 24071
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.63494857142857e-06,
+      "loss": 0.9985,
+      "step": 24072
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.63468e-06,
+      "loss": 1.4636,
+      "step": 24073
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.634411428571428e-06,
+      "loss": 0.8559,
+      "step": 24074
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.634142857142858e-06,
+      "loss": 1.4208,
+      "step": 24075
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.633874285714286e-06,
+      "loss": 1.0017,
+      "step": 24076
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.633605714285715e-06,
+      "loss": 1.5236,
+      "step": 24077
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.633337142857144e-06,
+      "loss": 1.0921,
+      "step": 24078
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.633068571428571e-06,
+      "loss": 1.2286,
+      "step": 24079
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.632800000000001e-06,
+      "loss": 0.9286,
+      "step": 24080
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.632531428571429e-06,
+      "loss": 1.1148,
+      "step": 24081
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.632262857142857e-06,
+      "loss": 0.9741,
+      "step": 24082
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.631994285714285e-06,
+      "loss": 1.0802,
+      "step": 24083
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.631725714285715e-06,
+      "loss": 1.3618,
+      "step": 24084
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.631457142857142e-06,
+      "loss": 0.9908,
+      "step": 24085
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.631188571428572e-06,
+      "loss": 1.4023,
+      "step": 24086
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.63092e-06,
+      "loss": 1.3264,
+      "step": 24087
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.63065142857143e-06,
+      "loss": 1.3584,
+      "step": 24088
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.630382857142858e-06,
+      "loss": 0.9874,
+      "step": 24089
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.630114285714286e-06,
+      "loss": 1.0912,
+      "step": 24090
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.629845714285716e-06,
+      "loss": 1.6113,
+      "step": 24091
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.629577142857142e-06,
+      "loss": 1.1417,
+      "step": 24092
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.629308571428571e-06,
+      "loss": 0.8344,
+      "step": 24093
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.62904e-06,
+      "loss": 1.0237,
+      "step": 24094
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6287714285714286e-06,
+      "loss": 1.3054,
+      "step": 24095
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.628502857142858e-06,
+      "loss": 1.4767,
+      "step": 24096
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.628234285714287e-06,
+      "loss": 1.4813,
+      "step": 24097
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.627965714285716e-06,
+      "loss": 1.3196,
+      "step": 24098
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.627697142857143e-06,
+      "loss": 1.1111,
+      "step": 24099
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.627428571428572e-06,
+      "loss": 1.6148,
+      "step": 24100
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.627160000000001e-06,
+      "loss": 0.8044,
+      "step": 24101
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.626891428571428e-06,
+      "loss": 0.7976,
+      "step": 24102
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.626622857142857e-06,
+      "loss": 1.5191,
+      "step": 24103
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.626354285714286e-06,
+      "loss": 1.0309,
+      "step": 24104
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.626085714285714e-06,
+      "loss": 0.83,
+      "step": 24105
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.625817142857143e-06,
+      "loss": 1.3367,
+      "step": 24106
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.625548571428572e-06,
+      "loss": 0.9774,
+      "step": 24107
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.62528e-06,
+      "loss": 0.7932,
+      "step": 24108
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.625011428571429e-06,
+      "loss": 1.078,
+      "step": 24109
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.624742857142858e-06,
+      "loss": 0.8443,
+      "step": 24110
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.624474285714287e-06,
+      "loss": 1.376,
+      "step": 24111
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.624205714285714e-06,
+      "loss": 1.2213,
+      "step": 24112
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.623937142857143e-06,
+      "loss": 1.1049,
+      "step": 24113
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.623668571428572e-06,
+      "loss": 1.0584,
+      "step": 24114
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6234e-06,
+      "loss": 1.2362,
+      "step": 24115
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.623131428571429e-06,
+      "loss": 1.079,
+      "step": 24116
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.622862857142858e-06,
+      "loss": 0.7811,
+      "step": 24117
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.622594285714285e-06,
+      "loss": 0.7785,
+      "step": 24118
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.622325714285714e-06,
+      "loss": 1.5221,
+      "step": 24119
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.622057142857143e-06,
+      "loss": 1.0654,
+      "step": 24120
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.621788571428572e-06,
+      "loss": 1.4581,
+      "step": 24121
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.62152e-06,
+      "loss": 1.0474,
+      "step": 24122
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.621251428571429e-06,
+      "loss": 1.065,
+      "step": 24123
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.620982857142858e-06,
+      "loss": 1.3393,
+      "step": 24124
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.620714285714286e-06,
+      "loss": 1.068,
+      "step": 24125
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.620445714285715e-06,
+      "loss": 1.2563,
+      "step": 24126
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.620177142857144e-06,
+      "loss": 0.9724,
+      "step": 24127
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.619908571428571e-06,
+      "loss": 1.2191,
+      "step": 24128
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.61964e-06,
+      "loss": 1.1489,
+      "step": 24129
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.619371428571429e-06,
+      "loss": 1.2581,
+      "step": 24130
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6191028571428565e-06,
+      "loss": 1.0909,
+      "step": 24131
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6188342857142855e-06,
+      "loss": 1.5021,
+      "step": 24132
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6185657142857145e-06,
+      "loss": 1.2281,
+      "step": 24133
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6182971428571435e-06,
+      "loss": 1.0009,
+      "step": 24134
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.618028571428572e-06,
+      "loss": 1.0696,
+      "step": 24135
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.617760000000001e-06,
+      "loss": 0.9833,
+      "step": 24136
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.61749142857143e-06,
+      "loss": 1.2184,
+      "step": 24137
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.617222857142857e-06,
+      "loss": 1.2305,
+      "step": 24138
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.616954285714286e-06,
+      "loss": 1.2237,
+      "step": 24139
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.616685714285715e-06,
+      "loss": 0.9868,
+      "step": 24140
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6164171428571425e-06,
+      "loss": 1.2638,
+      "step": 24141
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6161485714285715e-06,
+      "loss": 0.8973,
+      "step": 24142
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6158800000000005e-06,
+      "loss": 1.2621,
+      "step": 24143
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6156114285714295e-06,
+      "loss": 0.9495,
+      "step": 24144
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.615342857142857e-06,
+      "loss": 1.0251,
+      "step": 24145
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.615074285714286e-06,
+      "loss": 1.0649,
+      "step": 24146
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.614805714285715e-06,
+      "loss": 1.1062,
+      "step": 24147
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.614537142857143e-06,
+      "loss": 1.2086,
+      "step": 24148
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.614268571428572e-06,
+      "loss": 1.2732,
+      "step": 24149
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.614000000000001e-06,
+      "loss": 1.2385,
+      "step": 24150
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6137314285714285e-06,
+      "loss": 1.1286,
+      "step": 24151
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6134628571428575e-06,
+      "loss": 1.2626,
+      "step": 24152
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6131942857142865e-06,
+      "loss": 1.1643,
+      "step": 24153
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.612925714285714e-06,
+      "loss": 1.0817,
+      "step": 24154
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.612657142857143e-06,
+      "loss": 1.1861,
+      "step": 24155
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.612388571428572e-06,
+      "loss": 1.1905,
+      "step": 24156
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.612120000000001e-06,
+      "loss": 1.6546,
+      "step": 24157
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.611851428571428e-06,
+      "loss": 1.0058,
+      "step": 24158
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.611582857142857e-06,
+      "loss": 1.3782,
+      "step": 24159
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.611314285714286e-06,
+      "loss": 1.183,
+      "step": 24160
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6110457142857145e-06,
+      "loss": 1.3703,
+      "step": 24161
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6107771428571435e-06,
+      "loss": 1.0388,
+      "step": 24162
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6105085714285725e-06,
+      "loss": 0.9748,
+      "step": 24163
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.61024e-06,
+      "loss": 0.9269,
+      "step": 24164
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.609971428571429e-06,
+      "loss": 1.0289,
+      "step": 24165
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.609702857142858e-06,
+      "loss": 1.0305,
+      "step": 24166
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.609434285714287e-06,
+      "loss": 0.9344,
+      "step": 24167
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.609165714285714e-06,
+      "loss": 1.2203,
+      "step": 24168
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.608897142857143e-06,
+      "loss": 1.2471,
+      "step": 24169
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.608628571428572e-06,
+      "loss": 1.1699,
+      "step": 24170
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.60836e-06,
+      "loss": 1.0072,
+      "step": 24171
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.608091428571429e-06,
+      "loss": 1.0371,
+      "step": 24172
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.607822857142858e-06,
+      "loss": 1.1791,
+      "step": 24173
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.607554285714286e-06,
+      "loss": 1.0701,
+      "step": 24174
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.607285714285715e-06,
+      "loss": 1.3286,
+      "step": 24175
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.607017142857144e-06,
+      "loss": 1.1086,
+      "step": 24176
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.606748571428571e-06,
+      "loss": 1.0863,
+      "step": 24177
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.60648e-06,
+      "loss": 1.0821,
+      "step": 24178
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.606211428571429e-06,
+      "loss": 1.1189,
+      "step": 24179
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.605942857142858e-06,
+      "loss": 1.2462,
+      "step": 24180
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.605674285714286e-06,
+      "loss": 1.1717,
+      "step": 24181
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.605405714285715e-06,
+      "loss": 1.1772,
+      "step": 24182
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.605137142857144e-06,
+      "loss": 1.12,
+      "step": 24183
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.604868571428571e-06,
+      "loss": 1.2854,
+      "step": 24184
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6046e-06,
+      "loss": 1.3454,
+      "step": 24185
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.604331428571429e-06,
+      "loss": 1.2143,
+      "step": 24186
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.604062857142857e-06,
+      "loss": 0.9567,
+      "step": 24187
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.603794285714286e-06,
+      "loss": 1.6642,
+      "step": 24188
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.603525714285715e-06,
+      "loss": 1.1009,
+      "step": 24189
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6032571428571426e-06,
+      "loss": 1.021,
+      "step": 24190
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.602988571428572e-06,
+      "loss": 0.947,
+      "step": 24191
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.602720000000001e-06,
+      "loss": 1.2026,
+      "step": 24192
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.60245142857143e-06,
+      "loss": 1.5336,
+      "step": 24193
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.602182857142857e-06,
+      "loss": 1.2521,
+      "step": 24194
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.601914285714286e-06,
+      "loss": 1.2759,
+      "step": 24195
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.601645714285715e-06,
+      "loss": 1.0807,
+      "step": 24196
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.601377142857142e-06,
+      "loss": 1.3147,
+      "step": 24197
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.601108571428571e-06,
+      "loss": 1.0987,
+      "step": 24198
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.60084e-06,
+      "loss": 0.73,
+      "step": 24199
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.6005714285714286e-06,
+      "loss": 1.1567,
+      "step": 24200
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.600302857142858e-06,
+      "loss": 1.3435,
+      "step": 24201
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.600034285714287e-06,
+      "loss": 1.4376,
+      "step": 24202
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.599765714285716e-06,
+      "loss": 1.4982,
+      "step": 24203
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.599497142857143e-06,
+      "loss": 1.1524,
+      "step": 24204
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.599228571428572e-06,
+      "loss": 1.373,
+      "step": 24205
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.598960000000001e-06,
+      "loss": 1.4391,
+      "step": 24206
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.598691428571428e-06,
+      "loss": 1.0159,
+      "step": 24207
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.598422857142857e-06,
+      "loss": 1.0196,
+      "step": 24208
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.598154285714286e-06,
+      "loss": 1.0109,
+      "step": 24209
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.597885714285714e-06,
+      "loss": 0.8657,
+      "step": 24210
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.597617142857143e-06,
+      "loss": 1.864,
+      "step": 24211
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.597348571428572e-06,
+      "loss": 1.306,
+      "step": 24212
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.597079999999999e-06,
+      "loss": 0.934,
+      "step": 24213
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.596811428571429e-06,
+      "loss": 0.9259,
+      "step": 24214
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.596542857142858e-06,
+      "loss": 0.982,
+      "step": 24215
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.596274285714287e-06,
+      "loss": 0.9661,
+      "step": 24216
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.596005714285714e-06,
+      "loss": 1.0822,
+      "step": 24217
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.595737142857143e-06,
+      "loss": 0.9399,
+      "step": 24218
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.595468571428572e-06,
+      "loss": 1.0015,
+      "step": 24219
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.5952e-06,
+      "loss": 1.5799,
+      "step": 24220
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.594931428571429e-06,
+      "loss": 1.2177,
+      "step": 24221
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.594662857142858e-06,
+      "loss": 0.8515,
+      "step": 24222
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.594394285714285e-06,
+      "loss": 1.291,
+      "step": 24223
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.594125714285714e-06,
+      "loss": 1.3017,
+      "step": 24224
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.593857142857143e-06,
+      "loss": 1.4391,
+      "step": 24225
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.593588571428572e-06,
+      "loss": 1.2333,
+      "step": 24226
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.59332e-06,
+      "loss": 0.9928,
+      "step": 24227
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.593051428571429e-06,
+      "loss": 0.9155,
+      "step": 24228
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.592782857142858e-06,
+      "loss": 1.1479,
+      "step": 24229
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.592514285714286e-06,
+      "loss": 1.2387,
+      "step": 24230
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.592245714285715e-06,
+      "loss": 1.3698,
+      "step": 24231
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.591977142857144e-06,
+      "loss": 1.2187,
+      "step": 24232
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.591708571428571e-06,
+      "loss": 0.7983,
+      "step": 24233
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.59144e-06,
+      "loss": 1.5262,
+      "step": 24234
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.591171428571429e-06,
+      "loss": 1.4079,
+      "step": 24235
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.5909028571428565e-06,
+      "loss": 1.0151,
+      "step": 24236
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 7.5906342857142855e-06,
+      "loss": 0.9018,
+      "step": 24237
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5903657142857145e-06,
+      "loss": 1.2541,
+      "step": 24238
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5900971428571435e-06,
+      "loss": 1.3664,
+      "step": 24239
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.589828571428572e-06,
+      "loss": 1.2484,
+      "step": 24240
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.589560000000001e-06,
+      "loss": 1.0534,
+      "step": 24241
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.58929142857143e-06,
+      "loss": 1.8055,
+      "step": 24242
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.589022857142857e-06,
+      "loss": 1.4172,
+      "step": 24243
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.588754285714286e-06,
+      "loss": 1.2198,
+      "step": 24244
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.588485714285715e-06,
+      "loss": 1.5714,
+      "step": 24245
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5882171428571425e-06,
+      "loss": 1.0303,
+      "step": 24246
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5879485714285715e-06,
+      "loss": 1.1282,
+      "step": 24247
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5876800000000005e-06,
+      "loss": 1.1372,
+      "step": 24248
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5874114285714295e-06,
+      "loss": 0.8362,
+      "step": 24249
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.587142857142857e-06,
+      "loss": 1.2686,
+      "step": 24250
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.586874285714286e-06,
+      "loss": 1.3724,
+      "step": 24251
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.586605714285715e-06,
+      "loss": 1.2722,
+      "step": 24252
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.586337142857143e-06,
+      "loss": 1.2935,
+      "step": 24253
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.586068571428572e-06,
+      "loss": 1.2536,
+      "step": 24254
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.585800000000001e-06,
+      "loss": 1.0271,
+      "step": 24255
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5855314285714285e-06,
+      "loss": 1.7954,
+      "step": 24256
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5852628571428575e-06,
+      "loss": 1.3091,
+      "step": 24257
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5849942857142865e-06,
+      "loss": 1.1669,
+      "step": 24258
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.584725714285714e-06,
+      "loss": 0.9579,
+      "step": 24259
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.584457142857143e-06,
+      "loss": 1.2507,
+      "step": 24260
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.584188571428572e-06,
+      "loss": 1.0596,
+      "step": 24261
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.583920000000001e-06,
+      "loss": 0.8745,
+      "step": 24262
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.583651428571428e-06,
+      "loss": 0.9786,
+      "step": 24263
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.583382857142857e-06,
+      "loss": 0.9631,
+      "step": 24264
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.583114285714286e-06,
+      "loss": 1.4507,
+      "step": 24265
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.582845714285714e-06,
+      "loss": 1.0114,
+      "step": 24266
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5825771428571435e-06,
+      "loss": 1.5385,
+      "step": 24267
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5823085714285725e-06,
+      "loss": 0.7673,
+      "step": 24268
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.58204e-06,
+      "loss": 1.2742,
+      "step": 24269
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.581771428571429e-06,
+      "loss": 1.3146,
+      "step": 24270
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.581502857142858e-06,
+      "loss": 1.2354,
+      "step": 24271
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.581234285714287e-06,
+      "loss": 1.4709,
+      "step": 24272
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.580965714285714e-06,
+      "loss": 1.12,
+      "step": 24273
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.580697142857143e-06,
+      "loss": 1.3069,
+      "step": 24274
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.580428571428572e-06,
+      "loss": 1.7551,
+      "step": 24275
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.58016e-06,
+      "loss": 1.3605,
+      "step": 24276
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.579891428571429e-06,
+      "loss": 1.6193,
+      "step": 24277
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.579622857142858e-06,
+      "loss": 1.5376,
+      "step": 24278
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.579354285714285e-06,
+      "loss": 1.147,
+      "step": 24279
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.579085714285715e-06,
+      "loss": 1.5059,
+      "step": 24280
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.578817142857144e-06,
+      "loss": 1.2993,
+      "step": 24281
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.578548571428571e-06,
+      "loss": 0.9563,
+      "step": 24282
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.57828e-06,
+      "loss": 0.8583,
+      "step": 24283
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.578011428571429e-06,
+      "loss": 1.1159,
+      "step": 24284
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.577742857142858e-06,
+      "loss": 0.7292,
+      "step": 24285
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.577474285714286e-06,
+      "loss": 0.8354,
+      "step": 24286
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.577205714285715e-06,
+      "loss": 1.3429,
+      "step": 24287
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.576937142857144e-06,
+      "loss": 1.0552,
+      "step": 24288
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.576668571428571e-06,
+      "loss": 0.7931,
+      "step": 24289
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5764e-06,
+      "loss": 1.1878,
+      "step": 24290
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.576131428571429e-06,
+      "loss": 0.8919,
+      "step": 24291
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.575862857142856e-06,
+      "loss": 0.8622,
+      "step": 24292
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.575594285714286e-06,
+      "loss": 0.9472,
+      "step": 24293
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.575325714285715e-06,
+      "loss": 1.2438,
+      "step": 24294
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.575057142857144e-06,
+      "loss": 1.2182,
+      "step": 24295
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.574788571428572e-06,
+      "loss": 1.3281,
+      "step": 24296
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.574520000000001e-06,
+      "loss": 0.9083,
+      "step": 24297
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.57425142857143e-06,
+      "loss": 1.0388,
+      "step": 24298
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.573982857142857e-06,
+      "loss": 1.0837,
+      "step": 24299
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.573714285714286e-06,
+      "loss": 1.4235,
+      "step": 24300
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.573445714285715e-06,
+      "loss": 0.9178,
+      "step": 24301
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.573177142857142e-06,
+      "loss": 1.5142,
+      "step": 24302
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.572908571428571e-06,
+      "loss": 1.1909,
+      "step": 24303
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.57264e-06,
+      "loss": 1.0156,
+      "step": 24304
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.572371428571428e-06,
+      "loss": 1.2399,
+      "step": 24305
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.572102857142858e-06,
+      "loss": 1.2638,
+      "step": 24306
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.571834285714287e-06,
+      "loss": 1.2162,
+      "step": 24307
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.571565714285716e-06,
+      "loss": 1.4511,
+      "step": 24308
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.571297142857143e-06,
+      "loss": 0.8181,
+      "step": 24309
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.571028571428572e-06,
+      "loss": 1.0075,
+      "step": 24310
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.570760000000001e-06,
+      "loss": 1.0659,
+      "step": 24311
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.570491428571428e-06,
+      "loss": 1.0321,
+      "step": 24312
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.570222857142857e-06,
+      "loss": 0.8976,
+      "step": 24313
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.569954285714286e-06,
+      "loss": 1.3263,
+      "step": 24314
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.569685714285714e-06,
+      "loss": 1.0778,
+      "step": 24315
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.569417142857143e-06,
+      "loss": 0.9487,
+      "step": 24316
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.569148571428572e-06,
+      "loss": 1.3825,
+      "step": 24317
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.568880000000001e-06,
+      "loss": 1.3835,
+      "step": 24318
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.568611428571428e-06,
+      "loss": 1.0184,
+      "step": 24319
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.568342857142858e-06,
+      "loss": 1.2762,
+      "step": 24320
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.568074285714287e-06,
+      "loss": 1.0635,
+      "step": 24321
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.567805714285714e-06,
+      "loss": 1.1859,
+      "step": 24322
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.567537142857143e-06,
+      "loss": 1.1208,
+      "step": 24323
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.567268571428572e-06,
+      "loss": 1.5285,
+      "step": 24324
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.567e-06,
+      "loss": 1.0611,
+      "step": 24325
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.566731428571429e-06,
+      "loss": 1.0974,
+      "step": 24326
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.566462857142858e-06,
+      "loss": 0.8937,
+      "step": 24327
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.566194285714285e-06,
+      "loss": 1.1302,
+      "step": 24328
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.565925714285714e-06,
+      "loss": 1.0787,
+      "step": 24329
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.565657142857143e-06,
+      "loss": 1.1179,
+      "step": 24330
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.565388571428572e-06,
+      "loss": 1.3731,
+      "step": 24331
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5651199999999995e-06,
+      "loss": 1.2406,
+      "step": 24332
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.564851428571429e-06,
+      "loss": 1.2309,
+      "step": 24333
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.564582857142858e-06,
+      "loss": 1.1432,
+      "step": 24334
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.564314285714286e-06,
+      "loss": 0.6991,
+      "step": 24335
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.564045714285715e-06,
+      "loss": 1.1177,
+      "step": 24336
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.563777142857144e-06,
+      "loss": 1.1181,
+      "step": 24337
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.563508571428571e-06,
+      "loss": 0.9854,
+      "step": 24338
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.56324e-06,
+      "loss": 0.9637,
+      "step": 24339
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.562971428571429e-06,
+      "loss": 1.355,
+      "step": 24340
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5627028571428565e-06,
+      "loss": 1.2797,
+      "step": 24341
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5624342857142855e-06,
+      "loss": 1.2841,
+      "step": 24342
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5621657142857145e-06,
+      "loss": 0.7292,
+      "step": 24343
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5618971428571435e-06,
+      "loss": 1.2857,
+      "step": 24344
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.561628571428571e-06,
+      "loss": 1.3521,
+      "step": 24345
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.561360000000001e-06,
+      "loss": 0.9048,
+      "step": 24346
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.56109142857143e-06,
+      "loss": 0.7424,
+      "step": 24347
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.560822857142857e-06,
+      "loss": 1.2946,
+      "step": 24348
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.560554285714286e-06,
+      "loss": 1.4186,
+      "step": 24349
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.560285714285715e-06,
+      "loss": 1.2099,
+      "step": 24350
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5600171428571425e-06,
+      "loss": 1.4387,
+      "step": 24351
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5597485714285715e-06,
+      "loss": 1.3433,
+      "step": 24352
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5594800000000005e-06,
+      "loss": 1.5745,
+      "step": 24353
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5592114285714295e-06,
+      "loss": 1.1734,
+      "step": 24354
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.558942857142857e-06,
+      "loss": 1.1162,
+      "step": 24355
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.558674285714286e-06,
+      "loss": 1.1209,
+      "step": 24356
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.558405714285715e-06,
+      "loss": 1.4385,
+      "step": 24357
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.558137142857142e-06,
+      "loss": 1.3593,
+      "step": 24358
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.557868571428572e-06,
+      "loss": 0.9395,
+      "step": 24359
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.557600000000001e-06,
+      "loss": 1.1526,
+      "step": 24360
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5573314285714285e-06,
+      "loss": 1.2419,
+      "step": 24361
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5570628571428575e-06,
+      "loss": 1.299,
+      "step": 24362
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5567942857142865e-06,
+      "loss": 1.8383,
+      "step": 24363
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.556525714285714e-06,
+      "loss": 1.0065,
+      "step": 24364
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.556257142857143e-06,
+      "loss": 1.3671,
+      "step": 24365
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.555988571428572e-06,
+      "loss": 0.6852,
+      "step": 24366
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.555720000000001e-06,
+      "loss": 0.878,
+      "step": 24367
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.555451428571428e-06,
+      "loss": 1.4001,
+      "step": 24368
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.555182857142857e-06,
+      "loss": 0.9542,
+      "step": 24369
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.554914285714286e-06,
+      "loss": 1.4189,
+      "step": 24370
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.554645714285714e-06,
+      "loss": 0.9055,
+      "step": 24371
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.554377142857143e-06,
+      "loss": 1.1853,
+      "step": 24372
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5541085714285725e-06,
+      "loss": 1.1053,
+      "step": 24373
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.55384e-06,
+      "loss": 1.2842,
+      "step": 24374
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.553571428571429e-06,
+      "loss": 0.7682,
+      "step": 24375
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.553302857142858e-06,
+      "loss": 1.1864,
+      "step": 24376
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.553034285714287e-06,
+      "loss": 1.4428,
+      "step": 24377
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.552765714285714e-06,
+      "loss": 0.8089,
+      "step": 24378
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.552497142857143e-06,
+      "loss": 1.0984,
+      "step": 24379
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.552228571428572e-06,
+      "loss": 1.2609,
+      "step": 24380
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.55196e-06,
+      "loss": 0.9562,
+      "step": 24381
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.551691428571429e-06,
+      "loss": 1.1277,
+      "step": 24382
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.551422857142858e-06,
+      "loss": 0.9167,
+      "step": 24383
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.551154285714285e-06,
+      "loss": 0.9418,
+      "step": 24384
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.550885714285714e-06,
+      "loss": 1.2572,
+      "step": 24385
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.550617142857144e-06,
+      "loss": 1.1795,
+      "step": 24386
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.550348571428571e-06,
+      "loss": 1.5315,
+      "step": 24387
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.55008e-06,
+      "loss": 1.3596,
+      "step": 24388
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.549811428571429e-06,
+      "loss": 0.9046,
+      "step": 24389
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.549542857142858e-06,
+      "loss": 0.95,
+      "step": 24390
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.549274285714286e-06,
+      "loss": 1.0013,
+      "step": 24391
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.549005714285715e-06,
+      "loss": 1.5938,
+      "step": 24392
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.548737142857144e-06,
+      "loss": 1.4677,
+      "step": 24393
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.548468571428571e-06,
+      "loss": 1.5638,
+      "step": 24394
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.5482e-06,
+      "loss": 1.1329,
+      "step": 24395
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.547931428571429e-06,
+      "loss": 1.1671,
+      "step": 24396
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.547662857142856e-06,
+      "loss": 1.1923,
+      "step": 24397
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.547394285714285e-06,
+      "loss": 0.8817,
+      "step": 24398
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.547125714285715e-06,
+      "loss": 1.3619,
+      "step": 24399
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.546857142857144e-06,
+      "loss": 0.9423,
+      "step": 24400
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.546588571428572e-06,
+      "loss": 1.1348,
+      "step": 24401
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.546320000000001e-06,
+      "loss": 1.3135,
+      "step": 24402
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.54605142857143e-06,
+      "loss": 1.3043,
+      "step": 24403
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.545782857142857e-06,
+      "loss": 1.2768,
+      "step": 24404
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.545514285714286e-06,
+      "loss": 0.9938,
+      "step": 24405
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.545245714285715e-06,
+      "loss": 1.403,
+      "step": 24406
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.544977142857142e-06,
+      "loss": 1.3549,
+      "step": 24407
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.544708571428571e-06,
+      "loss": 1.3386,
+      "step": 24408
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.54444e-06,
+      "loss": 1.2374,
+      "step": 24409
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.544171428571428e-06,
+      "loss": 1.1972,
+      "step": 24410
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.543902857142857e-06,
+      "loss": 1.2493,
+      "step": 24411
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 7.543634285714287e-06,
+      "loss": 0.8628,
+      "step": 24412
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.543365714285716e-06,
+      "loss": 1.4696,
+      "step": 24413
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.543097142857143e-06,
+      "loss": 1.0397,
+      "step": 24414
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.542828571428572e-06,
+      "loss": 1.0341,
+      "step": 24415
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.542560000000001e-06,
+      "loss": 0.952,
+      "step": 24416
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.542291428571428e-06,
+      "loss": 1.042,
+      "step": 24417
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.542022857142857e-06,
+      "loss": 0.7395,
+      "step": 24418
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.541754285714286e-06,
+      "loss": 1.2778,
+      "step": 24419
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.541485714285714e-06,
+      "loss": 1.2633,
+      "step": 24420
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.541217142857143e-06,
+      "loss": 1.3138,
+      "step": 24421
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.540948571428572e-06,
+      "loss": 0.8988,
+      "step": 24422
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.540680000000001e-06,
+      "loss": 1.2387,
+      "step": 24423
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.540411428571428e-06,
+      "loss": 1.1458,
+      "step": 24424
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.540142857142857e-06,
+      "loss": 1.2038,
+      "step": 24425
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.539874285714287e-06,
+      "loss": 1.5118,
+      "step": 24426
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.539605714285714e-06,
+      "loss": 1.0236,
+      "step": 24427
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.539337142857143e-06,
+      "loss": 1.4386,
+      "step": 24428
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.539068571428572e-06,
+      "loss": 1.2804,
+      "step": 24429
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5388e-06,
+      "loss": 1.2145,
+      "step": 24430
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.538531428571429e-06,
+      "loss": 0.9588,
+      "step": 24431
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.538262857142858e-06,
+      "loss": 1.0622,
+      "step": 24432
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.537994285714285e-06,
+      "loss": 0.6437,
+      "step": 24433
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.537725714285714e-06,
+      "loss": 1.4004,
+      "step": 24434
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.537457142857143e-06,
+      "loss": 1.3829,
+      "step": 24435
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.537188571428572e-06,
+      "loss": 1.1802,
+      "step": 24436
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5369199999999995e-06,
+      "loss": 0.982,
+      "step": 24437
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5366514285714285e-06,
+      "loss": 1.4369,
+      "step": 24438
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.536382857142858e-06,
+      "loss": 0.9534,
+      "step": 24439
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.536114285714286e-06,
+      "loss": 0.8264,
+      "step": 24440
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.535845714285715e-06,
+      "loss": 1.0567,
+      "step": 24441
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.535577142857144e-06,
+      "loss": 0.9962,
+      "step": 24442
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.535308571428571e-06,
+      "loss": 1.6817,
+      "step": 24443
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.53504e-06,
+      "loss": 0.8755,
+      "step": 24444
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.534771428571429e-06,
+      "loss": 1.4826,
+      "step": 24445
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.534502857142858e-06,
+      "loss": 0.9249,
+      "step": 24446
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5342342857142855e-06,
+      "loss": 1.136,
+      "step": 24447
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5339657142857145e-06,
+      "loss": 1.4266,
+      "step": 24448
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5336971428571435e-06,
+      "loss": 1.6742,
+      "step": 24449
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.533428571428571e-06,
+      "loss": 1.4113,
+      "step": 24450
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.53316e-06,
+      "loss": 1.1267,
+      "step": 24451
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.53289142857143e-06,
+      "loss": 0.8528,
+      "step": 24452
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.532622857142857e-06,
+      "loss": 0.9401,
+      "step": 24453
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.532354285714286e-06,
+      "loss": 1.263,
+      "step": 24454
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.532085714285715e-06,
+      "loss": 0.8754,
+      "step": 24455
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5318171428571425e-06,
+      "loss": 1.2523,
+      "step": 24456
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5315485714285715e-06,
+      "loss": 1.2693,
+      "step": 24457
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5312800000000005e-06,
+      "loss": 1.4235,
+      "step": 24458
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5310114285714295e-06,
+      "loss": 1.2569,
+      "step": 24459
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.530742857142857e-06,
+      "loss": 1.5041,
+      "step": 24460
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.530474285714286e-06,
+      "loss": 1.0519,
+      "step": 24461
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.530205714285715e-06,
+      "loss": 1.14,
+      "step": 24462
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.529937142857142e-06,
+      "loss": 1.1513,
+      "step": 24463
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.529668571428571e-06,
+      "loss": 0.7661,
+      "step": 24464
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.529400000000001e-06,
+      "loss": 1.295,
+      "step": 24465
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5291314285714285e-06,
+      "loss": 1.1941,
+      "step": 24466
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5288628571428575e-06,
+      "loss": 1.2844,
+      "step": 24467
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5285942857142865e-06,
+      "loss": 1.0827,
+      "step": 24468
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5283257142857155e-06,
+      "loss": 1.4383,
+      "step": 24469
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.528057142857143e-06,
+      "loss": 1.1284,
+      "step": 24470
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.527788571428572e-06,
+      "loss": 1.2478,
+      "step": 24471
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.527520000000001e-06,
+      "loss": 0.9582,
+      "step": 24472
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.527251428571428e-06,
+      "loss": 1.1787,
+      "step": 24473
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.526982857142857e-06,
+      "loss": 1.3316,
+      "step": 24474
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.526714285714286e-06,
+      "loss": 1.3105,
+      "step": 24475
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.526445714285714e-06,
+      "loss": 0.9794,
+      "step": 24476
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.526177142857143e-06,
+      "loss": 1.1068,
+      "step": 24477
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.525908571428572e-06,
+      "loss": 1.2713,
+      "step": 24478
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.52564e-06,
+      "loss": 1.3499,
+      "step": 24479
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.525371428571429e-06,
+      "loss": 1.0629,
+      "step": 24480
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.525102857142858e-06,
+      "loss": 1.0095,
+      "step": 24481
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.524834285714287e-06,
+      "loss": 1.454,
+      "step": 24482
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.524565714285714e-06,
+      "loss": 1.2217,
+      "step": 24483
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.524297142857143e-06,
+      "loss": 1.3207,
+      "step": 24484
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.524028571428572e-06,
+      "loss": 1.1282,
+      "step": 24485
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.52376e-06,
+      "loss": 1.5434,
+      "step": 24486
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.523491428571429e-06,
+      "loss": 0.9742,
+      "step": 24487
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.523222857142858e-06,
+      "loss": 1.4485,
+      "step": 24488
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.522954285714285e-06,
+      "loss": 1.0701,
+      "step": 24489
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.522685714285714e-06,
+      "loss": 1.0211,
+      "step": 24490
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.522417142857143e-06,
+      "loss": 1.0962,
+      "step": 24491
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.522148571428573e-06,
+      "loss": 1.1478,
+      "step": 24492
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.52188e-06,
+      "loss": 0.743,
+      "step": 24493
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.521611428571429e-06,
+      "loss": 0.9355,
+      "step": 24494
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.521342857142858e-06,
+      "loss": 1.0722,
+      "step": 24495
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.521074285714286e-06,
+      "loss": 1.0219,
+      "step": 24496
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.520805714285715e-06,
+      "loss": 1.3644,
+      "step": 24497
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.520537142857144e-06,
+      "loss": 1.4819,
+      "step": 24498
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.520268571428571e-06,
+      "loss": 0.9286,
+      "step": 24499
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.52e-06,
+      "loss": 1.3584,
+      "step": 24500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.519731428571429e-06,
+      "loss": 1.0518,
+      "step": 24501
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.519462857142856e-06,
+      "loss": 1.0165,
+      "step": 24502
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.519194285714285e-06,
+      "loss": 1.0544,
+      "step": 24503
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.518925714285714e-06,
+      "loss": 0.6918,
+      "step": 24504
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.518657142857144e-06,
+      "loss": 1.5181,
+      "step": 24505
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.518388571428572e-06,
+      "loss": 1.3791,
+      "step": 24506
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.518120000000001e-06,
+      "loss": 0.9851,
+      "step": 24507
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.51785142857143e-06,
+      "loss": 1.0514,
+      "step": 24508
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.517582857142857e-06,
+      "loss": 1.4311,
+      "step": 24509
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.517314285714286e-06,
+      "loss": 1.646,
+      "step": 24510
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.517045714285715e-06,
+      "loss": 1.2566,
+      "step": 24511
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.516777142857142e-06,
+      "loss": 0.7601,
+      "step": 24512
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.516508571428571e-06,
+      "loss": 1.429,
+      "step": 24513
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.51624e-06,
+      "loss": 1.0781,
+      "step": 24514
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.515971428571428e-06,
+      "loss": 0.7965,
+      "step": 24515
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.515702857142857e-06,
+      "loss": 1.0548,
+      "step": 24516
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.515434285714286e-06,
+      "loss": 1.2905,
+      "step": 24517
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.515165714285716e-06,
+      "loss": 1.0908,
+      "step": 24518
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.514897142857143e-06,
+      "loss": 1.1686,
+      "step": 24519
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.514628571428572e-06,
+      "loss": 1.4577,
+      "step": 24520
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.514360000000001e-06,
+      "loss": 0.9111,
+      "step": 24521
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.514091428571428e-06,
+      "loss": 1.0188,
+      "step": 24522
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.513822857142857e-06,
+      "loss": 0.8985,
+      "step": 24523
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.513554285714286e-06,
+      "loss": 0.7945,
+      "step": 24524
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.513285714285714e-06,
+      "loss": 1.122,
+      "step": 24525
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.513017142857143e-06,
+      "loss": 1.2243,
+      "step": 24526
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.512748571428572e-06,
+      "loss": 1.2078,
+      "step": 24527
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.512480000000001e-06,
+      "loss": 1.2115,
+      "step": 24528
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.512211428571428e-06,
+      "loss": 0.7726,
+      "step": 24529
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.511942857142857e-06,
+      "loss": 1.3065,
+      "step": 24530
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.511674285714286e-06,
+      "loss": 1.1925,
+      "step": 24531
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.511405714285714e-06,
+      "loss": 1.3754,
+      "step": 24532
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.511137142857143e-06,
+      "loss": 0.8967,
+      "step": 24533
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.510868571428572e-06,
+      "loss": 1.42,
+      "step": 24534
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5106e-06,
+      "loss": 1.2788,
+      "step": 24535
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.510331428571429e-06,
+      "loss": 0.9838,
+      "step": 24536
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.510062857142858e-06,
+      "loss": 0.9093,
+      "step": 24537
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.509794285714285e-06,
+      "loss": 0.9977,
+      "step": 24538
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.509525714285714e-06,
+      "loss": 1.1739,
+      "step": 24539
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.509257142857143e-06,
+      "loss": 1.1762,
+      "step": 24540
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.508988571428572e-06,
+      "loss": 1.5588,
+      "step": 24541
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5087199999999995e-06,
+      "loss": 0.8836,
+      "step": 24542
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5084514285714285e-06,
+      "loss": 1.1187,
+      "step": 24543
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5081828571428575e-06,
+      "loss": 1.6526,
+      "step": 24544
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.507914285714286e-06,
+      "loss": 1.3506,
+      "step": 24545
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.507645714285715e-06,
+      "loss": 0.9607,
+      "step": 24546
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.507377142857144e-06,
+      "loss": 1.2772,
+      "step": 24547
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.507108571428571e-06,
+      "loss": 1.6817,
+      "step": 24548
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.50684e-06,
+      "loss": 0.8812,
+      "step": 24549
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.506571428571429e-06,
+      "loss": 0.8617,
+      "step": 24550
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.506302857142858e-06,
+      "loss": 1.2959,
+      "step": 24551
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5060342857142855e-06,
+      "loss": 1.1639,
+      "step": 24552
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5057657142857145e-06,
+      "loss": 1.1532,
+      "step": 24553
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5054971428571435e-06,
+      "loss": 1.1639,
+      "step": 24554
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.505228571428571e-06,
+      "loss": 1.0709,
+      "step": 24555
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.50496e-06,
+      "loss": 1.1704,
+      "step": 24556
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.504691428571429e-06,
+      "loss": 1.1252,
+      "step": 24557
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.504422857142857e-06,
+      "loss": 0.7894,
+      "step": 24558
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.504154285714286e-06,
+      "loss": 1.2275,
+      "step": 24559
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.503885714285715e-06,
+      "loss": 1.227,
+      "step": 24560
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5036171428571425e-06,
+      "loss": 1.6807,
+      "step": 24561
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5033485714285715e-06,
+      "loss": 1.1104,
+      "step": 24562
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5030800000000005e-06,
+      "loss": 1.0749,
+      "step": 24563
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5028114285714295e-06,
+      "loss": 1.0174,
+      "step": 24564
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.502542857142857e-06,
+      "loss": 1.346,
+      "step": 24565
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.502274285714286e-06,
+      "loss": 1.4016,
+      "step": 24566
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.502005714285715e-06,
+      "loss": 0.9534,
+      "step": 24567
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.501737142857142e-06,
+      "loss": 1.184,
+      "step": 24568
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.501468571428571e-06,
+      "loss": 1.0097,
+      "step": 24569
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5012e-06,
+      "loss": 1.1686,
+      "step": 24570
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5009314285714285e-06,
+      "loss": 0.9587,
+      "step": 24571
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5006628571428575e-06,
+      "loss": 1.222,
+      "step": 24572
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5003942857142865e-06,
+      "loss": 0.8631,
+      "step": 24573
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.5001257142857155e-06,
+      "loss": 1.2599,
+      "step": 24574
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.499857142857143e-06,
+      "loss": 1.2434,
+      "step": 24575
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.499588571428572e-06,
+      "loss": 1.3507,
+      "step": 24576
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.499320000000001e-06,
+      "loss": 1.1914,
+      "step": 24577
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.499051428571428e-06,
+      "loss": 0.8662,
+      "step": 24578
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.498782857142857e-06,
+      "loss": 0.9805,
+      "step": 24579
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.498514285714286e-06,
+      "loss": 1.474,
+      "step": 24580
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.498245714285714e-06,
+      "loss": 0.9375,
+      "step": 24581
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.497977142857143e-06,
+      "loss": 1.0884,
+      "step": 24582
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.497708571428572e-06,
+      "loss": 1.229,
+      "step": 24583
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.49744e-06,
+      "loss": 1.4817,
+      "step": 24584
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.497171428571429e-06,
+      "loss": 1.0728,
+      "step": 24585
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.496902857142858e-06,
+      "loss": 1.31,
+      "step": 24586
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 7.496634285714287e-06,
+      "loss": 1.0009,
+      "step": 24587
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.496365714285714e-06,
+      "loss": 1.0237,
+      "step": 24588
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.496097142857143e-06,
+      "loss": 1.3341,
+      "step": 24589
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.495828571428572e-06,
+      "loss": 1.0498,
+      "step": 24590
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.49556e-06,
+      "loss": 1.3799,
+      "step": 24591
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.495291428571429e-06,
+      "loss": 1.3036,
+      "step": 24592
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.495022857142858e-06,
+      "loss": 1.5852,
+      "step": 24593
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.494754285714285e-06,
+      "loss": 0.9699,
+      "step": 24594
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.494485714285714e-06,
+      "loss": 1.4229,
+      "step": 24595
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.494217142857143e-06,
+      "loss": 1.2,
+      "step": 24596
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.493948571428572e-06,
+      "loss": 0.8724,
+      "step": 24597
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.49368e-06,
+      "loss": 1.1869,
+      "step": 24598
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.493411428571429e-06,
+      "loss": 1.3307,
+      "step": 24599
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.493142857142858e-06,
+      "loss": 1.0307,
+      "step": 24600
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.492874285714286e-06,
+      "loss": 1.1262,
+      "step": 24601
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.492605714285715e-06,
+      "loss": 1.2687,
+      "step": 24602
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.492337142857144e-06,
+      "loss": 1.3761,
+      "step": 24603
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.492068571428571e-06,
+      "loss": 1.1685,
+      "step": 24604
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4918e-06,
+      "loss": 1.1589,
+      "step": 24605
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.491531428571429e-06,
+      "loss": 1.218,
+      "step": 24606
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.491262857142856e-06,
+      "loss": 1.159,
+      "step": 24607
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.490994285714285e-06,
+      "loss": 1.2059,
+      "step": 24608
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.490725714285714e-06,
+      "loss": 1.6455,
+      "step": 24609
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.490457142857143e-06,
+      "loss": 1.3904,
+      "step": 24610
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.490188571428572e-06,
+      "loss": 1.1856,
+      "step": 24611
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.489920000000001e-06,
+      "loss": 0.785,
+      "step": 24612
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.48965142857143e-06,
+      "loss": 1.5229,
+      "step": 24613
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.489382857142857e-06,
+      "loss": 0.9461,
+      "step": 24614
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.489114285714286e-06,
+      "loss": 1.0027,
+      "step": 24615
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.488845714285715e-06,
+      "loss": 1.4997,
+      "step": 24616
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.488577142857142e-06,
+      "loss": 1.217,
+      "step": 24617
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.488308571428571e-06,
+      "loss": 1.1083,
+      "step": 24618
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.48804e-06,
+      "loss": 1.136,
+      "step": 24619
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.487771428571429e-06,
+      "loss": 1.0983,
+      "step": 24620
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.487502857142857e-06,
+      "loss": 0.9568,
+      "step": 24621
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.487234285714286e-06,
+      "loss": 1.2895,
+      "step": 24622
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.486965714285715e-06,
+      "loss": 1.0392,
+      "step": 24623
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.486697142857143e-06,
+      "loss": 1.3384,
+      "step": 24624
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.486428571428572e-06,
+      "loss": 0.8106,
+      "step": 24625
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.486160000000001e-06,
+      "loss": 1.3401,
+      "step": 24626
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.485891428571428e-06,
+      "loss": 1.1981,
+      "step": 24627
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.485622857142857e-06,
+      "loss": 1.6307,
+      "step": 24628
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.485354285714286e-06,
+      "loss": 1.0784,
+      "step": 24629
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.485085714285714e-06,
+      "loss": 1.3265,
+      "step": 24630
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.484817142857143e-06,
+      "loss": 1.2744,
+      "step": 24631
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.484548571428572e-06,
+      "loss": 1.1085,
+      "step": 24632
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.484280000000001e-06,
+      "loss": 1.4718,
+      "step": 24633
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.484011428571428e-06,
+      "loss": 0.8845,
+      "step": 24634
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.483742857142857e-06,
+      "loss": 1.2048,
+      "step": 24635
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.483474285714286e-06,
+      "loss": 0.9266,
+      "step": 24636
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.483205714285714e-06,
+      "loss": 1.0332,
+      "step": 24637
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.482937142857143e-06,
+      "loss": 0.7837,
+      "step": 24638
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.482668571428572e-06,
+      "loss": 1.1469,
+      "step": 24639
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4824e-06,
+      "loss": 0.9975,
+      "step": 24640
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.482131428571429e-06,
+      "loss": 0.8797,
+      "step": 24641
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.481862857142858e-06,
+      "loss": 1.1584,
+      "step": 24642
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.481594285714287e-06,
+      "loss": 0.8308,
+      "step": 24643
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.481325714285714e-06,
+      "loss": 0.9035,
+      "step": 24644
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.481057142857143e-06,
+      "loss": 0.9826,
+      "step": 24645
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.480788571428572e-06,
+      "loss": 1.0503,
+      "step": 24646
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4805199999999995e-06,
+      "loss": 0.8789,
+      "step": 24647
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4802514285714285e-06,
+      "loss": 1.0219,
+      "step": 24648
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4799828571428575e-06,
+      "loss": 1.3284,
+      "step": 24649
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.479714285714286e-06,
+      "loss": 1.6315,
+      "step": 24650
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.479445714285715e-06,
+      "loss": 1.1065,
+      "step": 24651
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.479177142857144e-06,
+      "loss": 1.3377,
+      "step": 24652
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.478908571428571e-06,
+      "loss": 0.9352,
+      "step": 24653
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.47864e-06,
+      "loss": 1.506,
+      "step": 24654
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.478371428571429e-06,
+      "loss": 0.9476,
+      "step": 24655
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.478102857142858e-06,
+      "loss": 0.9215,
+      "step": 24656
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4778342857142855e-06,
+      "loss": 0.8497,
+      "step": 24657
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4775657142857145e-06,
+      "loss": 1.4321,
+      "step": 24658
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4772971428571435e-06,
+      "loss": 1.0414,
+      "step": 24659
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.477028571428571e-06,
+      "loss": 1.3225,
+      "step": 24660
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.47676e-06,
+      "loss": 0.95,
+      "step": 24661
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.476491428571429e-06,
+      "loss": 1.4351,
+      "step": 24662
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.476222857142857e-06,
+      "loss": 1.0813,
+      "step": 24663
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.475954285714286e-06,
+      "loss": 1.5499,
+      "step": 24664
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.475685714285715e-06,
+      "loss": 1.0189,
+      "step": 24665
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.475417142857144e-06,
+      "loss": 0.7557,
+      "step": 24666
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4751485714285715e-06,
+      "loss": 1.008,
+      "step": 24667
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4748800000000005e-06,
+      "loss": 0.9346,
+      "step": 24668
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4746114285714295e-06,
+      "loss": 1.1755,
+      "step": 24669
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.474342857142857e-06,
+      "loss": 1.3321,
+      "step": 24670
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.474074285714286e-06,
+      "loss": 1.5433,
+      "step": 24671
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.473805714285715e-06,
+      "loss": 1.0589,
+      "step": 24672
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.473537142857142e-06,
+      "loss": 1.2967,
+      "step": 24673
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.473268571428571e-06,
+      "loss": 1.3275,
+      "step": 24674
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.473e-06,
+      "loss": 1.4244,
+      "step": 24675
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4727314285714285e-06,
+      "loss": 1.4787,
+      "step": 24676
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4724628571428575e-06,
+      "loss": 1.1771,
+      "step": 24677
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4721942857142865e-06,
+      "loss": 1.0207,
+      "step": 24678
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4719257142857155e-06,
+      "loss": 1.2222,
+      "step": 24679
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.471657142857143e-06,
+      "loss": 1.0846,
+      "step": 24680
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.471388571428572e-06,
+      "loss": 1.3697,
+      "step": 24681
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.471120000000001e-06,
+      "loss": 0.9038,
+      "step": 24682
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.470851428571428e-06,
+      "loss": 1.1048,
+      "step": 24683
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.470582857142857e-06,
+      "loss": 1.1069,
+      "step": 24684
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.470314285714286e-06,
+      "loss": 0.9855,
+      "step": 24685
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.470045714285714e-06,
+      "loss": 0.7145,
+      "step": 24686
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.469777142857143e-06,
+      "loss": 1.1538,
+      "step": 24687
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.469508571428572e-06,
+      "loss": 1.2954,
+      "step": 24688
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.46924e-06,
+      "loss": 1.0084,
+      "step": 24689
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.468971428571429e-06,
+      "loss": 0.8401,
+      "step": 24690
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.468702857142858e-06,
+      "loss": 1.383,
+      "step": 24691
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.468434285714287e-06,
+      "loss": 1.0625,
+      "step": 24692
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.468165714285714e-06,
+      "loss": 1.1429,
+      "step": 24693
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.467897142857143e-06,
+      "loss": 1.1935,
+      "step": 24694
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.467628571428572e-06,
+      "loss": 1.5275,
+      "step": 24695
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.46736e-06,
+      "loss": 1.1479,
+      "step": 24696
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.467091428571429e-06,
+      "loss": 1.3285,
+      "step": 24697
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.466822857142858e-06,
+      "loss": 1.033,
+      "step": 24698
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.466554285714285e-06,
+      "loss": 1.45,
+      "step": 24699
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.466285714285714e-06,
+      "loss": 0.8357,
+      "step": 24700
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.466017142857143e-06,
+      "loss": 0.9102,
+      "step": 24701
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.465748571428572e-06,
+      "loss": 1.5853,
+      "step": 24702
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.46548e-06,
+      "loss": 1.0499,
+      "step": 24703
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.465211428571429e-06,
+      "loss": 1.3488,
+      "step": 24704
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.464942857142858e-06,
+      "loss": 0.8789,
+      "step": 24705
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.464674285714286e-06,
+      "loss": 1.2088,
+      "step": 24706
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.464405714285715e-06,
+      "loss": 0.8531,
+      "step": 24707
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.464137142857144e-06,
+      "loss": 1.1788,
+      "step": 24708
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.463868571428571e-06,
+      "loss": 1.2978,
+      "step": 24709
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4636e-06,
+      "loss": 1.0396,
+      "step": 24710
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.463331428571429e-06,
+      "loss": 1.0209,
+      "step": 24711
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.463062857142856e-06,
+      "loss": 1.3419,
+      "step": 24712
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.462794285714285e-06,
+      "loss": 0.9821,
+      "step": 24713
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.462525714285714e-06,
+      "loss": 0.8477,
+      "step": 24714
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.462257142857143e-06,
+      "loss": 1.3341,
+      "step": 24715
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.461988571428572e-06,
+      "loss": 0.8242,
+      "step": 24716
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.461720000000001e-06,
+      "loss": 1.1846,
+      "step": 24717
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.46145142857143e-06,
+      "loss": 0.8463,
+      "step": 24718
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.461182857142857e-06,
+      "loss": 1.1823,
+      "step": 24719
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.460914285714286e-06,
+      "loss": 1.2635,
+      "step": 24720
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.460645714285715e-06,
+      "loss": 1.3463,
+      "step": 24721
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.460377142857142e-06,
+      "loss": 1.2777,
+      "step": 24722
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.460108571428571e-06,
+      "loss": 0.8569,
+      "step": 24723
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.45984e-06,
+      "loss": 1.0741,
+      "step": 24724
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.459571428571429e-06,
+      "loss": 1.0432,
+      "step": 24725
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.459302857142857e-06,
+      "loss": 0.8431,
+      "step": 24726
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.459034285714286e-06,
+      "loss": 1.3796,
+      "step": 24727
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.458765714285715e-06,
+      "loss": 1.5237,
+      "step": 24728
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.458497142857143e-06,
+      "loss": 1.1131,
+      "step": 24729
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.458228571428572e-06,
+      "loss": 1.0776,
+      "step": 24730
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.457960000000001e-06,
+      "loss": 1.4142,
+      "step": 24731
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.457691428571428e-06,
+      "loss": 1.223,
+      "step": 24732
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.457422857142857e-06,
+      "loss": 1.2293,
+      "step": 24733
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.457154285714286e-06,
+      "loss": 0.9005,
+      "step": 24734
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.456885714285714e-06,
+      "loss": 1.1936,
+      "step": 24735
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.456617142857143e-06,
+      "loss": 0.9282,
+      "step": 24736
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.456348571428572e-06,
+      "loss": 1.8705,
+      "step": 24737
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.456080000000001e-06,
+      "loss": 1.2546,
+      "step": 24738
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.455811428571428e-06,
+      "loss": 1.2784,
+      "step": 24739
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.455542857142857e-06,
+      "loss": 1.1773,
+      "step": 24740
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.455274285714286e-06,
+      "loss": 1.0999,
+      "step": 24741
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.455005714285714e-06,
+      "loss": 1.3367,
+      "step": 24742
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.454737142857143e-06,
+      "loss": 1.4006,
+      "step": 24743
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.454468571428572e-06,
+      "loss": 1.1834,
+      "step": 24744
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4542e-06,
+      "loss": 1.5044,
+      "step": 24745
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.453931428571429e-06,
+      "loss": 1.5223,
+      "step": 24746
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.453662857142858e-06,
+      "loss": 0.9312,
+      "step": 24747
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.453394285714287e-06,
+      "loss": 1.2329,
+      "step": 24748
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.453125714285714e-06,
+      "loss": 1.0338,
+      "step": 24749
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.452857142857143e-06,
+      "loss": 1.1501,
+      "step": 24750
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.452588571428572e-06,
+      "loss": 1.6367,
+      "step": 24751
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4523199999999995e-06,
+      "loss": 1.4636,
+      "step": 24752
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4520514285714285e-06,
+      "loss": 1.4862,
+      "step": 24753
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4517828571428575e-06,
+      "loss": 1.0561,
+      "step": 24754
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.451514285714286e-06,
+      "loss": 1.1195,
+      "step": 24755
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.451245714285715e-06,
+      "loss": 1.614,
+      "step": 24756
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.450977142857144e-06,
+      "loss": 1.0846,
+      "step": 24757
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.450708571428571e-06,
+      "loss": 1.4945,
+      "step": 24758
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.45044e-06,
+      "loss": 1.6062,
+      "step": 24759
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.450171428571429e-06,
+      "loss": 1.311,
+      "step": 24760
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.449902857142858e-06,
+      "loss": 0.9508,
+      "step": 24761
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.4496342857142855e-06,
+      "loss": 1.3679,
+      "step": 24762
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4493657142857145e-06,
+      "loss": 1.3167,
+      "step": 24763
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4490971428571435e-06,
+      "loss": 1.2624,
+      "step": 24764
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.448828571428571e-06,
+      "loss": 1.093,
+      "step": 24765
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.44856e-06,
+      "loss": 1.2737,
+      "step": 24766
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.448291428571429e-06,
+      "loss": 0.8466,
+      "step": 24767
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.448022857142857e-06,
+      "loss": 1.3377,
+      "step": 24768
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.447754285714286e-06,
+      "loss": 1.2366,
+      "step": 24769
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.447485714285715e-06,
+      "loss": 1.0999,
+      "step": 24770
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.447217142857144e-06,
+      "loss": 1.2505,
+      "step": 24771
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4469485714285715e-06,
+      "loss": 0.8521,
+      "step": 24772
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4466800000000005e-06,
+      "loss": 1.1055,
+      "step": 24773
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4464114285714295e-06,
+      "loss": 0.9095,
+      "step": 24774
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.446142857142857e-06,
+      "loss": 0.8905,
+      "step": 24775
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.445874285714286e-06,
+      "loss": 1.0255,
+      "step": 24776
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.445605714285715e-06,
+      "loss": 1.186,
+      "step": 24777
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.445337142857142e-06,
+      "loss": 1.1509,
+      "step": 24778
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.445068571428571e-06,
+      "loss": 1.4543,
+      "step": 24779
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4448e-06,
+      "loss": 0.7356,
+      "step": 24780
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4445314285714285e-06,
+      "loss": 0.8616,
+      "step": 24781
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4442628571428575e-06,
+      "loss": 0.6811,
+      "step": 24782
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4439942857142865e-06,
+      "loss": 1.042,
+      "step": 24783
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4437257142857155e-06,
+      "loss": 1.1675,
+      "step": 24784
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.443457142857143e-06,
+      "loss": 1.1789,
+      "step": 24785
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.443188571428572e-06,
+      "loss": 0.6398,
+      "step": 24786
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.442920000000001e-06,
+      "loss": 1.5492,
+      "step": 24787
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.442651428571428e-06,
+      "loss": 1.0738,
+      "step": 24788
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.442382857142857e-06,
+      "loss": 1.5725,
+      "step": 24789
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.442114285714286e-06,
+      "loss": 1.3335,
+      "step": 24790
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.441845714285714e-06,
+      "loss": 1.4305,
+      "step": 24791
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.441577142857143e-06,
+      "loss": 1.295,
+      "step": 24792
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.441308571428572e-06,
+      "loss": 1.2357,
+      "step": 24793
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.441040000000001e-06,
+      "loss": 1.1599,
+      "step": 24794
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.440771428571429e-06,
+      "loss": 1.0448,
+      "step": 24795
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.440502857142858e-06,
+      "loss": 1.7182,
+      "step": 24796
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.440234285714287e-06,
+      "loss": 1.336,
+      "step": 24797
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.439965714285714e-06,
+      "loss": 0.9845,
+      "step": 24798
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.439697142857143e-06,
+      "loss": 1.4481,
+      "step": 24799
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.439428571428572e-06,
+      "loss": 1.4883,
+      "step": 24800
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.43916e-06,
+      "loss": 1.2059,
+      "step": 24801
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.438891428571429e-06,
+      "loss": 0.6885,
+      "step": 24802
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.438622857142858e-06,
+      "loss": 1.1838,
+      "step": 24803
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.438354285714285e-06,
+      "loss": 0.9374,
+      "step": 24804
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.438085714285714e-06,
+      "loss": 1.4725,
+      "step": 24805
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.437817142857143e-06,
+      "loss": 0.8418,
+      "step": 24806
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.437548571428572e-06,
+      "loss": 1.6524,
+      "step": 24807
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.43728e-06,
+      "loss": 1.0498,
+      "step": 24808
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.437011428571429e-06,
+      "loss": 1.0341,
+      "step": 24809
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.436742857142858e-06,
+      "loss": 1.1272,
+      "step": 24810
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.436474285714286e-06,
+      "loss": 1.0947,
+      "step": 24811
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.436205714285715e-06,
+      "loss": 0.7524,
+      "step": 24812
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.435937142857144e-06,
+      "loss": 0.8979,
+      "step": 24813
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.435668571428571e-06,
+      "loss": 1.188,
+      "step": 24814
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4354e-06,
+      "loss": 1.0588,
+      "step": 24815
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.435131428571429e-06,
+      "loss": 0.8994,
+      "step": 24816
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.434862857142858e-06,
+      "loss": 1.1106,
+      "step": 24817
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.434594285714285e-06,
+      "loss": 1.0456,
+      "step": 24818
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.434325714285714e-06,
+      "loss": 1.0495,
+      "step": 24819
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.434057142857143e-06,
+      "loss": 1.0368,
+      "step": 24820
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.433788571428572e-06,
+      "loss": 1.4957,
+      "step": 24821
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.433520000000001e-06,
+      "loss": 0.914,
+      "step": 24822
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.43325142857143e-06,
+      "loss": 1.3508,
+      "step": 24823
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.432982857142857e-06,
+      "loss": 1.2096,
+      "step": 24824
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.432714285714286e-06,
+      "loss": 1.1698,
+      "step": 24825
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.432445714285715e-06,
+      "loss": 0.9974,
+      "step": 24826
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.432177142857142e-06,
+      "loss": 0.907,
+      "step": 24827
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.431908571428571e-06,
+      "loss": 1.3137,
+      "step": 24828
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.43164e-06,
+      "loss": 0.9589,
+      "step": 24829
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.431371428571429e-06,
+      "loss": 0.6752,
+      "step": 24830
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.431102857142857e-06,
+      "loss": 1.0558,
+      "step": 24831
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.430834285714286e-06,
+      "loss": 1.2529,
+      "step": 24832
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.430565714285715e-06,
+      "loss": 1.1228,
+      "step": 24833
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.430297142857143e-06,
+      "loss": 1.4503,
+      "step": 24834
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.430028571428572e-06,
+      "loss": 1.1337,
+      "step": 24835
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.429760000000001e-06,
+      "loss": 0.9443,
+      "step": 24836
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.429491428571428e-06,
+      "loss": 0.9638,
+      "step": 24837
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.429222857142857e-06,
+      "loss": 1.0146,
+      "step": 24838
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.428954285714286e-06,
+      "loss": 1.4789,
+      "step": 24839
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.428685714285714e-06,
+      "loss": 1.3176,
+      "step": 24840
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.428417142857143e-06,
+      "loss": 1.0841,
+      "step": 24841
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.428148571428572e-06,
+      "loss": 1.3778,
+      "step": 24842
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.427880000000001e-06,
+      "loss": 0.8709,
+      "step": 24843
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.427611428571428e-06,
+      "loss": 1.089,
+      "step": 24844
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.427342857142857e-06,
+      "loss": 1.3811,
+      "step": 24845
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.427074285714286e-06,
+      "loss": 1.2779,
+      "step": 24846
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.426805714285714e-06,
+      "loss": 1.3787,
+      "step": 24847
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.426537142857143e-06,
+      "loss": 1.2225,
+      "step": 24848
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.426268571428572e-06,
+      "loss": 1.2938,
+      "step": 24849
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.426e-06,
+      "loss": 1.1736,
+      "step": 24850
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.425731428571429e-06,
+      "loss": 0.9581,
+      "step": 24851
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.425462857142858e-06,
+      "loss": 0.789,
+      "step": 24852
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.425194285714287e-06,
+      "loss": 1.2956,
+      "step": 24853
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.424925714285714e-06,
+      "loss": 1.1074,
+      "step": 24854
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.424657142857143e-06,
+      "loss": 1.0659,
+      "step": 24855
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.424388571428572e-06,
+      "loss": 1.2786,
+      "step": 24856
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4241199999999995e-06,
+      "loss": 1.4442,
+      "step": 24857
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4238514285714285e-06,
+      "loss": 1.2194,
+      "step": 24858
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4235828571428575e-06,
+      "loss": 1.1641,
+      "step": 24859
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.423314285714286e-06,
+      "loss": 1.1293,
+      "step": 24860
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.423045714285715e-06,
+      "loss": 1.3588,
+      "step": 24861
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.422777142857144e-06,
+      "loss": 1.2222,
+      "step": 24862
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.422508571428571e-06,
+      "loss": 1.1885,
+      "step": 24863
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.42224e-06,
+      "loss": 1.5532,
+      "step": 24864
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.421971428571429e-06,
+      "loss": 1.3946,
+      "step": 24865
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.421702857142858e-06,
+      "loss": 1.2005,
+      "step": 24866
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4214342857142855e-06,
+      "loss": 1.2077,
+      "step": 24867
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4211657142857145e-06,
+      "loss": 0.6975,
+      "step": 24868
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4208971428571435e-06,
+      "loss": 1.2947,
+      "step": 24869
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.420628571428571e-06,
+      "loss": 0.9931,
+      "step": 24870
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.42036e-06,
+      "loss": 1.061,
+      "step": 24871
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.420091428571429e-06,
+      "loss": 1.3966,
+      "step": 24872
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.419822857142857e-06,
+      "loss": 0.9205,
+      "step": 24873
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.419554285714286e-06,
+      "loss": 1.3749,
+      "step": 24874
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.419285714285715e-06,
+      "loss": 1.3802,
+      "step": 24875
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.419017142857144e-06,
+      "loss": 1.1304,
+      "step": 24876
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4187485714285715e-06,
+      "loss": 1.1407,
+      "step": 24877
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4184800000000005e-06,
+      "loss": 0.9661,
+      "step": 24878
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4182114285714295e-06,
+      "loss": 0.9562,
+      "step": 24879
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.417942857142857e-06,
+      "loss": 1.2305,
+      "step": 24880
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.417674285714286e-06,
+      "loss": 1.016,
+      "step": 24881
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.417405714285715e-06,
+      "loss": 1.6438,
+      "step": 24882
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.417137142857142e-06,
+      "loss": 1.5255,
+      "step": 24883
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.416868571428571e-06,
+      "loss": 0.933,
+      "step": 24884
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4166e-06,
+      "loss": 1.1117,
+      "step": 24885
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4163314285714285e-06,
+      "loss": 1.0198,
+      "step": 24886
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4160628571428575e-06,
+      "loss": 1.1272,
+      "step": 24887
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4157942857142865e-06,
+      "loss": 1.2966,
+      "step": 24888
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4155257142857155e-06,
+      "loss": 1.17,
+      "step": 24889
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.415257142857143e-06,
+      "loss": 1.1794,
+      "step": 24890
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.414988571428572e-06,
+      "loss": 1.3401,
+      "step": 24891
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.414720000000001e-06,
+      "loss": 1.0236,
+      "step": 24892
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.414451428571428e-06,
+      "loss": 1.2591,
+      "step": 24893
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.414182857142857e-06,
+      "loss": 1.2915,
+      "step": 24894
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.413914285714286e-06,
+      "loss": 1.4109,
+      "step": 24895
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.413645714285714e-06,
+      "loss": 1.3431,
+      "step": 24896
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.413377142857143e-06,
+      "loss": 1.047,
+      "step": 24897
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.413108571428572e-06,
+      "loss": 1.3695,
+      "step": 24898
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.412840000000001e-06,
+      "loss": 1.5515,
+      "step": 24899
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.412571428571429e-06,
+      "loss": 1.212,
+      "step": 24900
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.412302857142858e-06,
+      "loss": 0.9456,
+      "step": 24901
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.412034285714287e-06,
+      "loss": 1.1788,
+      "step": 24902
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.411765714285714e-06,
+      "loss": 1.2004,
+      "step": 24903
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.411497142857143e-06,
+      "loss": 1.3722,
+      "step": 24904
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.411228571428572e-06,
+      "loss": 1.3843,
+      "step": 24905
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.41096e-06,
+      "loss": 0.9008,
+      "step": 24906
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.410691428571429e-06,
+      "loss": 1.1911,
+      "step": 24907
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.410422857142858e-06,
+      "loss": 1.4386,
+      "step": 24908
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.410154285714285e-06,
+      "loss": 1.2419,
+      "step": 24909
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.409885714285714e-06,
+      "loss": 1.5524,
+      "step": 24910
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.409617142857143e-06,
+      "loss": 1.2413,
+      "step": 24911
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.409348571428572e-06,
+      "loss": 0.9252,
+      "step": 24912
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.40908e-06,
+      "loss": 1.5798,
+      "step": 24913
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.408811428571429e-06,
+      "loss": 1.1608,
+      "step": 24914
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.408542857142858e-06,
+      "loss": 1.6134,
+      "step": 24915
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.408274285714286e-06,
+      "loss": 1.4372,
+      "step": 24916
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.408005714285715e-06,
+      "loss": 1.2447,
+      "step": 24917
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.407737142857144e-06,
+      "loss": 1.4966,
+      "step": 24918
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.407468571428571e-06,
+      "loss": 1.1394,
+      "step": 24919
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.4072e-06,
+      "loss": 1.0349,
+      "step": 24920
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.406931428571429e-06,
+      "loss": 0.9496,
+      "step": 24921
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.406662857142858e-06,
+      "loss": 1.1635,
+      "step": 24922
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.406394285714285e-06,
+      "loss": 0.7223,
+      "step": 24923
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.406125714285714e-06,
+      "loss": 1.5257,
+      "step": 24924
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.405857142857143e-06,
+      "loss": 1.206,
+      "step": 24925
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.405588571428572e-06,
+      "loss": 0.8475,
+      "step": 24926
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.405320000000001e-06,
+      "loss": 1.0423,
+      "step": 24927
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.40505142857143e-06,
+      "loss": 1.1779,
+      "step": 24928
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.404782857142857e-06,
+      "loss": 1.4274,
+      "step": 24929
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.404514285714286e-06,
+      "loss": 0.7094,
+      "step": 24930
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.404245714285715e-06,
+      "loss": 0.9332,
+      "step": 24931
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.403977142857142e-06,
+      "loss": 0.9262,
+      "step": 24932
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.403708571428571e-06,
+      "loss": 1.5795,
+      "step": 24933
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.40344e-06,
+      "loss": 1.031,
+      "step": 24934
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.403171428571429e-06,
+      "loss": 1.6393,
+      "step": 24935
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.402902857142857e-06,
+      "loss": 0.8507,
+      "step": 24936
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.402634285714286e-06,
+      "loss": 0.8369,
+      "step": 24937
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.402365714285715e-06,
+      "loss": 1.2858,
+      "step": 24938
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.402097142857143e-06,
+      "loss": 1.0427,
+      "step": 24939
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.401828571428572e-06,
+      "loss": 1.114,
+      "step": 24940
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.401560000000001e-06,
+      "loss": 0.9305,
+      "step": 24941
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.401291428571428e-06,
+      "loss": 1.1803,
+      "step": 24942
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.401022857142857e-06,
+      "loss": 1.4232,
+      "step": 24943
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.400754285714286e-06,
+      "loss": 1.3951,
+      "step": 24944
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.400485714285715e-06,
+      "loss": 0.9891,
+      "step": 24945
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.400217142857143e-06,
+      "loss": 1.1189,
+      "step": 24946
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.399948571428572e-06,
+      "loss": 1.0581,
+      "step": 24947
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.399680000000001e-06,
+      "loss": 0.9829,
+      "step": 24948
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.399411428571428e-06,
+      "loss": 1.068,
+      "step": 24949
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.399142857142857e-06,
+      "loss": 1.1907,
+      "step": 24950
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.398874285714286e-06,
+      "loss": 1.676,
+      "step": 24951
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.398605714285714e-06,
+      "loss": 1.1503,
+      "step": 24952
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.398337142857143e-06,
+      "loss": 0.9471,
+      "step": 24953
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.398068571428572e-06,
+      "loss": 1.0687,
+      "step": 24954
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3978e-06,
+      "loss": 1.2392,
+      "step": 24955
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.397531428571429e-06,
+      "loss": 1.6515,
+      "step": 24956
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.397262857142858e-06,
+      "loss": 1.0883,
+      "step": 24957
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.396994285714287e-06,
+      "loss": 0.8132,
+      "step": 24958
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.396725714285714e-06,
+      "loss": 1.2841,
+      "step": 24959
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.396457142857143e-06,
+      "loss": 1.3729,
+      "step": 24960
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.396188571428572e-06,
+      "loss": 1.4764,
+      "step": 24961
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3959199999999995e-06,
+      "loss": 1.1439,
+      "step": 24962
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3956514285714285e-06,
+      "loss": 0.782,
+      "step": 24963
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3953828571428575e-06,
+      "loss": 1.4723,
+      "step": 24964
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.395114285714286e-06,
+      "loss": 1.4902,
+      "step": 24965
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.394845714285715e-06,
+      "loss": 0.8764,
+      "step": 24966
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.394577142857144e-06,
+      "loss": 0.8931,
+      "step": 24967
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.394308571428573e-06,
+      "loss": 1.1268,
+      "step": 24968
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.39404e-06,
+      "loss": 1.1733,
+      "step": 24969
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.393771428571429e-06,
+      "loss": 0.9921,
+      "step": 24970
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.393502857142858e-06,
+      "loss": 1.3761,
+      "step": 24971
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3932342857142855e-06,
+      "loss": 0.9948,
+      "step": 24972
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3929657142857145e-06,
+      "loss": 1.2165,
+      "step": 24973
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3926971428571435e-06,
+      "loss": 1.4935,
+      "step": 24974
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.392428571428571e-06,
+      "loss": 1.2356,
+      "step": 24975
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.39216e-06,
+      "loss": 0.9974,
+      "step": 24976
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.391891428571429e-06,
+      "loss": 1.2334,
+      "step": 24977
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.391622857142857e-06,
+      "loss": 1.2963,
+      "step": 24978
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.391354285714286e-06,
+      "loss": 1.0876,
+      "step": 24979
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.391085714285715e-06,
+      "loss": 1.3663,
+      "step": 24980
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.390817142857144e-06,
+      "loss": 1.0664,
+      "step": 24981
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3905485714285715e-06,
+      "loss": 1.5931,
+      "step": 24982
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3902800000000005e-06,
+      "loss": 0.8352,
+      "step": 24983
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3900114285714295e-06,
+      "loss": 0.9016,
+      "step": 24984
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.389742857142857e-06,
+      "loss": 1.4215,
+      "step": 24985
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.389474285714286e-06,
+      "loss": 1.0629,
+      "step": 24986
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.389205714285715e-06,
+      "loss": 1.4387,
+      "step": 24987
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.388937142857142e-06,
+      "loss": 1.4222,
+      "step": 24988
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.388668571428571e-06,
+      "loss": 1.5896,
+      "step": 24989
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3884e-06,
+      "loss": 1.2951,
+      "step": 24990
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.388131428571429e-06,
+      "loss": 0.8336,
+      "step": 24991
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3878628571428575e-06,
+      "loss": 1.1157,
+      "step": 24992
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3875942857142865e-06,
+      "loss": 1.1312,
+      "step": 24993
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3873257142857155e-06,
+      "loss": 1.2609,
+      "step": 24994
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.387057142857143e-06,
+      "loss": 1.5413,
+      "step": 24995
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.386788571428572e-06,
+      "loss": 0.9469,
+      "step": 24996
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.386520000000001e-06,
+      "loss": 1.2389,
+      "step": 24997
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.386251428571428e-06,
+      "loss": 1.3483,
+      "step": 24998
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.385982857142857e-06,
+      "loss": 0.9577,
+      "step": 24999
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.385714285714286e-06,
+      "loss": 1.0361,
+      "step": 25000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.385445714285714e-06,
+      "loss": 0.9976,
+      "step": 25001
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.385177142857143e-06,
+      "loss": 1.3316,
+      "step": 25002
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.384908571428572e-06,
+      "loss": 1.3286,
+      "step": 25003
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.384640000000001e-06,
+      "loss": 1.355,
+      "step": 25004
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.384371428571429e-06,
+      "loss": 1.2838,
+      "step": 25005
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.384102857142858e-06,
+      "loss": 1.2322,
+      "step": 25006
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.383834285714287e-06,
+      "loss": 0.9309,
+      "step": 25007
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.383565714285714e-06,
+      "loss": 1.3659,
+      "step": 25008
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.383297142857143e-06,
+      "loss": 1.2179,
+      "step": 25009
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.383028571428572e-06,
+      "loss": 1.1398,
+      "step": 25010
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.38276e-06,
+      "loss": 0.6334,
+      "step": 25011
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.382491428571429e-06,
+      "loss": 1.1873,
+      "step": 25012
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.382222857142858e-06,
+      "loss": 1.3506,
+      "step": 25013
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.381954285714285e-06,
+      "loss": 1.0255,
+      "step": 25014
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.381685714285714e-06,
+      "loss": 0.7924,
+      "step": 25015
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.381417142857143e-06,
+      "loss": 1.2154,
+      "step": 25016
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.381148571428572e-06,
+      "loss": 1.4343,
+      "step": 25017
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.38088e-06,
+      "loss": 1.1186,
+      "step": 25018
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.380611428571429e-06,
+      "loss": 0.9867,
+      "step": 25019
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.380342857142858e-06,
+      "loss": 0.9519,
+      "step": 25020
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.380074285714286e-06,
+      "loss": 0.8402,
+      "step": 25021
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.379805714285715e-06,
+      "loss": 1.5812,
+      "step": 25022
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.379537142857144e-06,
+      "loss": 1.3488,
+      "step": 25023
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.379268571428571e-06,
+      "loss": 0.9036,
+      "step": 25024
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.379e-06,
+      "loss": 1.8846,
+      "step": 25025
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.378731428571429e-06,
+      "loss": 1.4791,
+      "step": 25026
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.378462857142858e-06,
+      "loss": 1.1493,
+      "step": 25027
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.378194285714285e-06,
+      "loss": 1.3073,
+      "step": 25028
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.377925714285714e-06,
+      "loss": 1.0013,
+      "step": 25029
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.377657142857143e-06,
+      "loss": 1.1479,
+      "step": 25030
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.377388571428572e-06,
+      "loss": 1.2131,
+      "step": 25031
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.377120000000001e-06,
+      "loss": 1.3308,
+      "step": 25032
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.37685142857143e-06,
+      "loss": 1.4518,
+      "step": 25033
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.376582857142857e-06,
+      "loss": 1.7873,
+      "step": 25034
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.376314285714286e-06,
+      "loss": 1.2344,
+      "step": 25035
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.376045714285715e-06,
+      "loss": 1.5432,
+      "step": 25036
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.375777142857142e-06,
+      "loss": 1.2345,
+      "step": 25037
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.375508571428571e-06,
+      "loss": 1.5967,
+      "step": 25038
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.37524e-06,
+      "loss": 1.2938,
+      "step": 25039
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.374971428571429e-06,
+      "loss": 0.9034,
+      "step": 25040
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.374702857142857e-06,
+      "loss": 0.9879,
+      "step": 25041
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.374434285714286e-06,
+      "loss": 1.4333,
+      "step": 25042
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.374165714285715e-06,
+      "loss": 1.3419,
+      "step": 25043
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.373897142857143e-06,
+      "loss": 1.2066,
+      "step": 25044
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.373628571428572e-06,
+      "loss": 1.1943,
+      "step": 25045
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.373360000000001e-06,
+      "loss": 0.9723,
+      "step": 25046
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.373091428571428e-06,
+      "loss": 1.1006,
+      "step": 25047
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.372822857142857e-06,
+      "loss": 1.0682,
+      "step": 25048
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.372554285714286e-06,
+      "loss": 1.3686,
+      "step": 25049
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.372285714285715e-06,
+      "loss": 1.3693,
+      "step": 25050
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.372017142857143e-06,
+      "loss": 1.3347,
+      "step": 25051
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.371748571428572e-06,
+      "loss": 1.1678,
+      "step": 25052
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.371480000000001e-06,
+      "loss": 0.9403,
+      "step": 25053
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.371211428571428e-06,
+      "loss": 1.1118,
+      "step": 25054
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.370942857142857e-06,
+      "loss": 1.274,
+      "step": 25055
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.370674285714286e-06,
+      "loss": 0.9699,
+      "step": 25056
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.370405714285714e-06,
+      "loss": 1.0001,
+      "step": 25057
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.370137142857143e-06,
+      "loss": 1.3414,
+      "step": 25058
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.369868571428572e-06,
+      "loss": 1.2187,
+      "step": 25059
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3696e-06,
+      "loss": 0.7936,
+      "step": 25060
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.369331428571429e-06,
+      "loss": 1.6332,
+      "step": 25061
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.369062857142858e-06,
+      "loss": 1.1182,
+      "step": 25062
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.368794285714287e-06,
+      "loss": 0.9403,
+      "step": 25063
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.368525714285714e-06,
+      "loss": 1.2627,
+      "step": 25064
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.368257142857143e-06,
+      "loss": 0.6603,
+      "step": 25065
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.367988571428572e-06,
+      "loss": 1.2761,
+      "step": 25066
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3677199999999995e-06,
+      "loss": 0.9999,
+      "step": 25067
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3674514285714285e-06,
+      "loss": 0.7677,
+      "step": 25068
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3671828571428575e-06,
+      "loss": 1.6237,
+      "step": 25069
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.366914285714286e-06,
+      "loss": 1.0823,
+      "step": 25070
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.366645714285715e-06,
+      "loss": 1.1945,
+      "step": 25071
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.366377142857144e-06,
+      "loss": 0.7682,
+      "step": 25072
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.366108571428573e-06,
+      "loss": 1.3045,
+      "step": 25073
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.36584e-06,
+      "loss": 0.9571,
+      "step": 25074
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.365571428571429e-06,
+      "loss": 0.8305,
+      "step": 25075
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.365302857142858e-06,
+      "loss": 1.1385,
+      "step": 25076
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3650342857142855e-06,
+      "loss": 1.21,
+      "step": 25077
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3647657142857145e-06,
+      "loss": 1.0603,
+      "step": 25078
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3644971428571435e-06,
+      "loss": 0.9564,
+      "step": 25079
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.364228571428571e-06,
+      "loss": 1.4211,
+      "step": 25080
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.36396e-06,
+      "loss": 0.8727,
+      "step": 25081
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.363691428571429e-06,
+      "loss": 0.9273,
+      "step": 25082
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.363422857142857e-06,
+      "loss": 0.8722,
+      "step": 25083
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.363154285714286e-06,
+      "loss": 1.3367,
+      "step": 25084
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.362885714285715e-06,
+      "loss": 1.2826,
+      "step": 25085
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.362617142857144e-06,
+      "loss": 1.0565,
+      "step": 25086
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3623485714285715e-06,
+      "loss": 0.8846,
+      "step": 25087
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3620800000000005e-06,
+      "loss": 0.8287,
+      "step": 25088
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3618114285714295e-06,
+      "loss": 1.1201,
+      "step": 25089
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.361542857142857e-06,
+      "loss": 0.9744,
+      "step": 25090
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.361274285714286e-06,
+      "loss": 1.0304,
+      "step": 25091
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.361005714285715e-06,
+      "loss": 1.1906,
+      "step": 25092
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.360737142857142e-06,
+      "loss": 1.1609,
+      "step": 25093
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.360468571428571e-06,
+      "loss": 1.3013,
+      "step": 25094
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3602e-06,
+      "loss": 0.9184,
+      "step": 25095
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.359931428571429e-06,
+      "loss": 1.3219,
+      "step": 25096
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3596628571428575e-06,
+      "loss": 1.1589,
+      "step": 25097
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3593942857142865e-06,
+      "loss": 1.3418,
+      "step": 25098
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.3591257142857155e-06,
+      "loss": 1.5794,
+      "step": 25099
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.358857142857143e-06,
+      "loss": 1.2826,
+      "step": 25100
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.358588571428572e-06,
+      "loss": 1.2908,
+      "step": 25101
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.358320000000001e-06,
+      "loss": 1.3535,
+      "step": 25102
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.358051428571428e-06,
+      "loss": 1.0607,
+      "step": 25103
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.357782857142857e-06,
+      "loss": 1.1641,
+      "step": 25104
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.357514285714286e-06,
+      "loss": 0.6943,
+      "step": 25105
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.357245714285714e-06,
+      "loss": 1.2518,
+      "step": 25106
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.356977142857143e-06,
+      "loss": 1.3454,
+      "step": 25107
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.356708571428572e-06,
+      "loss": 1.3808,
+      "step": 25108
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.356440000000001e-06,
+      "loss": 1.2032,
+      "step": 25109
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.356171428571429e-06,
+      "loss": 1.3701,
+      "step": 25110
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.355902857142858e-06,
+      "loss": 1.1765,
+      "step": 25111
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 7.355634285714287e-06,
+      "loss": 1.172,
+      "step": 25112
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.355365714285714e-06,
+      "loss": 1.3339,
+      "step": 25113
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.355097142857143e-06,
+      "loss": 1.4495,
+      "step": 25114
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.354828571428572e-06,
+      "loss": 1.064,
+      "step": 25115
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.35456e-06,
+      "loss": 1.1647,
+      "step": 25116
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.354291428571429e-06,
+      "loss": 1.0647,
+      "step": 25117
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.354022857142858e-06,
+      "loss": 1.5819,
+      "step": 25118
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.353754285714287e-06,
+      "loss": 1.0627,
+      "step": 25119
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.353485714285714e-06,
+      "loss": 1.4491,
+      "step": 25120
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.353217142857143e-06,
+      "loss": 0.776,
+      "step": 25121
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.352948571428572e-06,
+      "loss": 0.7981,
+      "step": 25122
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.35268e-06,
+      "loss": 1.1512,
+      "step": 25123
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.352411428571429e-06,
+      "loss": 1.4187,
+      "step": 25124
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.352142857142858e-06,
+      "loss": 0.8979,
+      "step": 25125
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.351874285714286e-06,
+      "loss": 0.9484,
+      "step": 25126
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.351605714285715e-06,
+      "loss": 1.2158,
+      "step": 25127
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.351337142857144e-06,
+      "loss": 1.5121,
+      "step": 25128
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.351068571428571e-06,
+      "loss": 0.9955,
+      "step": 25129
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3508e-06,
+      "loss": 1.5193,
+      "step": 25130
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.350531428571429e-06,
+      "loss": 1.1404,
+      "step": 25131
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.350262857142858e-06,
+      "loss": 1.0643,
+      "step": 25132
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.349994285714285e-06,
+      "loss": 1.6723,
+      "step": 25133
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.349725714285714e-06,
+      "loss": 1.0823,
+      "step": 25134
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.349457142857143e-06,
+      "loss": 1.0846,
+      "step": 25135
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.349188571428572e-06,
+      "loss": 1.4616,
+      "step": 25136
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.348920000000001e-06,
+      "loss": 0.8042,
+      "step": 25137
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.34865142857143e-06,
+      "loss": 0.8527,
+      "step": 25138
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.348382857142857e-06,
+      "loss": 0.6043,
+      "step": 25139
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.348114285714286e-06,
+      "loss": 1.2564,
+      "step": 25140
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.347845714285715e-06,
+      "loss": 1.0307,
+      "step": 25141
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.347577142857144e-06,
+      "loss": 1.1028,
+      "step": 25142
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.347308571428571e-06,
+      "loss": 1.0735,
+      "step": 25143
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.34704e-06,
+      "loss": 1.673,
+      "step": 25144
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.346771428571429e-06,
+      "loss": 1.2266,
+      "step": 25145
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.346502857142857e-06,
+      "loss": 1.3961,
+      "step": 25146
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.346234285714286e-06,
+      "loss": 1.2126,
+      "step": 25147
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.345965714285715e-06,
+      "loss": 1.3641,
+      "step": 25148
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.345697142857143e-06,
+      "loss": 0.81,
+      "step": 25149
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.345428571428572e-06,
+      "loss": 1.1085,
+      "step": 25150
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.345160000000001e-06,
+      "loss": 0.7348,
+      "step": 25151
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.344891428571428e-06,
+      "loss": 1.0225,
+      "step": 25152
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.344622857142857e-06,
+      "loss": 1.5373,
+      "step": 25153
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.344354285714286e-06,
+      "loss": 1.1958,
+      "step": 25154
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.344085714285715e-06,
+      "loss": 1.4748,
+      "step": 25155
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.343817142857143e-06,
+      "loss": 1.1116,
+      "step": 25156
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.343548571428572e-06,
+      "loss": 1.5009,
+      "step": 25157
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.343280000000001e-06,
+      "loss": 1.1548,
+      "step": 25158
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.343011428571428e-06,
+      "loss": 1.0214,
+      "step": 25159
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.342742857142857e-06,
+      "loss": 1.4228,
+      "step": 25160
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.342474285714286e-06,
+      "loss": 1.2265,
+      "step": 25161
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3422057142857135e-06,
+      "loss": 1.0919,
+      "step": 25162
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.341937142857143e-06,
+      "loss": 0.7346,
+      "step": 25163
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.341668571428572e-06,
+      "loss": 1.5941,
+      "step": 25164
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.341400000000001e-06,
+      "loss": 0.9099,
+      "step": 25165
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.341131428571429e-06,
+      "loss": 1.1377,
+      "step": 25166
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.340862857142858e-06,
+      "loss": 0.9001,
+      "step": 25167
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.340594285714287e-06,
+      "loss": 0.9691,
+      "step": 25168
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.340325714285714e-06,
+      "loss": 1.3608,
+      "step": 25169
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.340057142857143e-06,
+      "loss": 0.8212,
+      "step": 25170
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.339788571428572e-06,
+      "loss": 1.1262,
+      "step": 25171
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3395199999999995e-06,
+      "loss": 1.2481,
+      "step": 25172
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3392514285714285e-06,
+      "loss": 1.2541,
+      "step": 25173
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3389828571428575e-06,
+      "loss": 1.0806,
+      "step": 25174
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.338714285714285e-06,
+      "loss": 1.0398,
+      "step": 25175
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.338445714285715e-06,
+      "loss": 0.9405,
+      "step": 25176
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.338177142857144e-06,
+      "loss": 1.0634,
+      "step": 25177
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.337908571428573e-06,
+      "loss": 1.0897,
+      "step": 25178
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.33764e-06,
+      "loss": 1.0788,
+      "step": 25179
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.337371428571429e-06,
+      "loss": 1.5551,
+      "step": 25180
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.337102857142858e-06,
+      "loss": 1.7506,
+      "step": 25181
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3368342857142855e-06,
+      "loss": 0.8798,
+      "step": 25182
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3365657142857145e-06,
+      "loss": 1.2609,
+      "step": 25183
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3362971428571435e-06,
+      "loss": 0.8766,
+      "step": 25184
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.336028571428571e-06,
+      "loss": 1.0881,
+      "step": 25185
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.33576e-06,
+      "loss": 1.5392,
+      "step": 25186
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.335491428571429e-06,
+      "loss": 1.148,
+      "step": 25187
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.335222857142856e-06,
+      "loss": 1.3365,
+      "step": 25188
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.334954285714286e-06,
+      "loss": 1.2474,
+      "step": 25189
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.334685714285715e-06,
+      "loss": 1.2084,
+      "step": 25190
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.334417142857144e-06,
+      "loss": 1.0065,
+      "step": 25191
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3341485714285715e-06,
+      "loss": 1.2125,
+      "step": 25192
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3338800000000005e-06,
+      "loss": 1.0231,
+      "step": 25193
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3336114285714295e-06,
+      "loss": 1.3639,
+      "step": 25194
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.333342857142857e-06,
+      "loss": 1.0202,
+      "step": 25195
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.333074285714286e-06,
+      "loss": 1.1833,
+      "step": 25196
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.332805714285715e-06,
+      "loss": 1.074,
+      "step": 25197
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.332537142857142e-06,
+      "loss": 1.0263,
+      "step": 25198
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.332268571428571e-06,
+      "loss": 1.0985,
+      "step": 25199
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.332e-06,
+      "loss": 0.8975,
+      "step": 25200
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.331731428571429e-06,
+      "loss": 1.3246,
+      "step": 25201
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3314628571428575e-06,
+      "loss": 1.0476,
+      "step": 25202
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3311942857142865e-06,
+      "loss": 1.292,
+      "step": 25203
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3309257142857155e-06,
+      "loss": 1.1501,
+      "step": 25204
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.330657142857143e-06,
+      "loss": 1.285,
+      "step": 25205
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.330388571428572e-06,
+      "loss": 1.1515,
+      "step": 25206
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.330120000000001e-06,
+      "loss": 0.7978,
+      "step": 25207
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.329851428571428e-06,
+      "loss": 1.2994,
+      "step": 25208
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.329582857142857e-06,
+      "loss": 1.0461,
+      "step": 25209
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.329314285714286e-06,
+      "loss": 1.1964,
+      "step": 25210
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.329045714285714e-06,
+      "loss": 1.3685,
+      "step": 25211
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.328777142857143e-06,
+      "loss": 1.1495,
+      "step": 25212
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.328508571428572e-06,
+      "loss": 1.4431,
+      "step": 25213
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.328240000000001e-06,
+      "loss": 1.3972,
+      "step": 25214
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.327971428571428e-06,
+      "loss": 1.2228,
+      "step": 25215
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.327702857142858e-06,
+      "loss": 1.2305,
+      "step": 25216
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.327434285714287e-06,
+      "loss": 1.5168,
+      "step": 25217
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.327165714285714e-06,
+      "loss": 1.0366,
+      "step": 25218
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.326897142857143e-06,
+      "loss": 0.8592,
+      "step": 25219
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.326628571428572e-06,
+      "loss": 1.0379,
+      "step": 25220
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.32636e-06,
+      "loss": 1.1482,
+      "step": 25221
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.326091428571429e-06,
+      "loss": 0.7334,
+      "step": 25222
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.325822857142858e-06,
+      "loss": 0.862,
+      "step": 25223
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.325554285714287e-06,
+      "loss": 1.3633,
+      "step": 25224
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.325285714285714e-06,
+      "loss": 0.8566,
+      "step": 25225
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.325017142857143e-06,
+      "loss": 1.1067,
+      "step": 25226
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.324748571428572e-06,
+      "loss": 0.9966,
+      "step": 25227
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.324479999999999e-06,
+      "loss": 1.1096,
+      "step": 25228
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.324211428571429e-06,
+      "loss": 0.7426,
+      "step": 25229
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.323942857142858e-06,
+      "loss": 1.2139,
+      "step": 25230
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.323674285714286e-06,
+      "loss": 1.5573,
+      "step": 25231
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.323405714285715e-06,
+      "loss": 1.1539,
+      "step": 25232
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.323137142857144e-06,
+      "loss": 0.7834,
+      "step": 25233
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.322868571428571e-06,
+      "loss": 1.3558,
+      "step": 25234
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3226e-06,
+      "loss": 0.9231,
+      "step": 25235
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.322331428571429e-06,
+      "loss": 1.1033,
+      "step": 25236
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.322062857142858e-06,
+      "loss": 1.1231,
+      "step": 25237
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.321794285714285e-06,
+      "loss": 0.899,
+      "step": 25238
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.321525714285714e-06,
+      "loss": 1.2702,
+      "step": 25239
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.321257142857143e-06,
+      "loss": 1.0027,
+      "step": 25240
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.320988571428571e-06,
+      "loss": 1.5025,
+      "step": 25241
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.320720000000001e-06,
+      "loss": 1.3363,
+      "step": 25242
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.32045142857143e-06,
+      "loss": 1.2576,
+      "step": 25243
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.320182857142857e-06,
+      "loss": 0.7575,
+      "step": 25244
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.319914285714286e-06,
+      "loss": 1.0427,
+      "step": 25245
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.319645714285715e-06,
+      "loss": 0.8175,
+      "step": 25246
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.319377142857144e-06,
+      "loss": 1.4731,
+      "step": 25247
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.319108571428571e-06,
+      "loss": 1.0124,
+      "step": 25248
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.31884e-06,
+      "loss": 1.3181,
+      "step": 25249
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.318571428571429e-06,
+      "loss": 1.0562,
+      "step": 25250
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.318302857142857e-06,
+      "loss": 1.4162,
+      "step": 25251
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.318034285714286e-06,
+      "loss": 0.9303,
+      "step": 25252
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.317765714285715e-06,
+      "loss": 1.204,
+      "step": 25253
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.317497142857142e-06,
+      "loss": 1.5106,
+      "step": 25254
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.317228571428572e-06,
+      "loss": 1.5014,
+      "step": 25255
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.316960000000001e-06,
+      "loss": 1.2603,
+      "step": 25256
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.316691428571428e-06,
+      "loss": 1.163,
+      "step": 25257
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.316422857142857e-06,
+      "loss": 1.0919,
+      "step": 25258
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.316154285714286e-06,
+      "loss": 1.0562,
+      "step": 25259
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.315885714285715e-06,
+      "loss": 1.0451,
+      "step": 25260
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.315617142857143e-06,
+      "loss": 0.9451,
+      "step": 25261
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.315348571428572e-06,
+      "loss": 0.8173,
+      "step": 25262
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.315080000000001e-06,
+      "loss": 0.9883,
+      "step": 25263
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.314811428571428e-06,
+      "loss": 0.6475,
+      "step": 25264
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.314542857142857e-06,
+      "loss": 1.1411,
+      "step": 25265
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.314274285714286e-06,
+      "loss": 1.1147,
+      "step": 25266
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3140057142857135e-06,
+      "loss": 0.6344,
+      "step": 25267
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3137371428571425e-06,
+      "loss": 1.0762,
+      "step": 25268
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.313468571428572e-06,
+      "loss": 1.1053,
+      "step": 25269
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.313200000000001e-06,
+      "loss": 1.2722,
+      "step": 25270
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.312931428571429e-06,
+      "loss": 0.8916,
+      "step": 25271
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.312662857142858e-06,
+      "loss": 1.3297,
+      "step": 25272
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.312394285714287e-06,
+      "loss": 0.9829,
+      "step": 25273
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.312125714285714e-06,
+      "loss": 1.2611,
+      "step": 25274
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.311857142857143e-06,
+      "loss": 1.5083,
+      "step": 25275
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.311588571428572e-06,
+      "loss": 1.0434,
+      "step": 25276
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3113199999999995e-06,
+      "loss": 0.8,
+      "step": 25277
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3110514285714285e-06,
+      "loss": 0.9401,
+      "step": 25278
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3107828571428575e-06,
+      "loss": 1.2166,
+      "step": 25279
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.310514285714285e-06,
+      "loss": 1.3958,
+      "step": 25280
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.310245714285714e-06,
+      "loss": 1.2714,
+      "step": 25281
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.309977142857144e-06,
+      "loss": 1.0691,
+      "step": 25282
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.309708571428573e-06,
+      "loss": 1.3895,
+      "step": 25283
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.30944e-06,
+      "loss": 0.7823,
+      "step": 25284
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.309171428571429e-06,
+      "loss": 1.3066,
+      "step": 25285
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.308902857142858e-06,
+      "loss": 1.1971,
+      "step": 25286
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 7.3086342857142855e-06,
+      "loss": 1.3358,
+      "step": 25287
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3083657142857145e-06,
+      "loss": 1.2586,
+      "step": 25288
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3080971428571435e-06,
+      "loss": 0.9636,
+      "step": 25289
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.307828571428571e-06,
+      "loss": 1.1293,
+      "step": 25290
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.30756e-06,
+      "loss": 1.6875,
+      "step": 25291
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.307291428571429e-06,
+      "loss": 1.3823,
+      "step": 25292
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.307022857142858e-06,
+      "loss": 1.4656,
+      "step": 25293
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.306754285714285e-06,
+      "loss": 1.3122,
+      "step": 25294
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.306485714285715e-06,
+      "loss": 1.1664,
+      "step": 25295
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.306217142857144e-06,
+      "loss": 1.2651,
+      "step": 25296
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3059485714285715e-06,
+      "loss": 1.3224,
+      "step": 25297
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3056800000000005e-06,
+      "loss": 1.0472,
+      "step": 25298
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3054114285714295e-06,
+      "loss": 0.9504,
+      "step": 25299
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.305142857142857e-06,
+      "loss": 1.2816,
+      "step": 25300
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.304874285714286e-06,
+      "loss": 1.2441,
+      "step": 25301
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.304605714285715e-06,
+      "loss": 1.62,
+      "step": 25302
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.304337142857142e-06,
+      "loss": 1.571,
+      "step": 25303
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.304068571428571e-06,
+      "loss": 1.1518,
+      "step": 25304
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3038e-06,
+      "loss": 1.0964,
+      "step": 25305
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.303531428571429e-06,
+      "loss": 1.1068,
+      "step": 25306
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.303262857142857e-06,
+      "loss": 0.8218,
+      "step": 25307
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3029942857142865e-06,
+      "loss": 1.2594,
+      "step": 25308
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.3027257142857155e-06,
+      "loss": 1.27,
+      "step": 25309
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.302457142857143e-06,
+      "loss": 1.2333,
+      "step": 25310
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.302188571428572e-06,
+      "loss": 1.2907,
+      "step": 25311
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.301920000000001e-06,
+      "loss": 1.4078,
+      "step": 25312
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.301651428571428e-06,
+      "loss": 0.921,
+      "step": 25313
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.301382857142857e-06,
+      "loss": 1.2357,
+      "step": 25314
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.301114285714286e-06,
+      "loss": 1.3186,
+      "step": 25315
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.300845714285715e-06,
+      "loss": 1.1133,
+      "step": 25316
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.300577142857143e-06,
+      "loss": 1.5353,
+      "step": 25317
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.300308571428572e-06,
+      "loss": 1.0465,
+      "step": 25318
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.300040000000001e-06,
+      "loss": 1.2213,
+      "step": 25319
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.299771428571428e-06,
+      "loss": 0.898,
+      "step": 25320
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.299502857142857e-06,
+      "loss": 1.1796,
+      "step": 25321
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.299234285714287e-06,
+      "loss": 1.2989,
+      "step": 25322
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.298965714285714e-06,
+      "loss": 1.2069,
+      "step": 25323
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.298697142857143e-06,
+      "loss": 1.4592,
+      "step": 25324
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.298428571428572e-06,
+      "loss": 1.1978,
+      "step": 25325
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.29816e-06,
+      "loss": 0.8885,
+      "step": 25326
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.297891428571429e-06,
+      "loss": 1.4758,
+      "step": 25327
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.297622857142858e-06,
+      "loss": 1.2324,
+      "step": 25328
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.297354285714287e-06,
+      "loss": 1.2369,
+      "step": 25329
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.297085714285714e-06,
+      "loss": 0.9901,
+      "step": 25330
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.296817142857143e-06,
+      "loss": 0.9483,
+      "step": 25331
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.296548571428572e-06,
+      "loss": 1.3311,
+      "step": 25332
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.296279999999999e-06,
+      "loss": 1.0523,
+      "step": 25333
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.296011428571428e-06,
+      "loss": 1.2419,
+      "step": 25334
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.295742857142858e-06,
+      "loss": 0.95,
+      "step": 25335
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.295474285714286e-06,
+      "loss": 1.0394,
+      "step": 25336
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.295205714285715e-06,
+      "loss": 1.2348,
+      "step": 25337
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.294937142857144e-06,
+      "loss": 1.3165,
+      "step": 25338
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.294668571428571e-06,
+      "loss": 1.3848,
+      "step": 25339
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2944e-06,
+      "loss": 1.3801,
+      "step": 25340
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.294131428571429e-06,
+      "loss": 0.9941,
+      "step": 25341
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.293862857142858e-06,
+      "loss": 0.8599,
+      "step": 25342
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.293594285714285e-06,
+      "loss": 1.2749,
+      "step": 25343
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.293325714285714e-06,
+      "loss": 1.0794,
+      "step": 25344
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.293057142857143e-06,
+      "loss": 1.3419,
+      "step": 25345
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.292788571428571e-06,
+      "loss": 1.2318,
+      "step": 25346
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.29252e-06,
+      "loss": 1.2446,
+      "step": 25347
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.29225142857143e-06,
+      "loss": 1.2775,
+      "step": 25348
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.291982857142857e-06,
+      "loss": 1.0476,
+      "step": 25349
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.291714285714286e-06,
+      "loss": 1.0035,
+      "step": 25350
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.291445714285715e-06,
+      "loss": 1.2585,
+      "step": 25351
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.291177142857144e-06,
+      "loss": 1.1972,
+      "step": 25352
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.290908571428571e-06,
+      "loss": 1.5232,
+      "step": 25353
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.29064e-06,
+      "loss": 1.1341,
+      "step": 25354
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.290371428571429e-06,
+      "loss": 1.086,
+      "step": 25355
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.290102857142857e-06,
+      "loss": 1.0281,
+      "step": 25356
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.289834285714286e-06,
+      "loss": 1.1627,
+      "step": 25357
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.289565714285715e-06,
+      "loss": 0.9061,
+      "step": 25358
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.289297142857142e-06,
+      "loss": 0.7151,
+      "step": 25359
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.289028571428571e-06,
+      "loss": 1.1037,
+      "step": 25360
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.288760000000001e-06,
+      "loss": 1.1613,
+      "step": 25361
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.288491428571428e-06,
+      "loss": 1.2876,
+      "step": 25362
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.288222857142857e-06,
+      "loss": 1.0184,
+      "step": 25363
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.287954285714286e-06,
+      "loss": 1.5524,
+      "step": 25364
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.287685714285715e-06,
+      "loss": 1.5135,
+      "step": 25365
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.287417142857143e-06,
+      "loss": 1.3895,
+      "step": 25366
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.287148571428572e-06,
+      "loss": 1.2887,
+      "step": 25367
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.286880000000001e-06,
+      "loss": 1.3879,
+      "step": 25368
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.286611428571428e-06,
+      "loss": 1.2125,
+      "step": 25369
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.286342857142857e-06,
+      "loss": 0.9771,
+      "step": 25370
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.286074285714286e-06,
+      "loss": 1.2474,
+      "step": 25371
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2858057142857135e-06,
+      "loss": 1.09,
+      "step": 25372
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2855371428571425e-06,
+      "loss": 1.0186,
+      "step": 25373
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2852685714285715e-06,
+      "loss": 1.093,
+      "step": 25374
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.285000000000001e-06,
+      "loss": 1.4043,
+      "step": 25375
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.284731428571429e-06,
+      "loss": 1.3453,
+      "step": 25376
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.284462857142858e-06,
+      "loss": 1.0546,
+      "step": 25377
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.284194285714287e-06,
+      "loss": 1.2861,
+      "step": 25378
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.283925714285714e-06,
+      "loss": 0.7916,
+      "step": 25379
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.283657142857143e-06,
+      "loss": 1.0662,
+      "step": 25380
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.283388571428572e-06,
+      "loss": 1.3297,
+      "step": 25381
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2831199999999995e-06,
+      "loss": 1.0783,
+      "step": 25382
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2828514285714285e-06,
+      "loss": 0.6117,
+      "step": 25383
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2825828571428575e-06,
+      "loss": 1.1853,
+      "step": 25384
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.282314285714285e-06,
+      "loss": 1.1481,
+      "step": 25385
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.282045714285714e-06,
+      "loss": 1.0885,
+      "step": 25386
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.281777142857143e-06,
+      "loss": 1.4433,
+      "step": 25387
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.281508571428573e-06,
+      "loss": 0.9836,
+      "step": 25388
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.28124e-06,
+      "loss": 0.6466,
+      "step": 25389
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.280971428571429e-06,
+      "loss": 1.396,
+      "step": 25390
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.280702857142858e-06,
+      "loss": 1.3305,
+      "step": 25391
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2804342857142855e-06,
+      "loss": 0.9726,
+      "step": 25392
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2801657142857145e-06,
+      "loss": 0.7009,
+      "step": 25393
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2798971428571435e-06,
+      "loss": 1.1464,
+      "step": 25394
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.279628571428571e-06,
+      "loss": 0.694,
+      "step": 25395
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.27936e-06,
+      "loss": 1.454,
+      "step": 25396
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.279091428571429e-06,
+      "loss": 1.137,
+      "step": 25397
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.278822857142858e-06,
+      "loss": 0.9492,
+      "step": 25398
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.278554285714285e-06,
+      "loss": 0.9007,
+      "step": 25399
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.278285714285714e-06,
+      "loss": 1.4613,
+      "step": 25400
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.278017142857144e-06,
+      "loss": 1.0414,
+      "step": 25401
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2777485714285715e-06,
+      "loss": 0.8556,
+      "step": 25402
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2774800000000005e-06,
+      "loss": 1.0734,
+      "step": 25403
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2772114285714295e-06,
+      "loss": 1.3671,
+      "step": 25404
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.276942857142857e-06,
+      "loss": 1.2291,
+      "step": 25405
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.276674285714286e-06,
+      "loss": 0.8893,
+      "step": 25406
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.276405714285715e-06,
+      "loss": 1.5739,
+      "step": 25407
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.276137142857142e-06,
+      "loss": 1.2941,
+      "step": 25408
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.275868571428571e-06,
+      "loss": 0.9865,
+      "step": 25409
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2756e-06,
+      "loss": 1.0661,
+      "step": 25410
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.275331428571429e-06,
+      "loss": 1.1275,
+      "step": 25411
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.275062857142857e-06,
+      "loss": 1.1237,
+      "step": 25412
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.274794285714286e-06,
+      "loss": 1.3049,
+      "step": 25413
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2745257142857155e-06,
+      "loss": 0.9968,
+      "step": 25414
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.274257142857143e-06,
+      "loss": 1.3482,
+      "step": 25415
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.273988571428572e-06,
+      "loss": 1.1254,
+      "step": 25416
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.273720000000001e-06,
+      "loss": 1.3557,
+      "step": 25417
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.273451428571428e-06,
+      "loss": 0.6892,
+      "step": 25418
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.273182857142857e-06,
+      "loss": 1.02,
+      "step": 25419
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.272914285714286e-06,
+      "loss": 1.3399,
+      "step": 25420
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.272645714285715e-06,
+      "loss": 1.0102,
+      "step": 25421
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.272377142857143e-06,
+      "loss": 1.2431,
+      "step": 25422
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.272108571428572e-06,
+      "loss": 0.9219,
+      "step": 25423
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.271840000000001e-06,
+      "loss": 1.2465,
+      "step": 25424
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.271571428571428e-06,
+      "loss": 1.0618,
+      "step": 25425
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.271302857142857e-06,
+      "loss": 0.8879,
+      "step": 25426
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.271034285714286e-06,
+      "loss": 0.8377,
+      "step": 25427
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.270765714285714e-06,
+      "loss": 1.3747,
+      "step": 25428
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.270497142857143e-06,
+      "loss": 0.9023,
+      "step": 25429
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.270228571428572e-06,
+      "loss": 1.2183,
+      "step": 25430
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.26996e-06,
+      "loss": 1.489,
+      "step": 25431
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.269691428571429e-06,
+      "loss": 1.0239,
+      "step": 25432
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.269422857142858e-06,
+      "loss": 1.0117,
+      "step": 25433
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.269154285714287e-06,
+      "loss": 1.1113,
+      "step": 25434
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.268885714285714e-06,
+      "loss": 0.8948,
+      "step": 25435
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.268617142857143e-06,
+      "loss": 0.3706,
+      "step": 25436
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.268348571428572e-06,
+      "loss": 1.1786,
+      "step": 25437
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.268079999999999e-06,
+      "loss": 1.3202,
+      "step": 25438
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.267811428571428e-06,
+      "loss": 1.3509,
+      "step": 25439
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.267542857142857e-06,
+      "loss": 1.064,
+      "step": 25440
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.267274285714286e-06,
+      "loss": 0.984,
+      "step": 25441
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.267005714285715e-06,
+      "loss": 1.3406,
+      "step": 25442
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.266737142857144e-06,
+      "loss": 1.5486,
+      "step": 25443
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.266468571428573e-06,
+      "loss": 1.483,
+      "step": 25444
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.2662e-06,
+      "loss": 1.4003,
+      "step": 25445
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.265931428571429e-06,
+      "loss": 0.6855,
+      "step": 25446
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.265662857142858e-06,
+      "loss": 0.9682,
+      "step": 25447
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.265394285714285e-06,
+      "loss": 1.1349,
+      "step": 25448
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.265125714285714e-06,
+      "loss": 1.2694,
+      "step": 25449
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.264857142857143e-06,
+      "loss": 1.0559,
+      "step": 25450
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.264588571428571e-06,
+      "loss": 1.0872,
+      "step": 25451
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.26432e-06,
+      "loss": 1.7137,
+      "step": 25452
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.264051428571429e-06,
+      "loss": 1.0305,
+      "step": 25453
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.263782857142857e-06,
+      "loss": 0.8038,
+      "step": 25454
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.263514285714286e-06,
+      "loss": 1.3828,
+      "step": 25455
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.263245714285715e-06,
+      "loss": 0.8096,
+      "step": 25456
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.262977142857144e-06,
+      "loss": 1.0661,
+      "step": 25457
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.262708571428571e-06,
+      "loss": 1.5978,
+      "step": 25458
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.26244e-06,
+      "loss": 1.0768,
+      "step": 25459
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.262171428571429e-06,
+      "loss": 0.9185,
+      "step": 25460
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.261902857142857e-06,
+      "loss": 1.0937,
+      "step": 25461
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 7.261634285714286e-06,
+      "loss": 1.0319,
+      "step": 25462
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.261365714285715e-06,
+      "loss": 1.345,
+      "step": 25463
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.261097142857142e-06,
+      "loss": 0.9812,
+      "step": 25464
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.260828571428571e-06,
+      "loss": 1.0783,
+      "step": 25465
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.26056e-06,
+      "loss": 0.9886,
+      "step": 25466
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.26029142857143e-06,
+      "loss": 0.9622,
+      "step": 25467
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.260022857142857e-06,
+      "loss": 1.1239,
+      "step": 25468
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.259754285714286e-06,
+      "loss": 1.3833,
+      "step": 25469
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.259485714285715e-06,
+      "loss": 0.9664,
+      "step": 25470
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.259217142857143e-06,
+      "loss": 1.2865,
+      "step": 25471
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.258948571428572e-06,
+      "loss": 0.9371,
+      "step": 25472
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.258680000000001e-06,
+      "loss": 1.0683,
+      "step": 25473
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.258411428571428e-06,
+      "loss": 1.1256,
+      "step": 25474
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.258142857142857e-06,
+      "loss": 1.5945,
+      "step": 25475
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.257874285714286e-06,
+      "loss": 1.1222,
+      "step": 25476
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2576057142857135e-06,
+      "loss": 1.7722,
+      "step": 25477
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2573371428571425e-06,
+      "loss": 1.2743,
+      "step": 25478
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2570685714285715e-06,
+      "loss": 1.2201,
+      "step": 25479
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2568000000000006e-06,
+      "loss": 1.2592,
+      "step": 25480
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.256531428571429e-06,
+      "loss": 0.9238,
+      "step": 25481
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.256262857142858e-06,
+      "loss": 1.0475,
+      "step": 25482
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.255994285714287e-06,
+      "loss": 1.3203,
+      "step": 25483
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.255725714285714e-06,
+      "loss": 1.1167,
+      "step": 25484
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.255457142857143e-06,
+      "loss": 0.8668,
+      "step": 25485
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.255188571428572e-06,
+      "loss": 1.1345,
+      "step": 25486
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2549199999999995e-06,
+      "loss": 1.4832,
+      "step": 25487
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2546514285714285e-06,
+      "loss": 0.9932,
+      "step": 25488
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2543828571428575e-06,
+      "loss": 1.2812,
+      "step": 25489
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2541142857142865e-06,
+      "loss": 1.2608,
+      "step": 25490
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.253845714285714e-06,
+      "loss": 0.9932,
+      "step": 25491
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.253577142857143e-06,
+      "loss": 1.484,
+      "step": 25492
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.253308571428572e-06,
+      "loss": 1.2,
+      "step": 25493
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.25304e-06,
+      "loss": 0.9366,
+      "step": 25494
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.252771428571429e-06,
+      "loss": 0.838,
+      "step": 25495
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.252502857142858e-06,
+      "loss": 1.1272,
+      "step": 25496
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2522342857142855e-06,
+      "loss": 0.7288,
+      "step": 25497
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2519657142857145e-06,
+      "loss": 0.9772,
+      "step": 25498
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2516971428571435e-06,
+      "loss": 1.1847,
+      "step": 25499
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.251428571428571e-06,
+      "loss": 0.8294,
+      "step": 25500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.25116e-06,
+      "loss": 0.7085,
+      "step": 25501
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.250891428571429e-06,
+      "loss": 1.0899,
+      "step": 25502
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.250622857142858e-06,
+      "loss": 1.1267,
+      "step": 25503
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.250354285714285e-06,
+      "loss": 1.4595,
+      "step": 25504
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.250085714285714e-06,
+      "loss": 1.2123,
+      "step": 25505
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.249817142857143e-06,
+      "loss": 0.7435,
+      "step": 25506
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2495485714285715e-06,
+      "loss": 1.152,
+      "step": 25507
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2492800000000005e-06,
+      "loss": 1.4036,
+      "step": 25508
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2490114285714295e-06,
+      "loss": 1.3065,
+      "step": 25509
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.248742857142857e-06,
+      "loss": 1.0354,
+      "step": 25510
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.248474285714286e-06,
+      "loss": 1.3873,
+      "step": 25511
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.248205714285715e-06,
+      "loss": 1.0113,
+      "step": 25512
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.247937142857142e-06,
+      "loss": 0.9279,
+      "step": 25513
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.247668571428571e-06,
+      "loss": 1.2676,
+      "step": 25514
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2474e-06,
+      "loss": 1.173,
+      "step": 25515
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.247131428571429e-06,
+      "loss": 0.7721,
+      "step": 25516
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.246862857142857e-06,
+      "loss": 1.8624,
+      "step": 25517
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.246594285714286e-06,
+      "loss": 1.6681,
+      "step": 25518
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.246325714285715e-06,
+      "loss": 1.7834,
+      "step": 25519
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.246057142857143e-06,
+      "loss": 1.7689,
+      "step": 25520
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.245788571428572e-06,
+      "loss": 1.0386,
+      "step": 25521
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.245520000000001e-06,
+      "loss": 1.0268,
+      "step": 25522
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.245251428571428e-06,
+      "loss": 1.2117,
+      "step": 25523
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.244982857142857e-06,
+      "loss": 0.8198,
+      "step": 25524
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.244714285714286e-06,
+      "loss": 1.4965,
+      "step": 25525
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.244445714285715e-06,
+      "loss": 0.9111,
+      "step": 25526
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.244177142857143e-06,
+      "loss": 1.4743,
+      "step": 25527
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.243908571428572e-06,
+      "loss": 0.8597,
+      "step": 25528
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.243640000000001e-06,
+      "loss": 1.2018,
+      "step": 25529
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.243371428571428e-06,
+      "loss": 1.4065,
+      "step": 25530
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.243102857142857e-06,
+      "loss": 1.3819,
+      "step": 25531
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.242834285714286e-06,
+      "loss": 1.1983,
+      "step": 25532
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.242565714285714e-06,
+      "loss": 1.3362,
+      "step": 25533
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.242297142857143e-06,
+      "loss": 0.7774,
+      "step": 25534
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.242028571428572e-06,
+      "loss": 1.0622,
+      "step": 25535
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.24176e-06,
+      "loss": 0.9901,
+      "step": 25536
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.241491428571429e-06,
+      "loss": 1.4069,
+      "step": 25537
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.241222857142858e-06,
+      "loss": 0.8573,
+      "step": 25538
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.240954285714287e-06,
+      "loss": 1.3835,
+      "step": 25539
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.240685714285714e-06,
+      "loss": 1.2406,
+      "step": 25540
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.240417142857143e-06,
+      "loss": 1.101,
+      "step": 25541
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.240148571428572e-06,
+      "loss": 1.1127,
+      "step": 25542
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.239879999999999e-06,
+      "loss": 0.9981,
+      "step": 25543
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.239611428571428e-06,
+      "loss": 1.1654,
+      "step": 25544
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.239342857142857e-06,
+      "loss": 1.3361,
+      "step": 25545
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.239074285714286e-06,
+      "loss": 1.4765,
+      "step": 25546
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.238805714285715e-06,
+      "loss": 1.0408,
+      "step": 25547
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.238537142857144e-06,
+      "loss": 1.0643,
+      "step": 25548
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.238268571428573e-06,
+      "loss": 1.4907,
+      "step": 25549
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.238e-06,
+      "loss": 1.6107,
+      "step": 25550
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.237731428571429e-06,
+      "loss": 1.4661,
+      "step": 25551
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.237462857142858e-06,
+      "loss": 1.1475,
+      "step": 25552
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.237194285714285e-06,
+      "loss": 1.3915,
+      "step": 25553
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.236925714285714e-06,
+      "loss": 1.2,
+      "step": 25554
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.236657142857143e-06,
+      "loss": 1.2548,
+      "step": 25555
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.236388571428571e-06,
+      "loss": 1.3766,
+      "step": 25556
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.23612e-06,
+      "loss": 0.9299,
+      "step": 25557
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.235851428571429e-06,
+      "loss": 1.2574,
+      "step": 25558
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.235582857142857e-06,
+      "loss": 1.09,
+      "step": 25559
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.235314285714286e-06,
+      "loss": 1.1451,
+      "step": 25560
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.235045714285715e-06,
+      "loss": 0.6948,
+      "step": 25561
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.234777142857144e-06,
+      "loss": 0.9396,
+      "step": 25562
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.234508571428571e-06,
+      "loss": 1.1357,
+      "step": 25563
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.23424e-06,
+      "loss": 1.2486,
+      "step": 25564
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.233971428571429e-06,
+      "loss": 1.298,
+      "step": 25565
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.233702857142857e-06,
+      "loss": 1.194,
+      "step": 25566
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.233434285714286e-06,
+      "loss": 1.3442,
+      "step": 25567
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.233165714285715e-06,
+      "loss": 1.3868,
+      "step": 25568
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.232897142857142e-06,
+      "loss": 1.4365,
+      "step": 25569
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.232628571428571e-06,
+      "loss": 1.2784,
+      "step": 25570
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.23236e-06,
+      "loss": 1.4723,
+      "step": 25571
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.232091428571429e-06,
+      "loss": 1.0243,
+      "step": 25572
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.231822857142857e-06,
+      "loss": 0.7605,
+      "step": 25573
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.231554285714286e-06,
+      "loss": 1.1873,
+      "step": 25574
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.231285714285715e-06,
+      "loss": 1.4365,
+      "step": 25575
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.231017142857143e-06,
+      "loss": 1.5202,
+      "step": 25576
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.230748571428572e-06,
+      "loss": 1.177,
+      "step": 25577
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.230480000000001e-06,
+      "loss": 1.3076,
+      "step": 25578
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.230211428571428e-06,
+      "loss": 1.2907,
+      "step": 25579
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.229942857142857e-06,
+      "loss": 0.7076,
+      "step": 25580
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.229674285714286e-06,
+      "loss": 1.2848,
+      "step": 25581
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2294057142857135e-06,
+      "loss": 0.9281,
+      "step": 25582
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2291371428571425e-06,
+      "loss": 0.7733,
+      "step": 25583
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2288685714285715e-06,
+      "loss": 1.2196,
+      "step": 25584
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2286000000000006e-06,
+      "loss": 0.7833,
+      "step": 25585
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.228331428571429e-06,
+      "loss": 1.1325,
+      "step": 25586
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.228062857142858e-06,
+      "loss": 0.893,
+      "step": 25587
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.227794285714287e-06,
+      "loss": 1.0467,
+      "step": 25588
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.227525714285714e-06,
+      "loss": 1.3038,
+      "step": 25589
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.227257142857143e-06,
+      "loss": 1.0999,
+      "step": 25590
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.226988571428572e-06,
+      "loss": 1.2683,
+      "step": 25591
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2267199999999995e-06,
+      "loss": 1.3367,
+      "step": 25592
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2264514285714285e-06,
+      "loss": 0.9455,
+      "step": 25593
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2261828571428575e-06,
+      "loss": 1.4264,
+      "step": 25594
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2259142857142865e-06,
+      "loss": 1.4367,
+      "step": 25595
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.225645714285714e-06,
+      "loss": 0.7574,
+      "step": 25596
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.225377142857143e-06,
+      "loss": 1.0786,
+      "step": 25597
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.225108571428572e-06,
+      "loss": 1.044,
+      "step": 25598
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.22484e-06,
+      "loss": 1.2875,
+      "step": 25599
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.224571428571429e-06,
+      "loss": 1.1539,
+      "step": 25600
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.224302857142858e-06,
+      "loss": 1.1579,
+      "step": 25601
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2240342857142855e-06,
+      "loss": 1.2334,
+      "step": 25602
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2237657142857145e-06,
+      "loss": 1.1033,
+      "step": 25603
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2234971428571435e-06,
+      "loss": 1.2143,
+      "step": 25604
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.223228571428571e-06,
+      "loss": 1.4399,
+      "step": 25605
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.22296e-06,
+      "loss": 1.2553,
+      "step": 25606
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.222691428571429e-06,
+      "loss": 1.0274,
+      "step": 25607
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.222422857142858e-06,
+      "loss": 1.1902,
+      "step": 25608
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.222154285714285e-06,
+      "loss": 1.6136,
+      "step": 25609
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.221885714285714e-06,
+      "loss": 1.4381,
+      "step": 25610
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.221617142857143e-06,
+      "loss": 0.8532,
+      "step": 25611
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2213485714285715e-06,
+      "loss": 1.3972,
+      "step": 25612
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2210800000000005e-06,
+      "loss": 0.6768,
+      "step": 25613
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2208114285714295e-06,
+      "loss": 0.8874,
+      "step": 25614
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.220542857142857e-06,
+      "loss": 1.2446,
+      "step": 25615
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.220274285714286e-06,
+      "loss": 1.4387,
+      "step": 25616
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.220005714285715e-06,
+      "loss": 0.9768,
+      "step": 25617
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.219737142857144e-06,
+      "loss": 0.7837,
+      "step": 25618
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.219468571428571e-06,
+      "loss": 1.1854,
+      "step": 25619
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.2192e-06,
+      "loss": 0.8856,
+      "step": 25620
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.218931428571429e-06,
+      "loss": 1.1814,
+      "step": 25621
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.218662857142857e-06,
+      "loss": 0.7544,
+      "step": 25622
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.218394285714286e-06,
+      "loss": 1.6166,
+      "step": 25623
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.218125714285715e-06,
+      "loss": 1.4947,
+      "step": 25624
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.217857142857143e-06,
+      "loss": 0.9412,
+      "step": 25625
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.217588571428572e-06,
+      "loss": 0.9723,
+      "step": 25626
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.217320000000001e-06,
+      "loss": 1.2001,
+      "step": 25627
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.217051428571428e-06,
+      "loss": 0.87,
+      "step": 25628
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.216782857142857e-06,
+      "loss": 1.3938,
+      "step": 25629
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.216514285714286e-06,
+      "loss": 1.4674,
+      "step": 25630
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.216245714285715e-06,
+      "loss": 1.6719,
+      "step": 25631
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.215977142857143e-06,
+      "loss": 1.2374,
+      "step": 25632
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.215708571428572e-06,
+      "loss": 1.0934,
+      "step": 25633
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.215440000000001e-06,
+      "loss": 0.9204,
+      "step": 25634
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.215171428571428e-06,
+      "loss": 1.3921,
+      "step": 25635
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.214902857142857e-06,
+      "loss": 0.6824,
+      "step": 25636
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 7.214634285714286e-06,
+      "loss": 0.887,
+      "step": 25637
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.214365714285714e-06,
+      "loss": 1.1882,
+      "step": 25638
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.214097142857143e-06,
+      "loss": 1.039,
+      "step": 25639
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.213828571428572e-06,
+      "loss": 1.239,
+      "step": 25640
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.213560000000001e-06,
+      "loss": 1.1662,
+      "step": 25641
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.213291428571429e-06,
+      "loss": 0.8976,
+      "step": 25642
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.213022857142858e-06,
+      "loss": 0.9814,
+      "step": 25643
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.212754285714287e-06,
+      "loss": 1.1061,
+      "step": 25644
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.212485714285714e-06,
+      "loss": 1.1024,
+      "step": 25645
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.212217142857143e-06,
+      "loss": 1.1851,
+      "step": 25646
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.211948571428572e-06,
+      "loss": 0.763,
+      "step": 25647
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.211679999999999e-06,
+      "loss": 1.4795,
+      "step": 25648
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.211411428571428e-06,
+      "loss": 0.9494,
+      "step": 25649
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.211142857142857e-06,
+      "loss": 1.2416,
+      "step": 25650
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.210874285714286e-06,
+      "loss": 0.9538,
+      "step": 25651
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.210605714285715e-06,
+      "loss": 1.1004,
+      "step": 25652
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.210337142857144e-06,
+      "loss": 1.7243,
+      "step": 25653
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.210068571428573e-06,
+      "loss": 1.3842,
+      "step": 25654
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.2098e-06,
+      "loss": 1.1917,
+      "step": 25655
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.209531428571429e-06,
+      "loss": 1.0837,
+      "step": 25656
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.209262857142858e-06,
+      "loss": 1.3902,
+      "step": 25657
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.208994285714285e-06,
+      "loss": 0.9315,
+      "step": 25658
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.208725714285714e-06,
+      "loss": 0.8486,
+      "step": 25659
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.208457142857143e-06,
+      "loss": 1.5861,
+      "step": 25660
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.208188571428571e-06,
+      "loss": 0.8861,
+      "step": 25661
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.20792e-06,
+      "loss": 1.4447,
+      "step": 25662
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.207651428571429e-06,
+      "loss": 1.0937,
+      "step": 25663
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.207382857142858e-06,
+      "loss": 1.0216,
+      "step": 25664
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.207114285714286e-06,
+      "loss": 0.9225,
+      "step": 25665
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.206845714285715e-06,
+      "loss": 1.2107,
+      "step": 25666
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.206577142857144e-06,
+      "loss": 1.3056,
+      "step": 25667
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.206308571428571e-06,
+      "loss": 1.3063,
+      "step": 25668
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.20604e-06,
+      "loss": 0.895,
+      "step": 25669
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.205771428571429e-06,
+      "loss": 1.1748,
+      "step": 25670
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.205502857142857e-06,
+      "loss": 1.1034,
+      "step": 25671
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.205234285714286e-06,
+      "loss": 1.3358,
+      "step": 25672
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.204965714285715e-06,
+      "loss": 1.3862,
+      "step": 25673
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.204697142857142e-06,
+      "loss": 1.5182,
+      "step": 25674
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.204428571428571e-06,
+      "loss": 1.2129,
+      "step": 25675
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.20416e-06,
+      "loss": 1.2146,
+      "step": 25676
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.203891428571429e-06,
+      "loss": 1.4799,
+      "step": 25677
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.203622857142857e-06,
+      "loss": 1.5209,
+      "step": 25678
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.203354285714286e-06,
+      "loss": 0.8612,
+      "step": 25679
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.203085714285715e-06,
+      "loss": 1.5007,
+      "step": 25680
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.202817142857143e-06,
+      "loss": 0.8882,
+      "step": 25681
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.202548571428572e-06,
+      "loss": 0.8719,
+      "step": 25682
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.202280000000001e-06,
+      "loss": 1.0779,
+      "step": 25683
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.202011428571428e-06,
+      "loss": 1.4718,
+      "step": 25684
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.201742857142857e-06,
+      "loss": 1.0796,
+      "step": 25685
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.201474285714286e-06,
+      "loss": 0.983,
+      "step": 25686
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.2012057142857135e-06,
+      "loss": 1.4299,
+      "step": 25687
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.2009371428571425e-06,
+      "loss": 1.0548,
+      "step": 25688
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.2006685714285715e-06,
+      "loss": 1.0371,
+      "step": 25689
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.2004000000000006e-06,
+      "loss": 0.8988,
+      "step": 25690
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.200131428571429e-06,
+      "loss": 1.4598,
+      "step": 25691
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.199862857142858e-06,
+      "loss": 1.3161,
+      "step": 25692
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.199594285714287e-06,
+      "loss": 1.1181,
+      "step": 25693
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.199325714285714e-06,
+      "loss": 1.665,
+      "step": 25694
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.199057142857143e-06,
+      "loss": 0.9839,
+      "step": 25695
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.198788571428572e-06,
+      "loss": 0.8946,
+      "step": 25696
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1985199999999995e-06,
+      "loss": 0.8458,
+      "step": 25697
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1982514285714285e-06,
+      "loss": 0.8435,
+      "step": 25698
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1979828571428575e-06,
+      "loss": 1.1712,
+      "step": 25699
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1977142857142865e-06,
+      "loss": 1.0083,
+      "step": 25700
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.197445714285714e-06,
+      "loss": 1.6096,
+      "step": 25701
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.197177142857143e-06,
+      "loss": 1.2447,
+      "step": 25702
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.196908571428572e-06,
+      "loss": 1.2625,
+      "step": 25703
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.19664e-06,
+      "loss": 1.4262,
+      "step": 25704
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.196371428571429e-06,
+      "loss": 0.7664,
+      "step": 25705
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.196102857142858e-06,
+      "loss": 0.9623,
+      "step": 25706
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1958342857142855e-06,
+      "loss": 1.4577,
+      "step": 25707
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1955657142857145e-06,
+      "loss": 1.1234,
+      "step": 25708
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1952971428571435e-06,
+      "loss": 1.1777,
+      "step": 25709
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.195028571428571e-06,
+      "loss": 1.0837,
+      "step": 25710
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.19476e-06,
+      "loss": 0.8451,
+      "step": 25711
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.194491428571429e-06,
+      "loss": 1.0738,
+      "step": 25712
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.194222857142858e-06,
+      "loss": 1.3013,
+      "step": 25713
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.193954285714285e-06,
+      "loss": 1.4586,
+      "step": 25714
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.193685714285714e-06,
+      "loss": 1.2527,
+      "step": 25715
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.193417142857143e-06,
+      "loss": 1.2638,
+      "step": 25716
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1931485714285715e-06,
+      "loss": 1.5582,
+      "step": 25717
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1928800000000005e-06,
+      "loss": 0.7452,
+      "step": 25718
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1926114285714295e-06,
+      "loss": 1.586,
+      "step": 25719
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.192342857142857e-06,
+      "loss": 1.1679,
+      "step": 25720
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.192074285714286e-06,
+      "loss": 1.5568,
+      "step": 25721
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.191805714285715e-06,
+      "loss": 1.0555,
+      "step": 25722
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.191537142857144e-06,
+      "loss": 0.6291,
+      "step": 25723
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.191268571428571e-06,
+      "loss": 1.0444,
+      "step": 25724
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.191e-06,
+      "loss": 0.6553,
+      "step": 25725
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.190731428571429e-06,
+      "loss": 0.9455,
+      "step": 25726
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.190462857142857e-06,
+      "loss": 0.732,
+      "step": 25727
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.190194285714286e-06,
+      "loss": 0.9542,
+      "step": 25728
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.189925714285715e-06,
+      "loss": 1.0555,
+      "step": 25729
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.189657142857143e-06,
+      "loss": 0.8639,
+      "step": 25730
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.189388571428572e-06,
+      "loss": 0.8788,
+      "step": 25731
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.189120000000001e-06,
+      "loss": 1.3325,
+      "step": 25732
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.188851428571428e-06,
+      "loss": 0.9391,
+      "step": 25733
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.188582857142857e-06,
+      "loss": 1.1911,
+      "step": 25734
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.188314285714286e-06,
+      "loss": 1.13,
+      "step": 25735
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.188045714285715e-06,
+      "loss": 1.1969,
+      "step": 25736
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.187777142857143e-06,
+      "loss": 0.7658,
+      "step": 25737
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.187508571428572e-06,
+      "loss": 1.279,
+      "step": 25738
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.187240000000001e-06,
+      "loss": 1.5141,
+      "step": 25739
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.186971428571428e-06,
+      "loss": 0.9351,
+      "step": 25740
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.186702857142857e-06,
+      "loss": 0.8498,
+      "step": 25741
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.186434285714286e-06,
+      "loss": 1.2735,
+      "step": 25742
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.186165714285714e-06,
+      "loss": 1.4113,
+      "step": 25743
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.185897142857143e-06,
+      "loss": 0.9421,
+      "step": 25744
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.185628571428572e-06,
+      "loss": 1.2771,
+      "step": 25745
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.185360000000001e-06,
+      "loss": 1.4688,
+      "step": 25746
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.185091428571429e-06,
+      "loss": 1.2412,
+      "step": 25747
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.184822857142858e-06,
+      "loss": 0.9644,
+      "step": 25748
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.184554285714287e-06,
+      "loss": 1.0651,
+      "step": 25749
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.184285714285714e-06,
+      "loss": 1.1201,
+      "step": 25750
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.184017142857143e-06,
+      "loss": 1.0624,
+      "step": 25751
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.183748571428572e-06,
+      "loss": 1.2159,
+      "step": 25752
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.183479999999999e-06,
+      "loss": 1.5568,
+      "step": 25753
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.183211428571428e-06,
+      "loss": 1.267,
+      "step": 25754
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.182942857142857e-06,
+      "loss": 0.9867,
+      "step": 25755
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.182674285714286e-06,
+      "loss": 1.4812,
+      "step": 25756
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.182405714285715e-06,
+      "loss": 1.0747,
+      "step": 25757
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.182137142857144e-06,
+      "loss": 1.1841,
+      "step": 25758
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.181868571428573e-06,
+      "loss": 0.8246,
+      "step": 25759
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1816e-06,
+      "loss": 1.1173,
+      "step": 25760
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.181331428571429e-06,
+      "loss": 1.5539,
+      "step": 25761
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.181062857142858e-06,
+      "loss": 1.5461,
+      "step": 25762
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.180794285714285e-06,
+      "loss": 1.1449,
+      "step": 25763
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.180525714285714e-06,
+      "loss": 0.9868,
+      "step": 25764
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.180257142857143e-06,
+      "loss": 0.9436,
+      "step": 25765
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.179988571428571e-06,
+      "loss": 1.1034,
+      "step": 25766
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.17972e-06,
+      "loss": 1.4527,
+      "step": 25767
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.179451428571429e-06,
+      "loss": 0.8912,
+      "step": 25768
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.179182857142858e-06,
+      "loss": 1.0585,
+      "step": 25769
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.178914285714286e-06,
+      "loss": 1.027,
+      "step": 25770
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.178645714285715e-06,
+      "loss": 1.6144,
+      "step": 25771
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.178377142857144e-06,
+      "loss": 0.9245,
+      "step": 25772
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.178108571428571e-06,
+      "loss": 1.3492,
+      "step": 25773
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.17784e-06,
+      "loss": 0.5916,
+      "step": 25774
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.177571428571429e-06,
+      "loss": 1.4585,
+      "step": 25775
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.177302857142857e-06,
+      "loss": 1.0635,
+      "step": 25776
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.177034285714286e-06,
+      "loss": 1.3471,
+      "step": 25777
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.176765714285715e-06,
+      "loss": 1.4081,
+      "step": 25778
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.176497142857142e-06,
+      "loss": 0.98,
+      "step": 25779
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.176228571428571e-06,
+      "loss": 0.9095,
+      "step": 25780
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.17596e-06,
+      "loss": 1.1976,
+      "step": 25781
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.175691428571429e-06,
+      "loss": 1.4555,
+      "step": 25782
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.175422857142857e-06,
+      "loss": 1.1202,
+      "step": 25783
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.175154285714286e-06,
+      "loss": 1.0786,
+      "step": 25784
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.174885714285715e-06,
+      "loss": 1.0298,
+      "step": 25785
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.174617142857143e-06,
+      "loss": 1.2159,
+      "step": 25786
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.174348571428572e-06,
+      "loss": 0.9554,
+      "step": 25787
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.174080000000001e-06,
+      "loss": 1.1155,
+      "step": 25788
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.173811428571428e-06,
+      "loss": 1.0399,
+      "step": 25789
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.173542857142857e-06,
+      "loss": 1.6084,
+      "step": 25790
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.173274285714286e-06,
+      "loss": 1.0963,
+      "step": 25791
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.173005714285715e-06,
+      "loss": 1.3414,
+      "step": 25792
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1727371428571425e-06,
+      "loss": 0.7689,
+      "step": 25793
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1724685714285715e-06,
+      "loss": 1.0797,
+      "step": 25794
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1722000000000006e-06,
+      "loss": 1.0605,
+      "step": 25795
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.171931428571429e-06,
+      "loss": 1.536,
+      "step": 25796
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.171662857142858e-06,
+      "loss": 0.8783,
+      "step": 25797
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.171394285714287e-06,
+      "loss": 1.2521,
+      "step": 25798
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.171125714285714e-06,
+      "loss": 1.1471,
+      "step": 25799
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.170857142857143e-06,
+      "loss": 1.1835,
+      "step": 25800
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.170588571428572e-06,
+      "loss": 1.3162,
+      "step": 25801
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1703199999999995e-06,
+      "loss": 0.7578,
+      "step": 25802
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1700514285714285e-06,
+      "loss": 1.2961,
+      "step": 25803
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1697828571428575e-06,
+      "loss": 1.6855,
+      "step": 25804
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1695142857142865e-06,
+      "loss": 1.0481,
+      "step": 25805
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.169245714285714e-06,
+      "loss": 1.7234,
+      "step": 25806
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.168977142857143e-06,
+      "loss": 1.3628,
+      "step": 25807
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.168708571428572e-06,
+      "loss": 1.3492,
+      "step": 25808
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.16844e-06,
+      "loss": 0.8622,
+      "step": 25809
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.168171428571429e-06,
+      "loss": 1.1271,
+      "step": 25810
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.167902857142858e-06,
+      "loss": 1.4342,
+      "step": 25811
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.1676342857142855e-06,
+      "loss": 1.2448,
+      "step": 25812
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1673657142857145e-06,
+      "loss": 1.0728,
+      "step": 25813
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1670971428571435e-06,
+      "loss": 1.4484,
+      "step": 25814
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1668285714285725e-06,
+      "loss": 0.9674,
+      "step": 25815
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.16656e-06,
+      "loss": 0.9768,
+      "step": 25816
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.166291428571429e-06,
+      "loss": 0.9247,
+      "step": 25817
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.166022857142858e-06,
+      "loss": 1.4894,
+      "step": 25818
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.165754285714285e-06,
+      "loss": 1.2787,
+      "step": 25819
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.165485714285714e-06,
+      "loss": 1.346,
+      "step": 25820
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.165217142857143e-06,
+      "loss": 1.2028,
+      "step": 25821
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1649485714285715e-06,
+      "loss": 1.3928,
+      "step": 25822
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1646800000000005e-06,
+      "loss": 1.3321,
+      "step": 25823
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1644114285714295e-06,
+      "loss": 1.2973,
+      "step": 25824
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.164142857142857e-06,
+      "loss": 1.2829,
+      "step": 25825
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.163874285714286e-06,
+      "loss": 1.3137,
+      "step": 25826
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.163605714285715e-06,
+      "loss": 1.0638,
+      "step": 25827
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.163337142857144e-06,
+      "loss": 1.2407,
+      "step": 25828
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.163068571428571e-06,
+      "loss": 1.4435,
+      "step": 25829
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1628e-06,
+      "loss": 1.1285,
+      "step": 25830
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.162531428571429e-06,
+      "loss": 1.238,
+      "step": 25831
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.162262857142857e-06,
+      "loss": 1.3078,
+      "step": 25832
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.161994285714286e-06,
+      "loss": 1.0736,
+      "step": 25833
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.161725714285715e-06,
+      "loss": 1.1277,
+      "step": 25834
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.161457142857143e-06,
+      "loss": 0.931,
+      "step": 25835
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.161188571428572e-06,
+      "loss": 0.8768,
+      "step": 25836
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.160920000000001e-06,
+      "loss": 1.091,
+      "step": 25837
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.160651428571428e-06,
+      "loss": 1.4172,
+      "step": 25838
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.160382857142857e-06,
+      "loss": 1.1422,
+      "step": 25839
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.160114285714286e-06,
+      "loss": 1.2167,
+      "step": 25840
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.159845714285715e-06,
+      "loss": 1.0454,
+      "step": 25841
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.159577142857143e-06,
+      "loss": 1.2752,
+      "step": 25842
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.159308571428572e-06,
+      "loss": 0.8829,
+      "step": 25843
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.159040000000001e-06,
+      "loss": 0.9565,
+      "step": 25844
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.158771428571428e-06,
+      "loss": 1.1605,
+      "step": 25845
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.158502857142857e-06,
+      "loss": 1.6179,
+      "step": 25846
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.158234285714286e-06,
+      "loss": 0.8132,
+      "step": 25847
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.157965714285714e-06,
+      "loss": 0.9814,
+      "step": 25848
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.157697142857143e-06,
+      "loss": 1.2583,
+      "step": 25849
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.157428571428572e-06,
+      "loss": 0.6331,
+      "step": 25850
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.157160000000001e-06,
+      "loss": 1.2839,
+      "step": 25851
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.156891428571429e-06,
+      "loss": 1.4251,
+      "step": 25852
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.156622857142858e-06,
+      "loss": 0.6761,
+      "step": 25853
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.156354285714287e-06,
+      "loss": 1.1771,
+      "step": 25854
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.156085714285714e-06,
+      "loss": 1.0484,
+      "step": 25855
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.155817142857143e-06,
+      "loss": 1.3721,
+      "step": 25856
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.155548571428572e-06,
+      "loss": 1.184,
+      "step": 25857
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.155279999999999e-06,
+      "loss": 1.5853,
+      "step": 25858
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.155011428571428e-06,
+      "loss": 1.6451,
+      "step": 25859
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.154742857142857e-06,
+      "loss": 1.0358,
+      "step": 25860
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.154474285714286e-06,
+      "loss": 0.9311,
+      "step": 25861
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.154205714285715e-06,
+      "loss": 1.401,
+      "step": 25862
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.153937142857144e-06,
+      "loss": 1.2466,
+      "step": 25863
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.153668571428573e-06,
+      "loss": 0.6022,
+      "step": 25864
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1534e-06,
+      "loss": 0.7334,
+      "step": 25865
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.153131428571429e-06,
+      "loss": 1.4352,
+      "step": 25866
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.152862857142858e-06,
+      "loss": 1.3251,
+      "step": 25867
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.152594285714285e-06,
+      "loss": 1.0925,
+      "step": 25868
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.152325714285714e-06,
+      "loss": 1.2701,
+      "step": 25869
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.152057142857143e-06,
+      "loss": 1.2999,
+      "step": 25870
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.151788571428571e-06,
+      "loss": 1.15,
+      "step": 25871
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.15152e-06,
+      "loss": 1.252,
+      "step": 25872
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.151251428571429e-06,
+      "loss": 0.8307,
+      "step": 25873
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.150982857142858e-06,
+      "loss": 1.111,
+      "step": 25874
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.150714285714286e-06,
+      "loss": 0.83,
+      "step": 25875
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.150445714285715e-06,
+      "loss": 0.8607,
+      "step": 25876
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.150177142857144e-06,
+      "loss": 0.9175,
+      "step": 25877
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.149908571428571e-06,
+      "loss": 0.8845,
+      "step": 25878
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.14964e-06,
+      "loss": 1.2007,
+      "step": 25879
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.149371428571429e-06,
+      "loss": 1.0804,
+      "step": 25880
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.149102857142857e-06,
+      "loss": 1.1173,
+      "step": 25881
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.148834285714286e-06,
+      "loss": 1.4328,
+      "step": 25882
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.148565714285715e-06,
+      "loss": 1.2701,
+      "step": 25883
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.148297142857142e-06,
+      "loss": 1.5534,
+      "step": 25884
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.148028571428571e-06,
+      "loss": 1.458,
+      "step": 25885
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.14776e-06,
+      "loss": 1.2263,
+      "step": 25886
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.147491428571429e-06,
+      "loss": 0.8486,
+      "step": 25887
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.147222857142857e-06,
+      "loss": 0.9741,
+      "step": 25888
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.146954285714286e-06,
+      "loss": 1.4056,
+      "step": 25889
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.146685714285715e-06,
+      "loss": 1.5069,
+      "step": 25890
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.146417142857143e-06,
+      "loss": 1.4952,
+      "step": 25891
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.146148571428572e-06,
+      "loss": 1.2116,
+      "step": 25892
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.145880000000001e-06,
+      "loss": 1.1019,
+      "step": 25893
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.145611428571428e-06,
+      "loss": 1.0519,
+      "step": 25894
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.145342857142857e-06,
+      "loss": 0.992,
+      "step": 25895
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.145074285714286e-06,
+      "loss": 1.2361,
+      "step": 25896
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.144805714285715e-06,
+      "loss": 1.2375,
+      "step": 25897
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1445371428571425e-06,
+      "loss": 1.1947,
+      "step": 25898
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1442685714285715e-06,
+      "loss": 1.1137,
+      "step": 25899
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1440000000000005e-06,
+      "loss": 1.2339,
+      "step": 25900
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.143731428571429e-06,
+      "loss": 1.2955,
+      "step": 25901
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.143462857142858e-06,
+      "loss": 1.1574,
+      "step": 25902
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.143194285714287e-06,
+      "loss": 1.1467,
+      "step": 25903
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.142925714285714e-06,
+      "loss": 1.0405,
+      "step": 25904
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.142657142857143e-06,
+      "loss": 1.0009,
+      "step": 25905
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.142388571428572e-06,
+      "loss": 1.4212,
+      "step": 25906
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1421199999999995e-06,
+      "loss": 1.0379,
+      "step": 25907
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1418514285714285e-06,
+      "loss": 1.7686,
+      "step": 25908
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1415828571428575e-06,
+      "loss": 1.085,
+      "step": 25909
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1413142857142865e-06,
+      "loss": 1.0469,
+      "step": 25910
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.141045714285714e-06,
+      "loss": 0.8442,
+      "step": 25911
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.140777142857143e-06,
+      "loss": 0.8702,
+      "step": 25912
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.140508571428572e-06,
+      "loss": 0.9507,
+      "step": 25913
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.14024e-06,
+      "loss": 1.1295,
+      "step": 25914
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.139971428571429e-06,
+      "loss": 0.9174,
+      "step": 25915
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.139702857142858e-06,
+      "loss": 0.8454,
+      "step": 25916
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1394342857142855e-06,
+      "loss": 1.28,
+      "step": 25917
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1391657142857145e-06,
+      "loss": 1.6195,
+      "step": 25918
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1388971428571435e-06,
+      "loss": 1.6452,
+      "step": 25919
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1386285714285725e-06,
+      "loss": 0.7986,
+      "step": 25920
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.13836e-06,
+      "loss": 1.2587,
+      "step": 25921
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.138091428571429e-06,
+      "loss": 1.2454,
+      "step": 25922
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.137822857142858e-06,
+      "loss": 1.4745,
+      "step": 25923
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.137554285714285e-06,
+      "loss": 1.1834,
+      "step": 25924
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.137285714285714e-06,
+      "loss": 1.2098,
+      "step": 25925
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.137017142857143e-06,
+      "loss": 1.1954,
+      "step": 25926
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1367485714285715e-06,
+      "loss": 1.1245,
+      "step": 25927
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1364800000000005e-06,
+      "loss": 1.2008,
+      "step": 25928
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1362114285714295e-06,
+      "loss": 1.5172,
+      "step": 25929
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.135942857142857e-06,
+      "loss": 0.8951,
+      "step": 25930
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.135674285714286e-06,
+      "loss": 1.0839,
+      "step": 25931
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.135405714285715e-06,
+      "loss": 0.8759,
+      "step": 25932
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.135137142857144e-06,
+      "loss": 1.1903,
+      "step": 25933
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.134868571428571e-06,
+      "loss": 1.1944,
+      "step": 25934
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1346e-06,
+      "loss": 1.0501,
+      "step": 25935
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.134331428571429e-06,
+      "loss": 0.9781,
+      "step": 25936
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.134062857142857e-06,
+      "loss": 1.0043,
+      "step": 25937
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.133794285714286e-06,
+      "loss": 1.7294,
+      "step": 25938
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.133525714285715e-06,
+      "loss": 1.0788,
+      "step": 25939
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.133257142857143e-06,
+      "loss": 1.3728,
+      "step": 25940
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.132988571428572e-06,
+      "loss": 1.258,
+      "step": 25941
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.132720000000001e-06,
+      "loss": 1.1895,
+      "step": 25942
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.13245142857143e-06,
+      "loss": 1.3992,
+      "step": 25943
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.132182857142857e-06,
+      "loss": 0.8976,
+      "step": 25944
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.131914285714286e-06,
+      "loss": 0.9524,
+      "step": 25945
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.131645714285715e-06,
+      "loss": 1.3828,
+      "step": 25946
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.131377142857143e-06,
+      "loss": 0.9284,
+      "step": 25947
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.131108571428572e-06,
+      "loss": 0.9479,
+      "step": 25948
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.130840000000001e-06,
+      "loss": 1.4556,
+      "step": 25949
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.130571428571428e-06,
+      "loss": 0.6089,
+      "step": 25950
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.130302857142857e-06,
+      "loss": 1.2429,
+      "step": 25951
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.130034285714286e-06,
+      "loss": 1.3735,
+      "step": 25952
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.129765714285714e-06,
+      "loss": 1.0865,
+      "step": 25953
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.129497142857143e-06,
+      "loss": 1.0329,
+      "step": 25954
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.129228571428572e-06,
+      "loss": 0.883,
+      "step": 25955
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.128960000000001e-06,
+      "loss": 0.8152,
+      "step": 25956
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.128691428571429e-06,
+      "loss": 0.7614,
+      "step": 25957
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.128422857142858e-06,
+      "loss": 1.1719,
+      "step": 25958
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.128154285714287e-06,
+      "loss": 1.1613,
+      "step": 25959
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.127885714285714e-06,
+      "loss": 1.2504,
+      "step": 25960
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.127617142857143e-06,
+      "loss": 0.9403,
+      "step": 25961
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.127348571428572e-06,
+      "loss": 1.1591,
+      "step": 25962
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.127079999999999e-06,
+      "loss": 0.8466,
+      "step": 25963
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.126811428571428e-06,
+      "loss": 1.071,
+      "step": 25964
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.126542857142857e-06,
+      "loss": 0.6818,
+      "step": 25965
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1262742857142864e-06,
+      "loss": 0.7696,
+      "step": 25966
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.126005714285715e-06,
+      "loss": 1.0393,
+      "step": 25967
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.125737142857144e-06,
+      "loss": 0.8147,
+      "step": 25968
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.125468571428573e-06,
+      "loss": 1.2299,
+      "step": 25969
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.1252e-06,
+      "loss": 1.3006,
+      "step": 25970
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.124931428571429e-06,
+      "loss": 1.4469,
+      "step": 25971
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.124662857142858e-06,
+      "loss": 1.4192,
+      "step": 25972
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.124394285714285e-06,
+      "loss": 1.2185,
+      "step": 25973
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.124125714285714e-06,
+      "loss": 0.678,
+      "step": 25974
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.123857142857143e-06,
+      "loss": 1.1275,
+      "step": 25975
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.123588571428571e-06,
+      "loss": 1.2835,
+      "step": 25976
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.12332e-06,
+      "loss": 1.2126,
+      "step": 25977
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.123051428571429e-06,
+      "loss": 1.0123,
+      "step": 25978
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.122782857142858e-06,
+      "loss": 1.4684,
+      "step": 25979
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.122514285714286e-06,
+      "loss": 1.0915,
+      "step": 25980
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.122245714285715e-06,
+      "loss": 1.0268,
+      "step": 25981
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.121977142857144e-06,
+      "loss": 1.0685,
+      "step": 25982
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.121708571428571e-06,
+      "loss": 1.1648,
+      "step": 25983
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.12144e-06,
+      "loss": 0.8783,
+      "step": 25984
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.121171428571429e-06,
+      "loss": 1.2927,
+      "step": 25985
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.120902857142857e-06,
+      "loss": 1.4005,
+      "step": 25986
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.120634285714286e-06,
+      "loss": 1.3202,
+      "step": 25987
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.120365714285715e-06,
+      "loss": 1.2194,
+      "step": 25988
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.120097142857144e-06,
+      "loss": 1.0617,
+      "step": 25989
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.119828571428571e-06,
+      "loss": 1.0304,
+      "step": 25990
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.11956e-06,
+      "loss": 1.3114,
+      "step": 25991
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.119291428571429e-06,
+      "loss": 1.55,
+      "step": 25992
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.119022857142857e-06,
+      "loss": 1.3435,
+      "step": 25993
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.118754285714286e-06,
+      "loss": 0.9229,
+      "step": 25994
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.118485714285715e-06,
+      "loss": 1.2256,
+      "step": 25995
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.118217142857143e-06,
+      "loss": 1.493,
+      "step": 25996
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.117948571428572e-06,
+      "loss": 0.9206,
+      "step": 25997
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.117680000000001e-06,
+      "loss": 1.4771,
+      "step": 25998
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.117411428571428e-06,
+      "loss": 0.9179,
+      "step": 25999
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.117142857142857e-06,
+      "loss": 1.417,
+      "step": 26000
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.116874285714286e-06,
+      "loss": 1.2321,
+      "step": 26001
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.116605714285715e-06,
+      "loss": 0.8471,
+      "step": 26002
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1163371428571425e-06,
+      "loss": 1.1663,
+      "step": 26003
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1160685714285715e-06,
+      "loss": 1.1658,
+      "step": 26004
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1158000000000005e-06,
+      "loss": 0.8269,
+      "step": 26005
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.115531428571429e-06,
+      "loss": 1.3146,
+      "step": 26006
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.115262857142858e-06,
+      "loss": 0.9396,
+      "step": 26007
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.114994285714287e-06,
+      "loss": 1.5864,
+      "step": 26008
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.114725714285714e-06,
+      "loss": 1.3821,
+      "step": 26009
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.114457142857143e-06,
+      "loss": 0.9691,
+      "step": 26010
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.114188571428572e-06,
+      "loss": 1.122,
+      "step": 26011
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1139199999999995e-06,
+      "loss": 1.3961,
+      "step": 26012
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1136514285714285e-06,
+      "loss": 1.1454,
+      "step": 26013
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1133828571428575e-06,
+      "loss": 1.288,
+      "step": 26014
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1131142857142865e-06,
+      "loss": 1.3523,
+      "step": 26015
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.112845714285714e-06,
+      "loss": 0.8152,
+      "step": 26016
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.112577142857143e-06,
+      "loss": 1.3122,
+      "step": 26017
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.112308571428572e-06,
+      "loss": 0.9378,
+      "step": 26018
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.11204e-06,
+      "loss": 1.2858,
+      "step": 26019
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.111771428571429e-06,
+      "loss": 1.0428,
+      "step": 26020
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.111502857142858e-06,
+      "loss": 1.2426,
+      "step": 26021
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1112342857142855e-06,
+      "loss": 1.355,
+      "step": 26022
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1109657142857145e-06,
+      "loss": 1.3512,
+      "step": 26023
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1106971428571435e-06,
+      "loss": 1.2733,
+      "step": 26024
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1104285714285725e-06,
+      "loss": 0.9756,
+      "step": 26025
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.11016e-06,
+      "loss": 1.2771,
+      "step": 26026
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.109891428571429e-06,
+      "loss": 0.8278,
+      "step": 26027
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.109622857142858e-06,
+      "loss": 0.7743,
+      "step": 26028
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.109354285714285e-06,
+      "loss": 1.3247,
+      "step": 26029
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.109085714285714e-06,
+      "loss": 1.2958,
+      "step": 26030
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.108817142857143e-06,
+      "loss": 1.1683,
+      "step": 26031
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1085485714285715e-06,
+      "loss": 0.6452,
+      "step": 26032
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1082800000000005e-06,
+      "loss": 1.0873,
+      "step": 26033
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1080114285714295e-06,
+      "loss": 1.0607,
+      "step": 26034
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.107742857142857e-06,
+      "loss": 1.033,
+      "step": 26035
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.107474285714286e-06,
+      "loss": 1.2422,
+      "step": 26036
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.107205714285715e-06,
+      "loss": 0.8383,
+      "step": 26037
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.106937142857144e-06,
+      "loss": 1.2002,
+      "step": 26038
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.106668571428571e-06,
+      "loss": 1.6997,
+      "step": 26039
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.1064e-06,
+      "loss": 1.4008,
+      "step": 26040
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.106131428571429e-06,
+      "loss": 1.5958,
+      "step": 26041
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.105862857142857e-06,
+      "loss": 0.9715,
+      "step": 26042
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.105594285714286e-06,
+      "loss": 0.9793,
+      "step": 26043
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.105325714285715e-06,
+      "loss": 1.1251,
+      "step": 26044
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.105057142857143e-06,
+      "loss": 1.2357,
+      "step": 26045
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.104788571428572e-06,
+      "loss": 1.3636,
+      "step": 26046
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.104520000000001e-06,
+      "loss": 0.8192,
+      "step": 26047
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.10425142857143e-06,
+      "loss": 0.7528,
+      "step": 26048
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.103982857142857e-06,
+      "loss": 1.1759,
+      "step": 26049
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.103714285714286e-06,
+      "loss": 1.4157,
+      "step": 26050
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.103445714285715e-06,
+      "loss": 1.5472,
+      "step": 26051
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.103177142857143e-06,
+      "loss": 0.9218,
+      "step": 26052
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.102908571428572e-06,
+      "loss": 1.0745,
+      "step": 26053
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.102640000000001e-06,
+      "loss": 1.1887,
+      "step": 26054
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.102371428571428e-06,
+      "loss": 1.0649,
+      "step": 26055
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.102102857142857e-06,
+      "loss": 1.1649,
+      "step": 26056
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.101834285714286e-06,
+      "loss": 1.3024,
+      "step": 26057
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.101565714285714e-06,
+      "loss": 1.5201,
+      "step": 26058
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.101297142857143e-06,
+      "loss": 1.28,
+      "step": 26059
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.101028571428572e-06,
+      "loss": 1.3084,
+      "step": 26060
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.100760000000001e-06,
+      "loss": 0.896,
+      "step": 26061
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.100491428571429e-06,
+      "loss": 1.1852,
+      "step": 26062
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.100222857142858e-06,
+      "loss": 1.5188,
+      "step": 26063
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.099954285714287e-06,
+      "loss": 1.0361,
+      "step": 26064
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.099685714285714e-06,
+      "loss": 0.9032,
+      "step": 26065
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.099417142857143e-06,
+      "loss": 1.3323,
+      "step": 26066
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.099148571428572e-06,
+      "loss": 1.3161,
+      "step": 26067
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.098879999999999e-06,
+      "loss": 0.8758,
+      "step": 26068
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.098611428571428e-06,
+      "loss": 1.0622,
+      "step": 26069
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.098342857142857e-06,
+      "loss": 1.1203,
+      "step": 26070
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0980742857142864e-06,
+      "loss": 1.4151,
+      "step": 26071
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.097805714285715e-06,
+      "loss": 0.942,
+      "step": 26072
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.097537142857144e-06,
+      "loss": 0.9218,
+      "step": 26073
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.097268571428573e-06,
+      "loss": 1.4119,
+      "step": 26074
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.097e-06,
+      "loss": 1.2188,
+      "step": 26075
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.096731428571429e-06,
+      "loss": 0.9539,
+      "step": 26076
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.096462857142858e-06,
+      "loss": 1.0027,
+      "step": 26077
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.096194285714285e-06,
+      "loss": 1.1868,
+      "step": 26078
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.095925714285714e-06,
+      "loss": 1.1957,
+      "step": 26079
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.095657142857143e-06,
+      "loss": 1.2874,
+      "step": 26080
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.095388571428571e-06,
+      "loss": 1.0953,
+      "step": 26081
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.09512e-06,
+      "loss": 1.1406,
+      "step": 26082
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.094851428571429e-06,
+      "loss": 0.8661,
+      "step": 26083
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.094582857142858e-06,
+      "loss": 0.911,
+      "step": 26084
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.094314285714286e-06,
+      "loss": 0.9721,
+      "step": 26085
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.094045714285715e-06,
+      "loss": 1.0848,
+      "step": 26086
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.093777142857144e-06,
+      "loss": 1.1639,
+      "step": 26087
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.093508571428571e-06,
+      "loss": 1.2187,
+      "step": 26088
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.09324e-06,
+      "loss": 0.9722,
+      "step": 26089
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.092971428571429e-06,
+      "loss": 1.062,
+      "step": 26090
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.092702857142857e-06,
+      "loss": 1.3154,
+      "step": 26091
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.092434285714286e-06,
+      "loss": 1.0926,
+      "step": 26092
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.092165714285715e-06,
+      "loss": 1.3925,
+      "step": 26093
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.091897142857144e-06,
+      "loss": 1.1515,
+      "step": 26094
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.091628571428571e-06,
+      "loss": 1.3137,
+      "step": 26095
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.09136e-06,
+      "loss": 0.5864,
+      "step": 26096
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.091091428571429e-06,
+      "loss": 1.296,
+      "step": 26097
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.090822857142857e-06,
+      "loss": 1.1238,
+      "step": 26098
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.090554285714286e-06,
+      "loss": 0.9743,
+      "step": 26099
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.090285714285715e-06,
+      "loss": 1.0497,
+      "step": 26100
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.090017142857143e-06,
+      "loss": 1.7349,
+      "step": 26101
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.089748571428572e-06,
+      "loss": 1.752,
+      "step": 26102
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.089480000000001e-06,
+      "loss": 1.2457,
+      "step": 26103
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.089211428571428e-06,
+      "loss": 1.2712,
+      "step": 26104
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.088942857142857e-06,
+      "loss": 1.2016,
+      "step": 26105
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.088674285714286e-06,
+      "loss": 1.1796,
+      "step": 26106
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.088405714285715e-06,
+      "loss": 0.8662,
+      "step": 26107
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0881371428571425e-06,
+      "loss": 0.958,
+      "step": 26108
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0878685714285715e-06,
+      "loss": 1.3498,
+      "step": 26109
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0876000000000005e-06,
+      "loss": 1.0648,
+      "step": 26110
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.087331428571429e-06,
+      "loss": 1.3047,
+      "step": 26111
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.087062857142858e-06,
+      "loss": 0.9255,
+      "step": 26112
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.086794285714287e-06,
+      "loss": 0.9893,
+      "step": 26113
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.086525714285714e-06,
+      "loss": 1.1725,
+      "step": 26114
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.086257142857143e-06,
+      "loss": 1.2881,
+      "step": 26115
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.085988571428572e-06,
+      "loss": 0.7589,
+      "step": 26116
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.085720000000001e-06,
+      "loss": 0.9617,
+      "step": 26117
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0854514285714285e-06,
+      "loss": 1.5578,
+      "step": 26118
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0851828571428575e-06,
+      "loss": 0.9805,
+      "step": 26119
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0849142857142865e-06,
+      "loss": 0.8418,
+      "step": 26120
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.084645714285714e-06,
+      "loss": 0.8193,
+      "step": 26121
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.084377142857143e-06,
+      "loss": 0.4789,
+      "step": 26122
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.084108571428572e-06,
+      "loss": 1.1619,
+      "step": 26123
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.083839999999999e-06,
+      "loss": 1.1767,
+      "step": 26124
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.083571428571429e-06,
+      "loss": 1.2951,
+      "step": 26125
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.083302857142858e-06,
+      "loss": 1.1021,
+      "step": 26126
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0830342857142855e-06,
+      "loss": 1.2179,
+      "step": 26127
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0827657142857145e-06,
+      "loss": 1.2046,
+      "step": 26128
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0824971428571435e-06,
+      "loss": 1.6173,
+      "step": 26129
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0822285714285725e-06,
+      "loss": 0.5742,
+      "step": 26130
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.08196e-06,
+      "loss": 0.7883,
+      "step": 26131
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.081691428571429e-06,
+      "loss": 0.7166,
+      "step": 26132
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.081422857142858e-06,
+      "loss": 1.0774,
+      "step": 26133
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.081154285714285e-06,
+      "loss": 1.1407,
+      "step": 26134
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.080885714285714e-06,
+      "loss": 1.5709,
+      "step": 26135
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.080617142857143e-06,
+      "loss": 1.2362,
+      "step": 26136
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.080348571428571e-06,
+      "loss": 1.2074,
+      "step": 26137
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0800800000000005e-06,
+      "loss": 1.1548,
+      "step": 26138
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0798114285714295e-06,
+      "loss": 1.1719,
+      "step": 26139
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0795428571428585e-06,
+      "loss": 1.1202,
+      "step": 26140
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.079274285714286e-06,
+      "loss": 1.2639,
+      "step": 26141
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.079005714285715e-06,
+      "loss": 1.1996,
+      "step": 26142
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.078737142857144e-06,
+      "loss": 0.8989,
+      "step": 26143
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.078468571428571e-06,
+      "loss": 1.1636,
+      "step": 26144
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.0782e-06,
+      "loss": 1.079,
+      "step": 26145
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.077931428571429e-06,
+      "loss": 0.9796,
+      "step": 26146
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.077662857142857e-06,
+      "loss": 1.3795,
+      "step": 26147
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.077394285714286e-06,
+      "loss": 0.784,
+      "step": 26148
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.077125714285715e-06,
+      "loss": 1.0943,
+      "step": 26149
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.076857142857142e-06,
+      "loss": 1.1121,
+      "step": 26150
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.076588571428572e-06,
+      "loss": 1.2044,
+      "step": 26151
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.076320000000001e-06,
+      "loss": 1.1463,
+      "step": 26152
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.07605142857143e-06,
+      "loss": 0.8908,
+      "step": 26153
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.075782857142857e-06,
+      "loss": 1.1001,
+      "step": 26154
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.075514285714286e-06,
+      "loss": 1.0356,
+      "step": 26155
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.075245714285715e-06,
+      "loss": 1.0921,
+      "step": 26156
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.074977142857143e-06,
+      "loss": 1.0145,
+      "step": 26157
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.074708571428572e-06,
+      "loss": 1.2458,
+      "step": 26158
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.074440000000001e-06,
+      "loss": 1.1559,
+      "step": 26159
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.074171428571428e-06,
+      "loss": 1.254,
+      "step": 26160
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.073902857142857e-06,
+      "loss": 1.0679,
+      "step": 26161
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.073634285714286e-06,
+      "loss": 1.2254,
+      "step": 26162
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.073365714285715e-06,
+      "loss": 0.8713,
+      "step": 26163
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.073097142857143e-06,
+      "loss": 1.0875,
+      "step": 26164
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.072828571428572e-06,
+      "loss": 1.3271,
+      "step": 26165
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.072560000000001e-06,
+      "loss": 1.0876,
+      "step": 26166
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.072291428571429e-06,
+      "loss": 1.3201,
+      "step": 26167
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.072022857142858e-06,
+      "loss": 1.0625,
+      "step": 26168
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.071754285714287e-06,
+      "loss": 1.3576,
+      "step": 26169
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.071485714285714e-06,
+      "loss": 1.1484,
+      "step": 26170
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.071217142857143e-06,
+      "loss": 0.9043,
+      "step": 26171
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.070948571428572e-06,
+      "loss": 1.1131,
+      "step": 26172
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.070679999999999e-06,
+      "loss": 1.2695,
+      "step": 26173
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.070411428571428e-06,
+      "loss": 0.9947,
+      "step": 26174
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.070142857142857e-06,
+      "loss": 1.0724,
+      "step": 26175
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0698742857142864e-06,
+      "loss": 1.2377,
+      "step": 26176
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.069605714285714e-06,
+      "loss": 1.2713,
+      "step": 26177
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.069337142857144e-06,
+      "loss": 1.1089,
+      "step": 26178
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.069068571428573e-06,
+      "loss": 1.0811,
+      "step": 26179
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0688e-06,
+      "loss": 1.2784,
+      "step": 26180
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.068531428571429e-06,
+      "loss": 0.873,
+      "step": 26181
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.068262857142858e-06,
+      "loss": 0.955,
+      "step": 26182
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.067994285714285e-06,
+      "loss": 1.4997,
+      "step": 26183
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.067725714285714e-06,
+      "loss": 1.0812,
+      "step": 26184
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.067457142857143e-06,
+      "loss": 1.1398,
+      "step": 26185
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.067188571428571e-06,
+      "loss": 1.3719,
+      "step": 26186
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.06692e-06,
+      "loss": 0.9818,
+      "step": 26187
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.066651428571429e-06,
+      "loss": 1.0626,
+      "step": 26188
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.066382857142858e-06,
+      "loss": 1.7472,
+      "step": 26189
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.066114285714285e-06,
+      "loss": 1.1297,
+      "step": 26190
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.065845714285715e-06,
+      "loss": 0.8394,
+      "step": 26191
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.065577142857144e-06,
+      "loss": 0.9364,
+      "step": 26192
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.065308571428571e-06,
+      "loss": 1.4581,
+      "step": 26193
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.06504e-06,
+      "loss": 0.9781,
+      "step": 26194
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.064771428571429e-06,
+      "loss": 0.945,
+      "step": 26195
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.064502857142857e-06,
+      "loss": 1.2107,
+      "step": 26196
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.064234285714286e-06,
+      "loss": 1.4164,
+      "step": 26197
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.063965714285715e-06,
+      "loss": 1.1752,
+      "step": 26198
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.063697142857144e-06,
+      "loss": 1.0355,
+      "step": 26199
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.063428571428571e-06,
+      "loss": 0.9248,
+      "step": 26200
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.06316e-06,
+      "loss": 1.3523,
+      "step": 26201
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.062891428571429e-06,
+      "loss": 0.9592,
+      "step": 26202
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0626228571428565e-06,
+      "loss": 0.8636,
+      "step": 26203
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.062354285714286e-06,
+      "loss": 1.3482,
+      "step": 26204
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.062085714285715e-06,
+      "loss": 0.8487,
+      "step": 26205
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.061817142857143e-06,
+      "loss": 1.1922,
+      "step": 26206
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.061548571428572e-06,
+      "loss": 1.0742,
+      "step": 26207
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.061280000000001e-06,
+      "loss": 0.9847,
+      "step": 26208
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.061011428571428e-06,
+      "loss": 1.0265,
+      "step": 26209
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.060742857142857e-06,
+      "loss": 1.5445,
+      "step": 26210
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.060474285714286e-06,
+      "loss": 0.8778,
+      "step": 26211
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.060205714285715e-06,
+      "loss": 1.0399,
+      "step": 26212
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0599371428571425e-06,
+      "loss": 1.0653,
+      "step": 26213
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0596685714285715e-06,
+      "loss": 1.011,
+      "step": 26214
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0594000000000005e-06,
+      "loss": 0.7017,
+      "step": 26215
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.059131428571428e-06,
+      "loss": 1.0967,
+      "step": 26216
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.058862857142858e-06,
+      "loss": 0.6953,
+      "step": 26217
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.058594285714287e-06,
+      "loss": 1.4228,
+      "step": 26218
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.058325714285714e-06,
+      "loss": 1.334,
+      "step": 26219
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.058057142857143e-06,
+      "loss": 1.2577,
+      "step": 26220
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.057788571428572e-06,
+      "loss": 1.1282,
+      "step": 26221
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.057520000000001e-06,
+      "loss": 1.1274,
+      "step": 26222
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0572514285714285e-06,
+      "loss": 0.5867,
+      "step": 26223
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0569828571428575e-06,
+      "loss": 0.765,
+      "step": 26224
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0567142857142865e-06,
+      "loss": 1.1065,
+      "step": 26225
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.056445714285714e-06,
+      "loss": 1.2422,
+      "step": 26226
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.056177142857143e-06,
+      "loss": 1.3748,
+      "step": 26227
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.055908571428572e-06,
+      "loss": 1.1416,
+      "step": 26228
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.055639999999999e-06,
+      "loss": 1.2396,
+      "step": 26229
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.055371428571428e-06,
+      "loss": 1.2012,
+      "step": 26230
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.055102857142858e-06,
+      "loss": 0.7775,
+      "step": 26231
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0548342857142855e-06,
+      "loss": 1.3935,
+      "step": 26232
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0545657142857145e-06,
+      "loss": 1.1343,
+      "step": 26233
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0542971428571435e-06,
+      "loss": 0.8539,
+      "step": 26234
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0540285714285725e-06,
+      "loss": 1.4339,
+      "step": 26235
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.05376e-06,
+      "loss": 1.4532,
+      "step": 26236
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.053491428571429e-06,
+      "loss": 1.6052,
+      "step": 26237
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.053222857142858e-06,
+      "loss": 0.8881,
+      "step": 26238
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.052954285714285e-06,
+      "loss": 1.1232,
+      "step": 26239
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.052685714285714e-06,
+      "loss": 1.3243,
+      "step": 26240
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.052417142857143e-06,
+      "loss": 1.092,
+      "step": 26241
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.052148571428571e-06,
+      "loss": 1.083,
+      "step": 26242
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.05188e-06,
+      "loss": 1.2487,
+      "step": 26243
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0516114285714295e-06,
+      "loss": 1.0156,
+      "step": 26244
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0513428571428585e-06,
+      "loss": 1.1404,
+      "step": 26245
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.051074285714286e-06,
+      "loss": 1.141,
+      "step": 26246
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.050805714285715e-06,
+      "loss": 1.3649,
+      "step": 26247
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.050537142857144e-06,
+      "loss": 1.0657,
+      "step": 26248
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.050268571428571e-06,
+      "loss": 0.8586,
+      "step": 26249
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.05e-06,
+      "loss": 1.2406,
+      "step": 26250
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0497314285714284e-06,
+      "loss": 1.1742,
+      "step": 26251
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0494628571428575e-06,
+      "loss": 1.2144,
+      "step": 26252
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.049194285714286e-06,
+      "loss": 1.2037,
+      "step": 26253
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.048925714285715e-06,
+      "loss": 1.3564,
+      "step": 26254
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.048657142857143e-06,
+      "loss": 1.3352,
+      "step": 26255
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.048388571428571e-06,
+      "loss": 1.4288,
+      "step": 26256
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.048120000000001e-06,
+      "loss": 1.0353,
+      "step": 26257
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.047851428571429e-06,
+      "loss": 1.5301,
+      "step": 26258
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.047582857142858e-06,
+      "loss": 1.3978,
+      "step": 26259
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.047314285714286e-06,
+      "loss": 0.8772,
+      "step": 26260
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0470457142857144e-06,
+      "loss": 1.0399,
+      "step": 26261
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0467771428571435e-06,
+      "loss": 1.3448,
+      "step": 26262
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.046508571428572e-06,
+      "loss": 0.7276,
+      "step": 26263
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.04624e-06,
+      "loss": 1.0058,
+      "step": 26264
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.045971428571429e-06,
+      "loss": 1.1942,
+      "step": 26265
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.045702857142857e-06,
+      "loss": 1.3523,
+      "step": 26266
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.045434285714286e-06,
+      "loss": 0.923,
+      "step": 26267
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.045165714285714e-06,
+      "loss": 0.8843,
+      "step": 26268
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.044897142857142e-06,
+      "loss": 1.1889,
+      "step": 26269
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.044628571428572e-06,
+      "loss": 1.1666,
+      "step": 26270
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0443600000000004e-06,
+      "loss": 1.3437,
+      "step": 26271
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0440914285714295e-06,
+      "loss": 1.1414,
+      "step": 26272
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.043822857142858e-06,
+      "loss": 0.9627,
+      "step": 26273
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.043554285714286e-06,
+      "loss": 1.2865,
+      "step": 26274
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.043285714285715e-06,
+      "loss": 1.2687,
+      "step": 26275
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.043017142857143e-06,
+      "loss": 1.3721,
+      "step": 26276
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.042748571428572e-06,
+      "loss": 1.231,
+      "step": 26277
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.04248e-06,
+      "loss": 1.0386,
+      "step": 26278
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.042211428571428e-06,
+      "loss": 1.1634,
+      "step": 26279
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.041942857142857e-06,
+      "loss": 1.1247,
+      "step": 26280
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.041674285714286e-06,
+      "loss": 0.9542,
+      "step": 26281
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.041405714285715e-06,
+      "loss": 1.2241,
+      "step": 26282
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.041137142857143e-06,
+      "loss": 1.1718,
+      "step": 26283
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.040868571428572e-06,
+      "loss": 0.7185,
+      "step": 26284
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.040600000000001e-06,
+      "loss": 1.0838,
+      "step": 26285
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.040331428571429e-06,
+      "loss": 1.5796,
+      "step": 26286
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.040062857142857e-06,
+      "loss": 1.2362,
+      "step": 26287
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.039794285714286e-06,
+      "loss": 1.1843,
+      "step": 26288
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.039525714285714e-06,
+      "loss": 0.8676,
+      "step": 26289
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.039257142857143e-06,
+      "loss": 1.2568,
+      "step": 26290
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.038988571428572e-06,
+      "loss": 1.1505,
+      "step": 26291
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.03872e-06,
+      "loss": 0.9073,
+      "step": 26292
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.038451428571429e-06,
+      "loss": 1.1977,
+      "step": 26293
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.038182857142857e-06,
+      "loss": 1.329,
+      "step": 26294
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.037914285714286e-06,
+      "loss": 1.1524,
+      "step": 26295
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.037645714285714e-06,
+      "loss": 0.8398,
+      "step": 26296
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.037377142857143e-06,
+      "loss": 0.6996,
+      "step": 26297
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.037108571428572e-06,
+      "loss": 1.0871,
+      "step": 26298
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.03684e-06,
+      "loss": 0.8983,
+      "step": 26299
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.036571428571429e-06,
+      "loss": 1.499,
+      "step": 26300
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0363028571428576e-06,
+      "loss": 1.2016,
+      "step": 26301
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.036034285714286e-06,
+      "loss": 0.9574,
+      "step": 26302
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.035765714285715e-06,
+      "loss": 1.1188,
+      "step": 26303
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.035497142857143e-06,
+      "loss": 1.3254,
+      "step": 26304
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.035228571428571e-06,
+      "loss": 1.279,
+      "step": 26305
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.03496e-06,
+      "loss": 1.113,
+      "step": 26306
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.034691428571428e-06,
+      "loss": 0.898,
+      "step": 26307
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.034422857142857e-06,
+      "loss": 0.9216,
+      "step": 26308
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0341542857142855e-06,
+      "loss": 0.7761,
+      "step": 26309
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0338857142857145e-06,
+      "loss": 1.3204,
+      "step": 26310
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0336171428571436e-06,
+      "loss": 0.9053,
+      "step": 26311
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.033348571428572e-06,
+      "loss": 1.4506,
+      "step": 26312
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.033080000000001e-06,
+      "loss": 0.8993,
+      "step": 26313
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.032811428571429e-06,
+      "loss": 1.1336,
+      "step": 26314
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.032542857142857e-06,
+      "loss": 0.9323,
+      "step": 26315
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.032274285714286e-06,
+      "loss": 1.0857,
+      "step": 26316
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.032005714285714e-06,
+      "loss": 1.0964,
+      "step": 26317
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.031737142857143e-06,
+      "loss": 0.8112,
+      "step": 26318
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0314685714285715e-06,
+      "loss": 0.627,
+      "step": 26319
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0312e-06,
+      "loss": 0.7938,
+      "step": 26320
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.030931428571429e-06,
+      "loss": 1.2384,
+      "step": 26321
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.030662857142857e-06,
+      "loss": 1.2517,
+      "step": 26322
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.030394285714287e-06,
+      "loss": 1.2725,
+      "step": 26323
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.030125714285715e-06,
+      "loss": 0.9583,
+      "step": 26324
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.029857142857143e-06,
+      "loss": 0.87,
+      "step": 26325
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.029588571428572e-06,
+      "loss": 1.3489,
+      "step": 26326
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.02932e-06,
+      "loss": 1.4448,
+      "step": 26327
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0290514285714285e-06,
+      "loss": 1.1607,
+      "step": 26328
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.0287828571428575e-06,
+      "loss": 1.1924,
+      "step": 26329
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.028514285714286e-06,
+      "loss": 1.5749,
+      "step": 26330
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.028245714285715e-06,
+      "loss": 1.2161,
+      "step": 26331
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.027977142857143e-06,
+      "loss": 0.978,
+      "step": 26332
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.027708571428571e-06,
+      "loss": 1.0179,
+      "step": 26333
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.02744e-06,
+      "loss": 1.148,
+      "step": 26334
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.027171428571428e-06,
+      "loss": 0.9953,
+      "step": 26335
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.026902857142857e-06,
+      "loss": 1.6829,
+      "step": 26336
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.026634285714286e-06,
+      "loss": 1.1448,
+      "step": 26337
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0263657142857145e-06,
+      "loss": 1.566,
+      "step": 26338
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0260971428571435e-06,
+      "loss": 0.9664,
+      "step": 26339
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.025828571428572e-06,
+      "loss": 0.726,
+      "step": 26340
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.025560000000001e-06,
+      "loss": 1.1403,
+      "step": 26341
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.025291428571429e-06,
+      "loss": 1.1285,
+      "step": 26342
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.025022857142857e-06,
+      "loss": 1.0658,
+      "step": 26343
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.024754285714286e-06,
+      "loss": 1.2177,
+      "step": 26344
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.024485714285714e-06,
+      "loss": 0.7449,
+      "step": 26345
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.024217142857143e-06,
+      "loss": 0.9623,
+      "step": 26346
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0239485714285715e-06,
+      "loss": 1.6859,
+      "step": 26347
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.02368e-06,
+      "loss": 0.9398,
+      "step": 26348
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.023411428571429e-06,
+      "loss": 1.4968,
+      "step": 26349
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.023142857142858e-06,
+      "loss": 1.074,
+      "step": 26350
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.022874285714286e-06,
+      "loss": 1.0545,
+      "step": 26351
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.022605714285715e-06,
+      "loss": 1.3671,
+      "step": 26352
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.022337142857143e-06,
+      "loss": 1.0214,
+      "step": 26353
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.022068571428572e-06,
+      "loss": 1.3526,
+      "step": 26354
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0218e-06,
+      "loss": 1.3601,
+      "step": 26355
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0215314285714284e-06,
+      "loss": 0.9274,
+      "step": 26356
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0212628571428575e-06,
+      "loss": 1.2548,
+      "step": 26357
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.020994285714286e-06,
+      "loss": 1.0374,
+      "step": 26358
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.020725714285715e-06,
+      "loss": 1.5609,
+      "step": 26359
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.020457142857143e-06,
+      "loss": 1.087,
+      "step": 26360
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.020188571428571e-06,
+      "loss": 0.8978,
+      "step": 26361
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.01992e-06,
+      "loss": 0.9458,
+      "step": 26362
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.019651428571429e-06,
+      "loss": 1.4977,
+      "step": 26363
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.019382857142858e-06,
+      "loss": 1.329,
+      "step": 26364
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.019114285714286e-06,
+      "loss": 0.9603,
+      "step": 26365
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0188457142857144e-06,
+      "loss": 1.2138,
+      "step": 26366
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0185771428571435e-06,
+      "loss": 1.2602,
+      "step": 26367
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.018308571428572e-06,
+      "loss": 1.5735,
+      "step": 26368
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.018040000000001e-06,
+      "loss": 1.4193,
+      "step": 26369
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.017771428571429e-06,
+      "loss": 0.8392,
+      "step": 26370
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.017502857142857e-06,
+      "loss": 1.4823,
+      "step": 26371
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.017234285714286e-06,
+      "loss": 1.1048,
+      "step": 26372
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.016965714285714e-06,
+      "loss": 1.4372,
+      "step": 26373
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.016697142857142e-06,
+      "loss": 1.1704,
+      "step": 26374
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.016428571428571e-06,
+      "loss": 1.5165,
+      "step": 26375
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0161600000000004e-06,
+      "loss": 1.0495,
+      "step": 26376
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0158914285714295e-06,
+      "loss": 1.3525,
+      "step": 26377
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.015622857142858e-06,
+      "loss": 1.5617,
+      "step": 26378
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.015354285714286e-06,
+      "loss": 1.177,
+      "step": 26379
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.015085714285715e-06,
+      "loss": 0.957,
+      "step": 26380
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.014817142857143e-06,
+      "loss": 1.2012,
+      "step": 26381
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.014548571428572e-06,
+      "loss": 1.3482,
+      "step": 26382
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.01428e-06,
+      "loss": 0.6995,
+      "step": 26383
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.014011428571428e-06,
+      "loss": 1.1278,
+      "step": 26384
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.013742857142857e-06,
+      "loss": 1.1832,
+      "step": 26385
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.013474285714286e-06,
+      "loss": 0.7861,
+      "step": 26386
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.013205714285715e-06,
+      "loss": 0.9913,
+      "step": 26387
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.012937142857143e-06,
+      "loss": 1.2328,
+      "step": 26388
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.012668571428572e-06,
+      "loss": 1.0836,
+      "step": 26389
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.012400000000001e-06,
+      "loss": 1.0175,
+      "step": 26390
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.012131428571429e-06,
+      "loss": 1.0137,
+      "step": 26391
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.011862857142857e-06,
+      "loss": 1.3267,
+      "step": 26392
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.011594285714286e-06,
+      "loss": 1.1662,
+      "step": 26393
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.011325714285714e-06,
+      "loss": 1.019,
+      "step": 26394
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.011057142857143e-06,
+      "loss": 1.1597,
+      "step": 26395
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.010788571428572e-06,
+      "loss": 1.0491,
+      "step": 26396
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.01052e-06,
+      "loss": 0.8816,
+      "step": 26397
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.010251428571429e-06,
+      "loss": 0.7999,
+      "step": 26398
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.009982857142857e-06,
+      "loss": 1.4834,
+      "step": 26399
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.009714285714286e-06,
+      "loss": 1.3123,
+      "step": 26400
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.009445714285714e-06,
+      "loss": 1.1312,
+      "step": 26401
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.009177142857143e-06,
+      "loss": 1.1373,
+      "step": 26402
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.008908571428572e-06,
+      "loss": 1.005,
+      "step": 26403
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.00864e-06,
+      "loss": 1.7041,
+      "step": 26404
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.008371428571429e-06,
+      "loss": 0.9693,
+      "step": 26405
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0081028571428576e-06,
+      "loss": 0.7997,
+      "step": 26406
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.007834285714286e-06,
+      "loss": 1.2253,
+      "step": 26407
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.007565714285715e-06,
+      "loss": 0.7902,
+      "step": 26408
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.007297142857143e-06,
+      "loss": 1.4011,
+      "step": 26409
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.007028571428572e-06,
+      "loss": 1.3218,
+      "step": 26410
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.00676e-06,
+      "loss": 0.8288,
+      "step": 26411
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.006491428571428e-06,
+      "loss": 1.3446,
+      "step": 26412
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.006222857142857e-06,
+      "loss": 1.4265,
+      "step": 26413
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0059542857142855e-06,
+      "loss": 1.425,
+      "step": 26414
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0056857142857145e-06,
+      "loss": 1.2916,
+      "step": 26415
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0054171428571436e-06,
+      "loss": 1.2317,
+      "step": 26416
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.005148571428572e-06,
+      "loss": 1.8289,
+      "step": 26417
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.004880000000001e-06,
+      "loss": 0.9539,
+      "step": 26418
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.004611428571429e-06,
+      "loss": 1.0986,
+      "step": 26419
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.004342857142857e-06,
+      "loss": 1.085,
+      "step": 26420
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.004074285714286e-06,
+      "loss": 1.267,
+      "step": 26421
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.003805714285714e-06,
+      "loss": 0.9903,
+      "step": 26422
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.003537142857143e-06,
+      "loss": 0.9344,
+      "step": 26423
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0032685714285715e-06,
+      "loss": 1.2025,
+      "step": 26424
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.003e-06,
+      "loss": 1.1348,
+      "step": 26425
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.002731428571429e-06,
+      "loss": 1.3223,
+      "step": 26426
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.002462857142857e-06,
+      "loss": 0.9611,
+      "step": 26427
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.002194285714286e-06,
+      "loss": 1.2091,
+      "step": 26428
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.001925714285715e-06,
+      "loss": 0.7604,
+      "step": 26429
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.001657142857143e-06,
+      "loss": 1.2718,
+      "step": 26430
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.001388571428572e-06,
+      "loss": 1.0632,
+      "step": 26431
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.00112e-06,
+      "loss": 1.2389,
+      "step": 26432
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.000851428571429e-06,
+      "loss": 1.2722,
+      "step": 26433
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.0005828571428575e-06,
+      "loss": 1.0938,
+      "step": 26434
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.000314285714286e-06,
+      "loss": 1.0257,
+      "step": 26435
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.000045714285715e-06,
+      "loss": 1.4611,
+      "step": 26436
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.999777142857143e-06,
+      "loss": 1.6849,
+      "step": 26437
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.999508571428571e-06,
+      "loss": 1.0642,
+      "step": 26438
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.99924e-06,
+      "loss": 1.0581,
+      "step": 26439
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.998971428571428e-06,
+      "loss": 0.9903,
+      "step": 26440
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.998702857142857e-06,
+      "loss": 1.2518,
+      "step": 26441
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.998434285714286e-06,
+      "loss": 1.276,
+      "step": 26442
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9981657142857145e-06,
+      "loss": 0.7928,
+      "step": 26443
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9978971428571435e-06,
+      "loss": 1.3364,
+      "step": 26444
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.997628571428572e-06,
+      "loss": 1.2472,
+      "step": 26445
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.997360000000001e-06,
+      "loss": 0.8854,
+      "step": 26446
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.997091428571429e-06,
+      "loss": 1.1795,
+      "step": 26447
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.996822857142857e-06,
+      "loss": 0.9954,
+      "step": 26448
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.996554285714286e-06,
+      "loss": 0.8374,
+      "step": 26449
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.996285714285714e-06,
+      "loss": 0.93,
+      "step": 26450
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.996017142857143e-06,
+      "loss": 0.8424,
+      "step": 26451
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9957485714285715e-06,
+      "loss": 0.9389,
+      "step": 26452
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.99548e-06,
+      "loss": 1.0156,
+      "step": 26453
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.995211428571429e-06,
+      "loss": 1.4686,
+      "step": 26454
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.994942857142858e-06,
+      "loss": 1.3253,
+      "step": 26455
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.994674285714287e-06,
+      "loss": 0.9967,
+      "step": 26456
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.994405714285715e-06,
+      "loss": 1.2391,
+      "step": 26457
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.994137142857143e-06,
+      "loss": 1.3527,
+      "step": 26458
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.993868571428572e-06,
+      "loss": 1.1285,
+      "step": 26459
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9936e-06,
+      "loss": 1.314,
+      "step": 26460
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9933314285714284e-06,
+      "loss": 1.1341,
+      "step": 26461
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9930628571428575e-06,
+      "loss": 1.2038,
+      "step": 26462
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.992794285714286e-06,
+      "loss": 1.3883,
+      "step": 26463
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.992525714285715e-06,
+      "loss": 1.6392,
+      "step": 26464
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.992257142857143e-06,
+      "loss": 1.3194,
+      "step": 26465
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.991988571428571e-06,
+      "loss": 1.1509,
+      "step": 26466
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.99172e-06,
+      "loss": 1.0984,
+      "step": 26467
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.991451428571429e-06,
+      "loss": 1.3033,
+      "step": 26468
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.991182857142858e-06,
+      "loss": 0.6491,
+      "step": 26469
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.990914285714286e-06,
+      "loss": 1.6127,
+      "step": 26470
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9906457142857144e-06,
+      "loss": 1.4203,
+      "step": 26471
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9903771428571435e-06,
+      "loss": 0.8351,
+      "step": 26472
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.990108571428572e-06,
+      "loss": 1.0638,
+      "step": 26473
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.989840000000001e-06,
+      "loss": 1.2769,
+      "step": 26474
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.989571428571429e-06,
+      "loss": 1.3609,
+      "step": 26475
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.989302857142857e-06,
+      "loss": 1.0097,
+      "step": 26476
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.989034285714286e-06,
+      "loss": 0.6589,
+      "step": 26477
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.988765714285714e-06,
+      "loss": 1.0418,
+      "step": 26478
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.988497142857142e-06,
+      "loss": 1.4142,
+      "step": 26479
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.988228571428571e-06,
+      "loss": 1.0436,
+      "step": 26480
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9879600000000004e-06,
+      "loss": 1.1132,
+      "step": 26481
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9876914285714295e-06,
+      "loss": 1.305,
+      "step": 26482
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.987422857142858e-06,
+      "loss": 0.824,
+      "step": 26483
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.987154285714286e-06,
+      "loss": 1.4304,
+      "step": 26484
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.986885714285715e-06,
+      "loss": 1.3184,
+      "step": 26485
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.986617142857143e-06,
+      "loss": 1.2241,
+      "step": 26486
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.986348571428572e-06,
+      "loss": 0.913,
+      "step": 26487
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.98608e-06,
+      "loss": 1.148,
+      "step": 26488
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.985811428571428e-06,
+      "loss": 1.23,
+      "step": 26489
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.985542857142857e-06,
+      "loss": 1.193,
+      "step": 26490
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.985274285714286e-06,
+      "loss": 1.3374,
+      "step": 26491
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.985005714285715e-06,
+      "loss": 1.1099,
+      "step": 26492
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.984737142857143e-06,
+      "loss": 0.8561,
+      "step": 26493
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.984468571428572e-06,
+      "loss": 1.1117,
+      "step": 26494
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.984200000000001e-06,
+      "loss": 0.8691,
+      "step": 26495
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.983931428571429e-06,
+      "loss": 0.9654,
+      "step": 26496
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.983662857142858e-06,
+      "loss": 1.2778,
+      "step": 26497
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.983394285714286e-06,
+      "loss": 0.7374,
+      "step": 26498
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.983125714285714e-06,
+      "loss": 0.804,
+      "step": 26499
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.982857142857143e-06,
+      "loss": 1.3257,
+      "step": 26500
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9825885714285716e-06,
+      "loss": 1.0698,
+      "step": 26501
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.98232e-06,
+      "loss": 0.7909,
+      "step": 26502
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.982051428571429e-06,
+      "loss": 1.488,
+      "step": 26503
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.981782857142857e-06,
+      "loss": 1.0042,
+      "step": 26504
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.981514285714286e-06,
+      "loss": 1.4352,
+      "step": 26505
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.981245714285714e-06,
+      "loss": 0.6822,
+      "step": 26506
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.980977142857143e-06,
+      "loss": 1.3963,
+      "step": 26507
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.980708571428572e-06,
+      "loss": 1.1379,
+      "step": 26508
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.98044e-06,
+      "loss": 1.306,
+      "step": 26509
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.980171428571429e-06,
+      "loss": 1.2629,
+      "step": 26510
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.9799028571428576e-06,
+      "loss": 1.0858,
+      "step": 26511
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 6.979634285714286e-06,
+      "loss": 1.2416,
+      "step": 26512
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.979365714285715e-06,
+      "loss": 1.0388,
+      "step": 26513
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.979097142857143e-06,
+      "loss": 1.0188,
+      "step": 26514
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.978828571428572e-06,
+      "loss": 1.0078,
+      "step": 26515
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.97856e-06,
+      "loss": 0.8543,
+      "step": 26516
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.978291428571428e-06,
+      "loss": 0.7736,
+      "step": 26517
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.978022857142857e-06,
+      "loss": 1.0304,
+      "step": 26518
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9777542857142855e-06,
+      "loss": 1.0775,
+      "step": 26519
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9774857142857145e-06,
+      "loss": 1.0599,
+      "step": 26520
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9772171428571436e-06,
+      "loss": 0.8967,
+      "step": 26521
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.976948571428572e-06,
+      "loss": 1.1312,
+      "step": 26522
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.976680000000001e-06,
+      "loss": 0.9498,
+      "step": 26523
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.976411428571429e-06,
+      "loss": 1.3063,
+      "step": 26524
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.976142857142857e-06,
+      "loss": 1.09,
+      "step": 26525
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.975874285714286e-06,
+      "loss": 1.4277,
+      "step": 26526
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.975605714285714e-06,
+      "loss": 0.6935,
+      "step": 26527
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.975337142857143e-06,
+      "loss": 1.027,
+      "step": 26528
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9750685714285715e-06,
+      "loss": 1.0501,
+      "step": 26529
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9748e-06,
+      "loss": 0.9798,
+      "step": 26530
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.974531428571429e-06,
+      "loss": 1.4405,
+      "step": 26531
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.974262857142857e-06,
+      "loss": 1.121,
+      "step": 26532
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.973994285714286e-06,
+      "loss": 1.3959,
+      "step": 26533
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.973725714285715e-06,
+      "loss": 1.3994,
+      "step": 26534
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.973457142857143e-06,
+      "loss": 1.7286,
+      "step": 26535
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.973188571428572e-06,
+      "loss": 1.3437,
+      "step": 26536
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.97292e-06,
+      "loss": 1.0495,
+      "step": 26537
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.972651428571429e-06,
+      "loss": 1.2761,
+      "step": 26538
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9723828571428575e-06,
+      "loss": 1.0529,
+      "step": 26539
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.972114285714286e-06,
+      "loss": 1.1148,
+      "step": 26540
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.971845714285715e-06,
+      "loss": 1.015,
+      "step": 26541
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.971577142857143e-06,
+      "loss": 1.2235,
+      "step": 26542
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.971308571428571e-06,
+      "loss": 1.4787,
+      "step": 26543
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.97104e-06,
+      "loss": 1.0221,
+      "step": 26544
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.970771428571428e-06,
+      "loss": 0.942,
+      "step": 26545
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.970502857142857e-06,
+      "loss": 1.9049,
+      "step": 26546
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.970234285714286e-06,
+      "loss": 0.8832,
+      "step": 26547
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9699657142857145e-06,
+      "loss": 1.017,
+      "step": 26548
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9696971428571435e-06,
+      "loss": 0.8357,
+      "step": 26549
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.969428571428572e-06,
+      "loss": 0.815,
+      "step": 26550
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.969160000000001e-06,
+      "loss": 1.4195,
+      "step": 26551
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.968891428571429e-06,
+      "loss": 1.374,
+      "step": 26552
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.968622857142857e-06,
+      "loss": 1.2794,
+      "step": 26553
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.968354285714286e-06,
+      "loss": 1.2402,
+      "step": 26554
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.968085714285714e-06,
+      "loss": 1.2719,
+      "step": 26555
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.967817142857143e-06,
+      "loss": 0.9602,
+      "step": 26556
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9675485714285715e-06,
+      "loss": 0.92,
+      "step": 26557
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.96728e-06,
+      "loss": 1.6147,
+      "step": 26558
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.967011428571429e-06,
+      "loss": 0.7768,
+      "step": 26559
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.966742857142858e-06,
+      "loss": 1.958,
+      "step": 26560
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.966474285714287e-06,
+      "loss": 1.2071,
+      "step": 26561
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.966205714285715e-06,
+      "loss": 1.0899,
+      "step": 26562
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.965937142857143e-06,
+      "loss": 1.1983,
+      "step": 26563
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.965668571428572e-06,
+      "loss": 1.5547,
+      "step": 26564
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9654e-06,
+      "loss": 0.7706,
+      "step": 26565
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9651314285714284e-06,
+      "loss": 0.7809,
+      "step": 26566
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9648628571428575e-06,
+      "loss": 1.4133,
+      "step": 26567
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.964594285714286e-06,
+      "loss": 1.0825,
+      "step": 26568
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.964325714285715e-06,
+      "loss": 1.1409,
+      "step": 26569
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.964057142857143e-06,
+      "loss": 1.3487,
+      "step": 26570
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.963788571428571e-06,
+      "loss": 1.3452,
+      "step": 26571
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.96352e-06,
+      "loss": 1.4738,
+      "step": 26572
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.963251428571429e-06,
+      "loss": 0.8349,
+      "step": 26573
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.962982857142858e-06,
+      "loss": 1.2543,
+      "step": 26574
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.962714285714286e-06,
+      "loss": 1.238,
+      "step": 26575
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9624457142857144e-06,
+      "loss": 1.0492,
+      "step": 26576
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9621771428571435e-06,
+      "loss": 0.8441,
+      "step": 26577
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.961908571428572e-06,
+      "loss": 0.8204,
+      "step": 26578
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.961640000000001e-06,
+      "loss": 1.0151,
+      "step": 26579
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.961371428571429e-06,
+      "loss": 1.1468,
+      "step": 26580
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.961102857142857e-06,
+      "loss": 1.0813,
+      "step": 26581
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.960834285714286e-06,
+      "loss": 1.1512,
+      "step": 26582
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.960565714285714e-06,
+      "loss": 0.9576,
+      "step": 26583
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.960297142857143e-06,
+      "loss": 1.2346,
+      "step": 26584
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.960028571428571e-06,
+      "loss": 0.9159,
+      "step": 26585
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9597600000000004e-06,
+      "loss": 1.3731,
+      "step": 26586
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9594914285714295e-06,
+      "loss": 1.2725,
+      "step": 26587
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.959222857142858e-06,
+      "loss": 1.1702,
+      "step": 26588
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.958954285714286e-06,
+      "loss": 1.3641,
+      "step": 26589
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.958685714285715e-06,
+      "loss": 1.1888,
+      "step": 26590
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.958417142857143e-06,
+      "loss": 1.5877,
+      "step": 26591
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.958148571428572e-06,
+      "loss": 1.3995,
+      "step": 26592
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.95788e-06,
+      "loss": 1.4147,
+      "step": 26593
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.957611428571428e-06,
+      "loss": 1.0533,
+      "step": 26594
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.957342857142857e-06,
+      "loss": 1.7579,
+      "step": 26595
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.957074285714286e-06,
+      "loss": 1.1014,
+      "step": 26596
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.956805714285715e-06,
+      "loss": 1.0478,
+      "step": 26597
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.956537142857143e-06,
+      "loss": 1.1442,
+      "step": 26598
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.956268571428572e-06,
+      "loss": 1.3312,
+      "step": 26599
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.956000000000001e-06,
+      "loss": 1.5002,
+      "step": 26600
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.955731428571429e-06,
+      "loss": 1.1973,
+      "step": 26601
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.955462857142858e-06,
+      "loss": 1.3883,
+      "step": 26602
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.955194285714286e-06,
+      "loss": 1.0929,
+      "step": 26603
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.954925714285714e-06,
+      "loss": 1.264,
+      "step": 26604
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.954657142857143e-06,
+      "loss": 1.2027,
+      "step": 26605
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9543885714285716e-06,
+      "loss": 1.0993,
+      "step": 26606
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.954120000000001e-06,
+      "loss": 1.1836,
+      "step": 26607
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.953851428571429e-06,
+      "loss": 1.3601,
+      "step": 26608
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.953582857142857e-06,
+      "loss": 1.3954,
+      "step": 26609
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.953314285714286e-06,
+      "loss": 1.1391,
+      "step": 26610
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.953045714285714e-06,
+      "loss": 0.8201,
+      "step": 26611
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.952777142857143e-06,
+      "loss": 1.1827,
+      "step": 26612
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.952508571428572e-06,
+      "loss": 0.7493,
+      "step": 26613
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.95224e-06,
+      "loss": 1.1521,
+      "step": 26614
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.951971428571429e-06,
+      "loss": 0.6201,
+      "step": 26615
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9517028571428576e-06,
+      "loss": 1.1787,
+      "step": 26616
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.951434285714286e-06,
+      "loss": 1.4175,
+      "step": 26617
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.951165714285715e-06,
+      "loss": 1.2122,
+      "step": 26618
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.950897142857143e-06,
+      "loss": 1.022,
+      "step": 26619
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.950628571428572e-06,
+      "loss": 1.1739,
+      "step": 26620
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.95036e-06,
+      "loss": 0.6764,
+      "step": 26621
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.950091428571428e-06,
+      "loss": 1.0189,
+      "step": 26622
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.949822857142857e-06,
+      "loss": 1.0757,
+      "step": 26623
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9495542857142855e-06,
+      "loss": 1.2003,
+      "step": 26624
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9492857142857145e-06,
+      "loss": 1.9297,
+      "step": 26625
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9490171428571436e-06,
+      "loss": 0.9448,
+      "step": 26626
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.948748571428572e-06,
+      "loss": 0.9242,
+      "step": 26627
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.948480000000001e-06,
+      "loss": 1.1763,
+      "step": 26628
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.948211428571429e-06,
+      "loss": 0.7836,
+      "step": 26629
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.947942857142857e-06,
+      "loss": 1.1437,
+      "step": 26630
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.947674285714286e-06,
+      "loss": 0.9923,
+      "step": 26631
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.947405714285714e-06,
+      "loss": 1.1657,
+      "step": 26632
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.947137142857143e-06,
+      "loss": 1.3821,
+      "step": 26633
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9468685714285715e-06,
+      "loss": 1.0567,
+      "step": 26634
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9466e-06,
+      "loss": 1.6153,
+      "step": 26635
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.946331428571429e-06,
+      "loss": 1.0095,
+      "step": 26636
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.946062857142857e-06,
+      "loss": 1.275,
+      "step": 26637
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.945794285714286e-06,
+      "loss": 1.1917,
+      "step": 26638
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.945525714285715e-06,
+      "loss": 1.2193,
+      "step": 26639
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.945257142857143e-06,
+      "loss": 0.6837,
+      "step": 26640
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.944988571428572e-06,
+      "loss": 1.2082,
+      "step": 26641
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.94472e-06,
+      "loss": 1.0679,
+      "step": 26642
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.944451428571429e-06,
+      "loss": 1.4212,
+      "step": 26643
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9441828571428575e-06,
+      "loss": 1.3745,
+      "step": 26644
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.943914285714286e-06,
+      "loss": 1.0584,
+      "step": 26645
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.943645714285715e-06,
+      "loss": 1.2275,
+      "step": 26646
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.943377142857143e-06,
+      "loss": 1.2472,
+      "step": 26647
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.943108571428572e-06,
+      "loss": 0.7565,
+      "step": 26648
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.94284e-06,
+      "loss": 1.0407,
+      "step": 26649
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.942571428571428e-06,
+      "loss": 1.5026,
+      "step": 26650
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.942302857142857e-06,
+      "loss": 1.6541,
+      "step": 26651
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.942034285714286e-06,
+      "loss": 1.3468,
+      "step": 26652
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9417657142857145e-06,
+      "loss": 0.3799,
+      "step": 26653
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9414971428571435e-06,
+      "loss": 0.7418,
+      "step": 26654
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.941228571428572e-06,
+      "loss": 0.738,
+      "step": 26655
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.940960000000001e-06,
+      "loss": 0.9871,
+      "step": 26656
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.940691428571429e-06,
+      "loss": 0.65,
+      "step": 26657
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.940422857142857e-06,
+      "loss": 1.319,
+      "step": 26658
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.940154285714286e-06,
+      "loss": 1.1303,
+      "step": 26659
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.939885714285714e-06,
+      "loss": 1.1902,
+      "step": 26660
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.939617142857143e-06,
+      "loss": 0.8489,
+      "step": 26661
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9393485714285715e-06,
+      "loss": 1.1996,
+      "step": 26662
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.93908e-06,
+      "loss": 1.1498,
+      "step": 26663
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.938811428571429e-06,
+      "loss": 1.1733,
+      "step": 26664
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.938542857142858e-06,
+      "loss": 1.2145,
+      "step": 26665
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.938274285714287e-06,
+      "loss": 0.9943,
+      "step": 26666
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.938005714285715e-06,
+      "loss": 0.7488,
+      "step": 26667
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.937737142857143e-06,
+      "loss": 1.1479,
+      "step": 26668
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.937468571428572e-06,
+      "loss": 1.4353,
+      "step": 26669
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9372e-06,
+      "loss": 1.0028,
+      "step": 26670
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.936931428571429e-06,
+      "loss": 1.0613,
+      "step": 26671
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9366628571428575e-06,
+      "loss": 0.9834,
+      "step": 26672
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.936394285714286e-06,
+      "loss": 1.2714,
+      "step": 26673
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.936125714285715e-06,
+      "loss": 1.3324,
+      "step": 26674
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.935857142857143e-06,
+      "loss": 1.4633,
+      "step": 26675
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.935588571428571e-06,
+      "loss": 0.794,
+      "step": 26676
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.93532e-06,
+      "loss": 1.1953,
+      "step": 26677
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.935051428571428e-06,
+      "loss": 1.6296,
+      "step": 26678
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.934782857142858e-06,
+      "loss": 0.9964,
+      "step": 26679
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.934514285714286e-06,
+      "loss": 1.281,
+      "step": 26680
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9342457142857144e-06,
+      "loss": 0.8527,
+      "step": 26681
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.9339771428571435e-06,
+      "loss": 1.428,
+      "step": 26682
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.933708571428572e-06,
+      "loss": 1.0921,
+      "step": 26683
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.933440000000001e-06,
+      "loss": 1.2185,
+      "step": 26684
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.933171428571429e-06,
+      "loss": 1.0078,
+      "step": 26685
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.932902857142857e-06,
+      "loss": 1.0959,
+      "step": 26686
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 6.932634285714286e-06,
+      "loss": 1.511,
+      "step": 26687
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.932365714285714e-06,
+      "loss": 0.5225,
+      "step": 26688
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.932097142857143e-06,
+      "loss": 1.3025,
+      "step": 26689
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.931828571428571e-06,
+      "loss": 1.586,
+      "step": 26690
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.93156e-06,
+      "loss": 1.1338,
+      "step": 26691
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9312914285714295e-06,
+      "loss": 1.1697,
+      "step": 26692
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.931022857142858e-06,
+      "loss": 1.3707,
+      "step": 26693
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.930754285714287e-06,
+      "loss": 0.7479,
+      "step": 26694
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.930485714285715e-06,
+      "loss": 1.1258,
+      "step": 26695
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.930217142857143e-06,
+      "loss": 1.3622,
+      "step": 26696
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.929948571428572e-06,
+      "loss": 0.8747,
+      "step": 26697
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.92968e-06,
+      "loss": 0.9562,
+      "step": 26698
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.929411428571428e-06,
+      "loss": 1.3095,
+      "step": 26699
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.929142857142857e-06,
+      "loss": 0.9879,
+      "step": 26700
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.928874285714286e-06,
+      "loss": 1.0191,
+      "step": 26701
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.928605714285715e-06,
+      "loss": 0.8164,
+      "step": 26702
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.928337142857143e-06,
+      "loss": 1.3826,
+      "step": 26703
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.928068571428571e-06,
+      "loss": 1.1731,
+      "step": 26704
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.927800000000001e-06,
+      "loss": 1.3791,
+      "step": 26705
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.927531428571429e-06,
+      "loss": 1.4403,
+      "step": 26706
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.927262857142858e-06,
+      "loss": 1.2477,
+      "step": 26707
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.926994285714286e-06,
+      "loss": 1.4689,
+      "step": 26708
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.926725714285714e-06,
+      "loss": 1.0949,
+      "step": 26709
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.926457142857143e-06,
+      "loss": 0.9178,
+      "step": 26710
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9261885714285716e-06,
+      "loss": 1.2081,
+      "step": 26711
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.925920000000001e-06,
+      "loss": 1.0687,
+      "step": 26712
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.925651428571429e-06,
+      "loss": 0.9327,
+      "step": 26713
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.925382857142857e-06,
+      "loss": 0.9418,
+      "step": 26714
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.925114285714286e-06,
+      "loss": 1.3517,
+      "step": 26715
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.924845714285714e-06,
+      "loss": 1.5391,
+      "step": 26716
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.924577142857142e-06,
+      "loss": 0.9018,
+      "step": 26717
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.924308571428572e-06,
+      "loss": 0.8656,
+      "step": 26718
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.92404e-06,
+      "loss": 1.0171,
+      "step": 26719
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.923771428571429e-06,
+      "loss": 1.2631,
+      "step": 26720
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9235028571428576e-06,
+      "loss": 0.722,
+      "step": 26721
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.923234285714286e-06,
+      "loss": 0.9669,
+      "step": 26722
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.922965714285715e-06,
+      "loss": 0.7983,
+      "step": 26723
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.922697142857143e-06,
+      "loss": 1.4029,
+      "step": 26724
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.922428571428572e-06,
+      "loss": 1.0591,
+      "step": 26725
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.92216e-06,
+      "loss": 1.1299,
+      "step": 26726
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.921891428571428e-06,
+      "loss": 0.8472,
+      "step": 26727
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.921622857142857e-06,
+      "loss": 1.4784,
+      "step": 26728
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9213542857142855e-06,
+      "loss": 1.001,
+      "step": 26729
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9210857142857145e-06,
+      "loss": 0.9037,
+      "step": 26730
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.920817142857143e-06,
+      "loss": 1.0286,
+      "step": 26731
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.920548571428572e-06,
+      "loss": 0.9231,
+      "step": 26732
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.920280000000001e-06,
+      "loss": 1.3301,
+      "step": 26733
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.920011428571429e-06,
+      "loss": 0.9386,
+      "step": 26734
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.919742857142858e-06,
+      "loss": 1.211,
+      "step": 26735
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.919474285714286e-06,
+      "loss": 1.0451,
+      "step": 26736
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.919205714285714e-06,
+      "loss": 1.1776,
+      "step": 26737
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.918937142857143e-06,
+      "loss": 1.2841,
+      "step": 26738
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9186685714285715e-06,
+      "loss": 1.2227,
+      "step": 26739
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9184e-06,
+      "loss": 1.2117,
+      "step": 26740
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.918131428571429e-06,
+      "loss": 1.194,
+      "step": 26741
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.917862857142857e-06,
+      "loss": 0.88,
+      "step": 26742
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.917594285714286e-06,
+      "loss": 1.6022,
+      "step": 26743
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.917325714285714e-06,
+      "loss": 0.9158,
+      "step": 26744
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.917057142857143e-06,
+      "loss": 0.8902,
+      "step": 26745
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.916788571428572e-06,
+      "loss": 1.4644,
+      "step": 26746
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.91652e-06,
+      "loss": 1.2621,
+      "step": 26747
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.916251428571429e-06,
+      "loss": 1.1924,
+      "step": 26748
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9159828571428575e-06,
+      "loss": 1.2103,
+      "step": 26749
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.915714285714286e-06,
+      "loss": 0.7166,
+      "step": 26750
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.915445714285715e-06,
+      "loss": 1.3233,
+      "step": 26751
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.915177142857143e-06,
+      "loss": 1.2499,
+      "step": 26752
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.914908571428572e-06,
+      "loss": 1.2888,
+      "step": 26753
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.91464e-06,
+      "loss": 1.1761,
+      "step": 26754
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.914371428571428e-06,
+      "loss": 1.157,
+      "step": 26755
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.914102857142857e-06,
+      "loss": 1.1156,
+      "step": 26756
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9138342857142855e-06,
+      "loss": 1.2775,
+      "step": 26757
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.913565714285715e-06,
+      "loss": 0.9444,
+      "step": 26758
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9132971428571435e-06,
+      "loss": 1.4605,
+      "step": 26759
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.913028571428572e-06,
+      "loss": 1.07,
+      "step": 26760
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.912760000000001e-06,
+      "loss": 1.0986,
+      "step": 26761
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.912491428571429e-06,
+      "loss": 0.6489,
+      "step": 26762
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.912222857142857e-06,
+      "loss": 1.2394,
+      "step": 26763
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.911954285714286e-06,
+      "loss": 1.1823,
+      "step": 26764
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.911685714285714e-06,
+      "loss": 1.3915,
+      "step": 26765
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.911417142857143e-06,
+      "loss": 1.1137,
+      "step": 26766
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9111485714285715e-06,
+      "loss": 1.3715,
+      "step": 26767
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.91088e-06,
+      "loss": 1.2639,
+      "step": 26768
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.910611428571429e-06,
+      "loss": 1.6817,
+      "step": 26769
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.910342857142857e-06,
+      "loss": 1.3627,
+      "step": 26770
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.910074285714287e-06,
+      "loss": 1.1613,
+      "step": 26771
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.909805714285715e-06,
+      "loss": 1.1839,
+      "step": 26772
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.909537142857143e-06,
+      "loss": 1.3624,
+      "step": 26773
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.909268571428572e-06,
+      "loss": 1.1322,
+      "step": 26774
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.909e-06,
+      "loss": 1.6531,
+      "step": 26775
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.908731428571429e-06,
+      "loss": 1.2248,
+      "step": 26776
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9084628571428575e-06,
+      "loss": 1.1001,
+      "step": 26777
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.908194285714286e-06,
+      "loss": 1.2316,
+      "step": 26778
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.907925714285715e-06,
+      "loss": 1.3315,
+      "step": 26779
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.907657142857143e-06,
+      "loss": 1.0338,
+      "step": 26780
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.907388571428572e-06,
+      "loss": 1.0223,
+      "step": 26781
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.90712e-06,
+      "loss": 1.8286,
+      "step": 26782
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.906851428571428e-06,
+      "loss": 1.0514,
+      "step": 26783
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.906582857142857e-06,
+      "loss": 0.8242,
+      "step": 26784
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.906314285714286e-06,
+      "loss": 0.7472,
+      "step": 26785
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9060457142857144e-06,
+      "loss": 1.1036,
+      "step": 26786
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.9057771428571435e-06,
+      "loss": 1.1506,
+      "step": 26787
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.905508571428572e-06,
+      "loss": 1.7245,
+      "step": 26788
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.905240000000001e-06,
+      "loss": 1.2159,
+      "step": 26789
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.904971428571429e-06,
+      "loss": 1.021,
+      "step": 26790
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.904702857142857e-06,
+      "loss": 0.8371,
+      "step": 26791
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.904434285714286e-06,
+      "loss": 1.3378,
+      "step": 26792
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.904165714285714e-06,
+      "loss": 1.0028,
+      "step": 26793
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.903897142857143e-06,
+      "loss": 1.3377,
+      "step": 26794
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.903628571428571e-06,
+      "loss": 1.2977,
+      "step": 26795
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.90336e-06,
+      "loss": 1.1847,
+      "step": 26796
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.903091428571429e-06,
+      "loss": 1.1286,
+      "step": 26797
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.902822857142858e-06,
+      "loss": 0.8854,
+      "step": 26798
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.902554285714287e-06,
+      "loss": 1.3102,
+      "step": 26799
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.902285714285715e-06,
+      "loss": 0.9718,
+      "step": 26800
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.902017142857143e-06,
+      "loss": 1.098,
+      "step": 26801
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.901748571428572e-06,
+      "loss": 1.2305,
+      "step": 26802
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.90148e-06,
+      "loss": 1.4789,
+      "step": 26803
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.901211428571428e-06,
+      "loss": 0.6516,
+      "step": 26804
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.900942857142857e-06,
+      "loss": 1.1978,
+      "step": 26805
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.900674285714286e-06,
+      "loss": 1.8472,
+      "step": 26806
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.900405714285715e-06,
+      "loss": 1.3149,
+      "step": 26807
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.900137142857143e-06,
+      "loss": 1.3782,
+      "step": 26808
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.899868571428571e-06,
+      "loss": 1.3291,
+      "step": 26809
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8996e-06,
+      "loss": 1.1813,
+      "step": 26810
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.899331428571429e-06,
+      "loss": 0.9784,
+      "step": 26811
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.899062857142858e-06,
+      "loss": 1.1256,
+      "step": 26812
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.898794285714286e-06,
+      "loss": 1.4598,
+      "step": 26813
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.898525714285714e-06,
+      "loss": 1.073,
+      "step": 26814
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.898257142857143e-06,
+      "loss": 1.3533,
+      "step": 26815
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8979885714285716e-06,
+      "loss": 1.0983,
+      "step": 26816
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.897720000000001e-06,
+      "loss": 1.1922,
+      "step": 26817
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.897451428571429e-06,
+      "loss": 1.1026,
+      "step": 26818
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.897182857142857e-06,
+      "loss": 0.9053,
+      "step": 26819
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.896914285714286e-06,
+      "loss": 1.1088,
+      "step": 26820
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.896645714285714e-06,
+      "loss": 1.3311,
+      "step": 26821
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.896377142857143e-06,
+      "loss": 0.9966,
+      "step": 26822
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.896108571428571e-06,
+      "loss": 1.2544,
+      "step": 26823
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.89584e-06,
+      "loss": 1.0211,
+      "step": 26824
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.895571428571429e-06,
+      "loss": 1.2585,
+      "step": 26825
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8953028571428576e-06,
+      "loss": 1.0857,
+      "step": 26826
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.895034285714286e-06,
+      "loss": 1.1161,
+      "step": 26827
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.894765714285715e-06,
+      "loss": 1.5168,
+      "step": 26828
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.894497142857143e-06,
+      "loss": 1.2414,
+      "step": 26829
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.894228571428572e-06,
+      "loss": 1.2368,
+      "step": 26830
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.89396e-06,
+      "loss": 0.9997,
+      "step": 26831
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.893691428571428e-06,
+      "loss": 1.0582,
+      "step": 26832
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.893422857142857e-06,
+      "loss": 0.9503,
+      "step": 26833
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8931542857142855e-06,
+      "loss": 0.9019,
+      "step": 26834
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8928857142857145e-06,
+      "loss": 1.194,
+      "step": 26835
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.892617142857143e-06,
+      "loss": 0.9432,
+      "step": 26836
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.892348571428572e-06,
+      "loss": 1.3086,
+      "step": 26837
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.892080000000001e-06,
+      "loss": 1.1696,
+      "step": 26838
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.891811428571429e-06,
+      "loss": 1.311,
+      "step": 26839
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.891542857142858e-06,
+      "loss": 1.1711,
+      "step": 26840
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.891274285714286e-06,
+      "loss": 1.0981,
+      "step": 26841
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.891005714285714e-06,
+      "loss": 1.0577,
+      "step": 26842
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.890737142857143e-06,
+      "loss": 1.1805,
+      "step": 26843
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8904685714285715e-06,
+      "loss": 1.1561,
+      "step": 26844
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8902000000000005e-06,
+      "loss": 1.0051,
+      "step": 26845
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.889931428571429e-06,
+      "loss": 1.1462,
+      "step": 26846
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.889662857142857e-06,
+      "loss": 1.0672,
+      "step": 26847
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.889394285714286e-06,
+      "loss": 1.539,
+      "step": 26848
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.889125714285714e-06,
+      "loss": 1.1859,
+      "step": 26849
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.888857142857143e-06,
+      "loss": 1.2749,
+      "step": 26850
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.888588571428572e-06,
+      "loss": 1.0015,
+      "step": 26851
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.88832e-06,
+      "loss": 1.1958,
+      "step": 26852
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.888051428571429e-06,
+      "loss": 0.909,
+      "step": 26853
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8877828571428575e-06,
+      "loss": 1.3798,
+      "step": 26854
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.887514285714286e-06,
+      "loss": 1.3101,
+      "step": 26855
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.887245714285715e-06,
+      "loss": 1.1411,
+      "step": 26856
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.886977142857143e-06,
+      "loss": 1.1422,
+      "step": 26857
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.886708571428572e-06,
+      "loss": 1.1474,
+      "step": 26858
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.88644e-06,
+      "loss": 1.1725,
+      "step": 26859
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.886171428571428e-06,
+      "loss": 1.4957,
+      "step": 26860
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.885902857142857e-06,
+      "loss": 1.2396,
+      "step": 26861
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 6.8856342857142855e-06,
+      "loss": 1.597,
+      "step": 26862
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8853657142857145e-06,
+      "loss": 1.0933,
+      "step": 26863
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8850971428571435e-06,
+      "loss": 0.8807,
+      "step": 26864
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.884828571428572e-06,
+      "loss": 0.8522,
+      "step": 26865
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.884560000000001e-06,
+      "loss": 1.2046,
+      "step": 26866
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.884291428571429e-06,
+      "loss": 0.7768,
+      "step": 26867
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.884022857142858e-06,
+      "loss": 0.7513,
+      "step": 26868
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.883754285714286e-06,
+      "loss": 1.23,
+      "step": 26869
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.883485714285714e-06,
+      "loss": 1.1612,
+      "step": 26870
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.883217142857143e-06,
+      "loss": 1.2873,
+      "step": 26871
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8829485714285715e-06,
+      "loss": 1.184,
+      "step": 26872
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.88268e-06,
+      "loss": 0.946,
+      "step": 26873
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.882411428571429e-06,
+      "loss": 1.0642,
+      "step": 26874
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.882142857142857e-06,
+      "loss": 0.674,
+      "step": 26875
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.881874285714286e-06,
+      "loss": 1.3485,
+      "step": 26876
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.881605714285715e-06,
+      "loss": 1.1764,
+      "step": 26877
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.881337142857143e-06,
+      "loss": 1.4331,
+      "step": 26878
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.881068571428572e-06,
+      "loss": 1.0746,
+      "step": 26879
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8808e-06,
+      "loss": 0.783,
+      "step": 26880
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.880531428571429e-06,
+      "loss": 0.9145,
+      "step": 26881
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8802628571428575e-06,
+      "loss": 0.9036,
+      "step": 26882
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.879994285714286e-06,
+      "loss": 0.9248,
+      "step": 26883
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.879725714285715e-06,
+      "loss": 0.8048,
+      "step": 26884
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.879457142857143e-06,
+      "loss": 1.1614,
+      "step": 26885
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.879188571428572e-06,
+      "loss": 1.3899,
+      "step": 26886
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.87892e-06,
+      "loss": 1.2597,
+      "step": 26887
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.878651428571428e-06,
+      "loss": 0.9303,
+      "step": 26888
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.878382857142857e-06,
+      "loss": 1.2652,
+      "step": 26889
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.878114285714286e-06,
+      "loss": 1.2456,
+      "step": 26890
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8778457142857144e-06,
+      "loss": 1.1419,
+      "step": 26891
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8775771428571435e-06,
+      "loss": 1.5375,
+      "step": 26892
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.877308571428572e-06,
+      "loss": 1.215,
+      "step": 26893
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.877040000000001e-06,
+      "loss": 1.2565,
+      "step": 26894
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.876771428571429e-06,
+      "loss": 0.9797,
+      "step": 26895
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.876502857142857e-06,
+      "loss": 0.9487,
+      "step": 26896
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.876234285714286e-06,
+      "loss": 1.105,
+      "step": 26897
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.875965714285714e-06,
+      "loss": 1.0183,
+      "step": 26898
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.875697142857143e-06,
+      "loss": 1.2641,
+      "step": 26899
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.875428571428571e-06,
+      "loss": 0.9006,
+      "step": 26900
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.87516e-06,
+      "loss": 0.8305,
+      "step": 26901
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.874891428571429e-06,
+      "loss": 1.5073,
+      "step": 26902
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.874622857142858e-06,
+      "loss": 1.2579,
+      "step": 26903
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.874354285714287e-06,
+      "loss": 0.5795,
+      "step": 26904
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.874085714285715e-06,
+      "loss": 1.094,
+      "step": 26905
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.873817142857143e-06,
+      "loss": 0.919,
+      "step": 26906
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.873548571428572e-06,
+      "loss": 1.2465,
+      "step": 26907
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.87328e-06,
+      "loss": 0.9527,
+      "step": 26908
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.873011428571429e-06,
+      "loss": 1.0449,
+      "step": 26909
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.872742857142857e-06,
+      "loss": 1.0222,
+      "step": 26910
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8724742857142856e-06,
+      "loss": 1.023,
+      "step": 26911
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.872205714285715e-06,
+      "loss": 1.3032,
+      "step": 26912
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.871937142857143e-06,
+      "loss": 1.1176,
+      "step": 26913
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.871668571428571e-06,
+      "loss": 0.8011,
+      "step": 26914
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8714e-06,
+      "loss": 1.0833,
+      "step": 26915
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.871131428571429e-06,
+      "loss": 1.0341,
+      "step": 26916
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.870862857142858e-06,
+      "loss": 0.8909,
+      "step": 26917
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.870594285714286e-06,
+      "loss": 1.4243,
+      "step": 26918
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.870325714285714e-06,
+      "loss": 1.1603,
+      "step": 26919
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.870057142857143e-06,
+      "loss": 1.0565,
+      "step": 26920
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8697885714285716e-06,
+      "loss": 1.1975,
+      "step": 26921
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.869520000000001e-06,
+      "loss": 1.4694,
+      "step": 26922
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.869251428571429e-06,
+      "loss": 1.3772,
+      "step": 26923
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.868982857142857e-06,
+      "loss": 1.2968,
+      "step": 26924
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.868714285714286e-06,
+      "loss": 1.1256,
+      "step": 26925
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.868445714285714e-06,
+      "loss": 1.5565,
+      "step": 26926
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.868177142857143e-06,
+      "loss": 1.0872,
+      "step": 26927
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.867908571428571e-06,
+      "loss": 1.0561,
+      "step": 26928
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.86764e-06,
+      "loss": 0.8932,
+      "step": 26929
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.867371428571429e-06,
+      "loss": 1.4698,
+      "step": 26930
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8671028571428576e-06,
+      "loss": 1.327,
+      "step": 26931
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.866834285714287e-06,
+      "loss": 1.1962,
+      "step": 26932
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.866565714285715e-06,
+      "loss": 0.8538,
+      "step": 26933
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.866297142857143e-06,
+      "loss": 1.2536,
+      "step": 26934
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.866028571428572e-06,
+      "loss": 0.8605,
+      "step": 26935
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.86576e-06,
+      "loss": 1.219,
+      "step": 26936
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.865491428571428e-06,
+      "loss": 1.315,
+      "step": 26937
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.865222857142857e-06,
+      "loss": 1.176,
+      "step": 26938
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8649542857142855e-06,
+      "loss": 1.4755,
+      "step": 26939
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8646857142857145e-06,
+      "loss": 1.4988,
+      "step": 26940
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.864417142857143e-06,
+      "loss": 0.9957,
+      "step": 26941
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.864148571428572e-06,
+      "loss": 0.7038,
+      "step": 26942
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.863880000000001e-06,
+      "loss": 1.1929,
+      "step": 26943
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.863611428571429e-06,
+      "loss": 1.4096,
+      "step": 26944
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.863342857142858e-06,
+      "loss": 1.5018,
+      "step": 26945
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.863074285714286e-06,
+      "loss": 1.0564,
+      "step": 26946
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.862805714285714e-06,
+      "loss": 1.1771,
+      "step": 26947
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.862537142857143e-06,
+      "loss": 0.8459,
+      "step": 26948
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8622685714285715e-06,
+      "loss": 1.0193,
+      "step": 26949
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8620000000000005e-06,
+      "loss": 0.9135,
+      "step": 26950
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.861731428571429e-06,
+      "loss": 0.6642,
+      "step": 26951
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.861462857142857e-06,
+      "loss": 1.2033,
+      "step": 26952
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.861194285714286e-06,
+      "loss": 1.3063,
+      "step": 26953
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.860925714285714e-06,
+      "loss": 1.3298,
+      "step": 26954
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.860657142857143e-06,
+      "loss": 1.1204,
+      "step": 26955
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.860388571428572e-06,
+      "loss": 1.1865,
+      "step": 26956
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.86012e-06,
+      "loss": 1.1494,
+      "step": 26957
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.859851428571429e-06,
+      "loss": 1.0616,
+      "step": 26958
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8595828571428575e-06,
+      "loss": 0.9982,
+      "step": 26959
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.859314285714286e-06,
+      "loss": 0.9166,
+      "step": 26960
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.859045714285715e-06,
+      "loss": 1.5126,
+      "step": 26961
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.858777142857143e-06,
+      "loss": 1.0343,
+      "step": 26962
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.858508571428572e-06,
+      "loss": 1.5473,
+      "step": 26963
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.85824e-06,
+      "loss": 1.608,
+      "step": 26964
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.857971428571428e-06,
+      "loss": 1.1901,
+      "step": 26965
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.857702857142857e-06,
+      "loss": 1.2688,
+      "step": 26966
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8574342857142855e-06,
+      "loss": 1.1723,
+      "step": 26967
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8571657142857145e-06,
+      "loss": 1.149,
+      "step": 26968
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8568971428571435e-06,
+      "loss": 1.2511,
+      "step": 26969
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.856628571428572e-06,
+      "loss": 1.112,
+      "step": 26970
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.856360000000001e-06,
+      "loss": 1.0391,
+      "step": 26971
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.856091428571429e-06,
+      "loss": 1.2919,
+      "step": 26972
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.855822857142858e-06,
+      "loss": 1.1237,
+      "step": 26973
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.855554285714286e-06,
+      "loss": 1.2855,
+      "step": 26974
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.855285714285714e-06,
+      "loss": 0.9363,
+      "step": 26975
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.855017142857143e-06,
+      "loss": 0.7798,
+      "step": 26976
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8547485714285715e-06,
+      "loss": 1.1644,
+      "step": 26977
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.85448e-06,
+      "loss": 1.0407,
+      "step": 26978
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.854211428571429e-06,
+      "loss": 1.6628,
+      "step": 26979
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.853942857142857e-06,
+      "loss": 1.0321,
+      "step": 26980
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.853674285714286e-06,
+      "loss": 1.4208,
+      "step": 26981
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.853405714285715e-06,
+      "loss": 1.2436,
+      "step": 26982
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.853137142857143e-06,
+      "loss": 0.6221,
+      "step": 26983
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.852868571428572e-06,
+      "loss": 1.0046,
+      "step": 26984
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8526e-06,
+      "loss": 1.0206,
+      "step": 26985
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.852331428571429e-06,
+      "loss": 0.748,
+      "step": 26986
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8520628571428575e-06,
+      "loss": 0.9154,
+      "step": 26987
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.851794285714286e-06,
+      "loss": 1.2671,
+      "step": 26988
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.851525714285715e-06,
+      "loss": 1.1325,
+      "step": 26989
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.851257142857143e-06,
+      "loss": 0.8923,
+      "step": 26990
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.850988571428572e-06,
+      "loss": 1.1115,
+      "step": 26991
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.85072e-06,
+      "loss": 1.2329,
+      "step": 26992
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.850451428571428e-06,
+      "loss": 1.035,
+      "step": 26993
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.850182857142857e-06,
+      "loss": 1.0796,
+      "step": 26994
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.849914285714286e-06,
+      "loss": 0.8249,
+      "step": 26995
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.849645714285715e-06,
+      "loss": 1.0369,
+      "step": 26996
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8493771428571435e-06,
+      "loss": 1.2452,
+      "step": 26997
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.849108571428572e-06,
+      "loss": 1.1788,
+      "step": 26998
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.848840000000001e-06,
+      "loss": 1.4388,
+      "step": 26999
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.848571428571429e-06,
+      "loss": 1.4212,
+      "step": 27000
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.848302857142857e-06,
+      "loss": 1.0208,
+      "step": 27001
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.848034285714286e-06,
+      "loss": 1.1038,
+      "step": 27002
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.847765714285714e-06,
+      "loss": 1.2299,
+      "step": 27003
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.847497142857143e-06,
+      "loss": 0.7191,
+      "step": 27004
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.847228571428571e-06,
+      "loss": 1.1166,
+      "step": 27005
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.84696e-06,
+      "loss": 1.4921,
+      "step": 27006
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.846691428571429e-06,
+      "loss": 1.4511,
+      "step": 27007
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.846422857142858e-06,
+      "loss": 1.6105,
+      "step": 27008
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.846154285714287e-06,
+      "loss": 1.2213,
+      "step": 27009
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.845885714285715e-06,
+      "loss": 1.2212,
+      "step": 27010
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.845617142857143e-06,
+      "loss": 0.9435,
+      "step": 27011
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.845348571428572e-06,
+      "loss": 0.8455,
+      "step": 27012
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.84508e-06,
+      "loss": 0.9097,
+      "step": 27013
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.844811428571429e-06,
+      "loss": 1.5894,
+      "step": 27014
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.844542857142857e-06,
+      "loss": 1.0956,
+      "step": 27015
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8442742857142856e-06,
+      "loss": 1.1675,
+      "step": 27016
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.844005714285715e-06,
+      "loss": 1.0581,
+      "step": 27017
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.843737142857143e-06,
+      "loss": 1.174,
+      "step": 27018
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.843468571428572e-06,
+      "loss": 1.0396,
+      "step": 27019
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8432e-06,
+      "loss": 1.0194,
+      "step": 27020
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.842931428571429e-06,
+      "loss": 0.7117,
+      "step": 27021
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.842662857142858e-06,
+      "loss": 1.1909,
+      "step": 27022
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.842394285714286e-06,
+      "loss": 1.4203,
+      "step": 27023
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.842125714285714e-06,
+      "loss": 1.2938,
+      "step": 27024
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.841857142857143e-06,
+      "loss": 1.3441,
+      "step": 27025
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8415885714285716e-06,
+      "loss": 0.9248,
+      "step": 27026
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.841320000000001e-06,
+      "loss": 1.3515,
+      "step": 27027
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.841051428571429e-06,
+      "loss": 1.3061,
+      "step": 27028
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.840782857142857e-06,
+      "loss": 1.1294,
+      "step": 27029
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.840514285714286e-06,
+      "loss": 1.6386,
+      "step": 27030
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.840245714285714e-06,
+      "loss": 0.6512,
+      "step": 27031
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.839977142857143e-06,
+      "loss": 1.3351,
+      "step": 27032
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.839708571428571e-06,
+      "loss": 1.6028,
+      "step": 27033
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.83944e-06,
+      "loss": 1.4117,
+      "step": 27034
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.839171428571429e-06,
+      "loss": 0.9186,
+      "step": 27035
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.8389028571428576e-06,
+      "loss": 1.4566,
+      "step": 27036
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.838634285714287e-06,
+      "loss": 0.8761,
+      "step": 27037
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.838365714285715e-06,
+      "loss": 1.0952,
+      "step": 27038
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.838097142857143e-06,
+      "loss": 1.0596,
+      "step": 27039
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.837828571428572e-06,
+      "loss": 0.8312,
+      "step": 27040
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.83756e-06,
+      "loss": 0.8453,
+      "step": 27041
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.837291428571428e-06,
+      "loss": 0.8447,
+      "step": 27042
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.837022857142857e-06,
+      "loss": 1.0915,
+      "step": 27043
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8367542857142855e-06,
+      "loss": 1.0827,
+      "step": 27044
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8364857142857145e-06,
+      "loss": 1.4677,
+      "step": 27045
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.836217142857143e-06,
+      "loss": 1.0616,
+      "step": 27046
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.835948571428572e-06,
+      "loss": 1.3055,
+      "step": 27047
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.835680000000001e-06,
+      "loss": 1.0888,
+      "step": 27048
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.835411428571429e-06,
+      "loss": 1.5239,
+      "step": 27049
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.835142857142858e-06,
+      "loss": 1.3314,
+      "step": 27050
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.834874285714286e-06,
+      "loss": 1.6153,
+      "step": 27051
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.834605714285714e-06,
+      "loss": 0.8189,
+      "step": 27052
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.834337142857143e-06,
+      "loss": 0.9091,
+      "step": 27053
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8340685714285715e-06,
+      "loss": 1.3039,
+      "step": 27054
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8338000000000005e-06,
+      "loss": 0.8018,
+      "step": 27055
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.833531428571429e-06,
+      "loss": 1.0629,
+      "step": 27056
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.833262857142857e-06,
+      "loss": 1.5093,
+      "step": 27057
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.832994285714286e-06,
+      "loss": 1.2799,
+      "step": 27058
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.832725714285714e-06,
+      "loss": 0.9345,
+      "step": 27059
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.832457142857143e-06,
+      "loss": 0.9172,
+      "step": 27060
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.832188571428572e-06,
+      "loss": 1.2352,
+      "step": 27061
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.83192e-06,
+      "loss": 1.0506,
+      "step": 27062
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.831651428571429e-06,
+      "loss": 1.1858,
+      "step": 27063
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8313828571428575e-06,
+      "loss": 0.9238,
+      "step": 27064
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.831114285714286e-06,
+      "loss": 1.188,
+      "step": 27065
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.830845714285715e-06,
+      "loss": 1.1394,
+      "step": 27066
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.830577142857143e-06,
+      "loss": 0.9091,
+      "step": 27067
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.830308571428572e-06,
+      "loss": 1.5031,
+      "step": 27068
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.83004e-06,
+      "loss": 0.9991,
+      "step": 27069
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.829771428571428e-06,
+      "loss": 0.7375,
+      "step": 27070
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.829502857142857e-06,
+      "loss": 1.1403,
+      "step": 27071
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8292342857142855e-06,
+      "loss": 1.1581,
+      "step": 27072
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8289657142857145e-06,
+      "loss": 0.8021,
+      "step": 27073
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8286971428571435e-06,
+      "loss": 1.2279,
+      "step": 27074
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.828428571428572e-06,
+      "loss": 1.1779,
+      "step": 27075
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.828160000000001e-06,
+      "loss": 1.0118,
+      "step": 27076
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.827891428571429e-06,
+      "loss": 1.0197,
+      "step": 27077
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.827622857142858e-06,
+      "loss": 1.0156,
+      "step": 27078
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.827354285714286e-06,
+      "loss": 1.3048,
+      "step": 27079
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.827085714285714e-06,
+      "loss": 1.3922,
+      "step": 27080
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.826817142857143e-06,
+      "loss": 1.2664,
+      "step": 27081
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8265485714285715e-06,
+      "loss": 0.9253,
+      "step": 27082
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8262800000000005e-06,
+      "loss": 0.9813,
+      "step": 27083
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.826011428571429e-06,
+      "loss": 0.8065,
+      "step": 27084
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.825742857142857e-06,
+      "loss": 1.1916,
+      "step": 27085
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.825474285714286e-06,
+      "loss": 1.4877,
+      "step": 27086
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.825205714285715e-06,
+      "loss": 0.945,
+      "step": 27087
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.824937142857143e-06,
+      "loss": 1.0584,
+      "step": 27088
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.824668571428572e-06,
+      "loss": 1.2564,
+      "step": 27089
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8244e-06,
+      "loss": 1.3567,
+      "step": 27090
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.824131428571429e-06,
+      "loss": 1.0215,
+      "step": 27091
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8238628571428575e-06,
+      "loss": 0.9937,
+      "step": 27092
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.823594285714286e-06,
+      "loss": 0.8829,
+      "step": 27093
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.823325714285715e-06,
+      "loss": 1.3506,
+      "step": 27094
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.823057142857143e-06,
+      "loss": 0.9362,
+      "step": 27095
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.822788571428572e-06,
+      "loss": 1.2251,
+      "step": 27096
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.82252e-06,
+      "loss": 1.0946,
+      "step": 27097
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.822251428571428e-06,
+      "loss": 1.2494,
+      "step": 27098
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.821982857142857e-06,
+      "loss": 0.95,
+      "step": 27099
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.821714285714286e-06,
+      "loss": 1.5278,
+      "step": 27100
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.821445714285715e-06,
+      "loss": 1.262,
+      "step": 27101
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8211771428571435e-06,
+      "loss": 0.6469,
+      "step": 27102
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.820908571428572e-06,
+      "loss": 1.1024,
+      "step": 27103
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.820640000000001e-06,
+      "loss": 1.0914,
+      "step": 27104
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.820371428571429e-06,
+      "loss": 1.3389,
+      "step": 27105
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.820102857142858e-06,
+      "loss": 1.0976,
+      "step": 27106
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.819834285714286e-06,
+      "loss": 0.964,
+      "step": 27107
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.819565714285714e-06,
+      "loss": 1.5164,
+      "step": 27108
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.819297142857143e-06,
+      "loss": 1.2477,
+      "step": 27109
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.819028571428571e-06,
+      "loss": 0.8465,
+      "step": 27110
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.81876e-06,
+      "loss": 1.3543,
+      "step": 27111
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.818491428571429e-06,
+      "loss": 1.061,
+      "step": 27112
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.818222857142858e-06,
+      "loss": 1.4096,
+      "step": 27113
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.817954285714287e-06,
+      "loss": 1.165,
+      "step": 27114
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.817685714285715e-06,
+      "loss": 1.425,
+      "step": 27115
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.817417142857143e-06,
+      "loss": 0.9605,
+      "step": 27116
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.817148571428572e-06,
+      "loss": 0.9574,
+      "step": 27117
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.81688e-06,
+      "loss": 1.2269,
+      "step": 27118
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.816611428571429e-06,
+      "loss": 1.0502,
+      "step": 27119
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.816342857142857e-06,
+      "loss": 1.151,
+      "step": 27120
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8160742857142856e-06,
+      "loss": 1.1468,
+      "step": 27121
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.815805714285715e-06,
+      "loss": 1.0412,
+      "step": 27122
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.815537142857143e-06,
+      "loss": 1.2755,
+      "step": 27123
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.815268571428572e-06,
+      "loss": 1.031,
+      "step": 27124
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.815e-06,
+      "loss": 1.3902,
+      "step": 27125
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.814731428571429e-06,
+      "loss": 1.3418,
+      "step": 27126
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.814462857142858e-06,
+      "loss": 1.0992,
+      "step": 27127
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.814194285714286e-06,
+      "loss": 1.0895,
+      "step": 27128
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.813925714285714e-06,
+      "loss": 1.144,
+      "step": 27129
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.813657142857143e-06,
+      "loss": 1.3779,
+      "step": 27130
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8133885714285716e-06,
+      "loss": 1.2449,
+      "step": 27131
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.813120000000001e-06,
+      "loss": 1.2933,
+      "step": 27132
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.812851428571429e-06,
+      "loss": 0.6605,
+      "step": 27133
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.812582857142857e-06,
+      "loss": 0.8869,
+      "step": 27134
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.812314285714286e-06,
+      "loss": 0.9532,
+      "step": 27135
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.812045714285714e-06,
+      "loss": 1.0319,
+      "step": 27136
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.811777142857143e-06,
+      "loss": 1.0552,
+      "step": 27137
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.811508571428571e-06,
+      "loss": 1.0027,
+      "step": 27138
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8112399999999995e-06,
+      "loss": 1.3427,
+      "step": 27139
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.810971428571429e-06,
+      "loss": 0.9154,
+      "step": 27140
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8107028571428576e-06,
+      "loss": 0.911,
+      "step": 27141
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.810434285714287e-06,
+      "loss": 1.2481,
+      "step": 27142
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.810165714285715e-06,
+      "loss": 1.1932,
+      "step": 27143
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.809897142857143e-06,
+      "loss": 1.0534,
+      "step": 27144
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.809628571428572e-06,
+      "loss": 1.1084,
+      "step": 27145
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.80936e-06,
+      "loss": 1.4925,
+      "step": 27146
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.809091428571429e-06,
+      "loss": 1.418,
+      "step": 27147
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.808822857142857e-06,
+      "loss": 1.0806,
+      "step": 27148
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8085542857142855e-06,
+      "loss": 1.1768,
+      "step": 27149
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8082857142857145e-06,
+      "loss": 0.64,
+      "step": 27150
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.808017142857143e-06,
+      "loss": 1.2484,
+      "step": 27151
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.807748571428571e-06,
+      "loss": 1.1289,
+      "step": 27152
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.807480000000001e-06,
+      "loss": 0.8005,
+      "step": 27153
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.807211428571429e-06,
+      "loss": 0.8784,
+      "step": 27154
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.806942857142858e-06,
+      "loss": 1.2466,
+      "step": 27155
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.806674285714286e-06,
+      "loss": 1.2874,
+      "step": 27156
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.806405714285714e-06,
+      "loss": 0.8377,
+      "step": 27157
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.806137142857143e-06,
+      "loss": 1.5191,
+      "step": 27158
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8058685714285715e-06,
+      "loss": 1.5278,
+      "step": 27159
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8056000000000005e-06,
+      "loss": 0.8204,
+      "step": 27160
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.805331428571429e-06,
+      "loss": 1.1903,
+      "step": 27161
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.805062857142857e-06,
+      "loss": 1.0797,
+      "step": 27162
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.804794285714286e-06,
+      "loss": 0.8956,
+      "step": 27163
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.804525714285714e-06,
+      "loss": 0.9751,
+      "step": 27164
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.804257142857143e-06,
+      "loss": 1.2557,
+      "step": 27165
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.803988571428572e-06,
+      "loss": 1.0917,
+      "step": 27166
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.80372e-06,
+      "loss": 1.3661,
+      "step": 27167
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.803451428571429e-06,
+      "loss": 0.7292,
+      "step": 27168
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8031828571428575e-06,
+      "loss": 1.0393,
+      "step": 27169
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8029142857142865e-06,
+      "loss": 0.8239,
+      "step": 27170
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.802645714285715e-06,
+      "loss": 1.4843,
+      "step": 27171
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.802377142857143e-06,
+      "loss": 1.1743,
+      "step": 27172
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.802108571428572e-06,
+      "loss": 1.1513,
+      "step": 27173
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.80184e-06,
+      "loss": 1.0621,
+      "step": 27174
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.801571428571428e-06,
+      "loss": 1.3295,
+      "step": 27175
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.801302857142857e-06,
+      "loss": 1.4179,
+      "step": 27176
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8010342857142855e-06,
+      "loss": 1.8796,
+      "step": 27177
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8007657142857145e-06,
+      "loss": 1.3365,
+      "step": 27178
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.8004971428571435e-06,
+      "loss": 0.9452,
+      "step": 27179
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.800228571428572e-06,
+      "loss": 1.0112,
+      "step": 27180
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.799960000000001e-06,
+      "loss": 1.2018,
+      "step": 27181
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.799691428571429e-06,
+      "loss": 0.993,
+      "step": 27182
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.799422857142858e-06,
+      "loss": 1.2017,
+      "step": 27183
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.799154285714286e-06,
+      "loss": 1.1691,
+      "step": 27184
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.798885714285714e-06,
+      "loss": 1.0865,
+      "step": 27185
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.798617142857143e-06,
+      "loss": 0.9409,
+      "step": 27186
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.7983485714285715e-06,
+      "loss": 1.4717,
+      "step": 27187
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.7980800000000005e-06,
+      "loss": 1.1419,
+      "step": 27188
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.797811428571429e-06,
+      "loss": 1.3387,
+      "step": 27189
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.797542857142857e-06,
+      "loss": 1.2419,
+      "step": 27190
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.797274285714286e-06,
+      "loss": 1.0506,
+      "step": 27191
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.797005714285714e-06,
+      "loss": 1.0322,
+      "step": 27192
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.796737142857144e-06,
+      "loss": 1.0543,
+      "step": 27193
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.796468571428572e-06,
+      "loss": 1.371,
+      "step": 27194
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.7962e-06,
+      "loss": 0.7933,
+      "step": 27195
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.795931428571429e-06,
+      "loss": 1.0246,
+      "step": 27196
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.7956628571428575e-06,
+      "loss": 1.2081,
+      "step": 27197
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.795394285714286e-06,
+      "loss": 1.4478,
+      "step": 27198
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.795125714285715e-06,
+      "loss": 0.8661,
+      "step": 27199
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.794857142857143e-06,
+      "loss": 1.1394,
+      "step": 27200
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.794588571428572e-06,
+      "loss": 1.1521,
+      "step": 27201
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.79432e-06,
+      "loss": 1.3256,
+      "step": 27202
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.794051428571428e-06,
+      "loss": 1.2043,
+      "step": 27203
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.793782857142857e-06,
+      "loss": 1.3604,
+      "step": 27204
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.793514285714285e-06,
+      "loss": 1.167,
+      "step": 27205
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.793245714285715e-06,
+      "loss": 1.3145,
+      "step": 27206
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.7929771428571434e-06,
+      "loss": 1.5444,
+      "step": 27207
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.792708571428572e-06,
+      "loss": 1.4627,
+      "step": 27208
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.792440000000001e-06,
+      "loss": 0.7905,
+      "step": 27209
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.792171428571429e-06,
+      "loss": 1.0408,
+      "step": 27210
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.791902857142858e-06,
+      "loss": 1.4829,
+      "step": 27211
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.791634285714286e-06,
+      "loss": 1.2264,
+      "step": 27212
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.791365714285714e-06,
+      "loss": 1.1567,
+      "step": 27213
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.791097142857143e-06,
+      "loss": 1.2379,
+      "step": 27214
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.790828571428571e-06,
+      "loss": 1.1335,
+      "step": 27215
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7905599999999996e-06,
+      "loss": 1.1111,
+      "step": 27216
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.790291428571429e-06,
+      "loss": 1.2239,
+      "step": 27217
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.790022857142857e-06,
+      "loss": 1.0849,
+      "step": 27218
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.789754285714287e-06,
+      "loss": 0.9291,
+      "step": 27219
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.789485714285715e-06,
+      "loss": 1.5634,
+      "step": 27220
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.789217142857143e-06,
+      "loss": 1.3029,
+      "step": 27221
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.788948571428572e-06,
+      "loss": 1.2165,
+      "step": 27222
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.78868e-06,
+      "loss": 1.0005,
+      "step": 27223
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.788411428571429e-06,
+      "loss": 1.0819,
+      "step": 27224
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.788142857142857e-06,
+      "loss": 0.9352,
+      "step": 27225
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7878742857142856e-06,
+      "loss": 0.9719,
+      "step": 27226
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.787605714285715e-06,
+      "loss": 0.9408,
+      "step": 27227
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.787337142857143e-06,
+      "loss": 0.977,
+      "step": 27228
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.787068571428572e-06,
+      "loss": 1.3342,
+      "step": 27229
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7868e-06,
+      "loss": 1.2544,
+      "step": 27230
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.786531428571428e-06,
+      "loss": 1.1864,
+      "step": 27231
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.786262857142858e-06,
+      "loss": 1.0524,
+      "step": 27232
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.785994285714286e-06,
+      "loss": 1.1055,
+      "step": 27233
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.785725714285715e-06,
+      "loss": 1.2655,
+      "step": 27234
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.785457142857143e-06,
+      "loss": 1.3786,
+      "step": 27235
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7851885714285716e-06,
+      "loss": 1.3496,
+      "step": 27236
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.784920000000001e-06,
+      "loss": 1.2263,
+      "step": 27237
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.784651428571429e-06,
+      "loss": 0.8999,
+      "step": 27238
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.784382857142857e-06,
+      "loss": 1.5222,
+      "step": 27239
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.784114285714286e-06,
+      "loss": 1.1328,
+      "step": 27240
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.783845714285714e-06,
+      "loss": 0.7083,
+      "step": 27241
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.783577142857143e-06,
+      "loss": 1.2857,
+      "step": 27242
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.783308571428571e-06,
+      "loss": 1.0204,
+      "step": 27243
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7830399999999995e-06,
+      "loss": 1.3792,
+      "step": 27244
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7827714285714285e-06,
+      "loss": 1.4097,
+      "step": 27245
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7825028571428576e-06,
+      "loss": 0.7016,
+      "step": 27246
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.782234285714287e-06,
+      "loss": 0.8567,
+      "step": 27247
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.781965714285715e-06,
+      "loss": 1.1065,
+      "step": 27248
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.781697142857143e-06,
+      "loss": 0.9427,
+      "step": 27249
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.781428571428572e-06,
+      "loss": 0.9799,
+      "step": 27250
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.78116e-06,
+      "loss": 1.0523,
+      "step": 27251
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.780891428571429e-06,
+      "loss": 1.1279,
+      "step": 27252
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.780622857142857e-06,
+      "loss": 1.0787,
+      "step": 27253
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7803542857142855e-06,
+      "loss": 1.1182,
+      "step": 27254
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7800857142857145e-06,
+      "loss": 1.107,
+      "step": 27255
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.779817142857143e-06,
+      "loss": 1.1036,
+      "step": 27256
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.779548571428572e-06,
+      "loss": 1.3644,
+      "step": 27257
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.77928e-06,
+      "loss": 1.0751,
+      "step": 27258
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.779011428571429e-06,
+      "loss": 1.3497,
+      "step": 27259
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.778742857142858e-06,
+      "loss": 1.217,
+      "step": 27260
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.778474285714286e-06,
+      "loss": 0.9403,
+      "step": 27261
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.778205714285714e-06,
+      "loss": 0.9124,
+      "step": 27262
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.777937142857143e-06,
+      "loss": 1.0656,
+      "step": 27263
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7776685714285715e-06,
+      "loss": 0.9664,
+      "step": 27264
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7774000000000005e-06,
+      "loss": 1.1572,
+      "step": 27265
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.777131428571429e-06,
+      "loss": 0.8695,
+      "step": 27266
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.776862857142857e-06,
+      "loss": 1.1913,
+      "step": 27267
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.776594285714286e-06,
+      "loss": 1.3575,
+      "step": 27268
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.776325714285714e-06,
+      "loss": 1.3536,
+      "step": 27269
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.776057142857143e-06,
+      "loss": 1.0665,
+      "step": 27270
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.775788571428571e-06,
+      "loss": 1.0509,
+      "step": 27271
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.77552e-06,
+      "loss": 1.3023,
+      "step": 27272
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.775251428571429e-06,
+      "loss": 0.9676,
+      "step": 27273
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7749828571428575e-06,
+      "loss": 0.9792,
+      "step": 27274
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7747142857142865e-06,
+      "loss": 1.0159,
+      "step": 27275
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.774445714285715e-06,
+      "loss": 0.9954,
+      "step": 27276
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.774177142857143e-06,
+      "loss": 0.5325,
+      "step": 27277
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.773908571428572e-06,
+      "loss": 1.2027,
+      "step": 27278
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.77364e-06,
+      "loss": 1.2803,
+      "step": 27279
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.773371428571429e-06,
+      "loss": 0.8416,
+      "step": 27280
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.773102857142857e-06,
+      "loss": 1.0371,
+      "step": 27281
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7728342857142855e-06,
+      "loss": 1.2981,
+      "step": 27282
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7725657142857145e-06,
+      "loss": 1.2299,
+      "step": 27283
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.772297142857143e-06,
+      "loss": 1.253,
+      "step": 27284
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.772028571428572e-06,
+      "loss": 0.8745,
+      "step": 27285
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.771760000000001e-06,
+      "loss": 0.8415,
+      "step": 27286
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.771491428571429e-06,
+      "loss": 0.9623,
+      "step": 27287
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.771222857142858e-06,
+      "loss": 1.0579,
+      "step": 27288
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.770954285714286e-06,
+      "loss": 1.1445,
+      "step": 27289
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.770685714285714e-06,
+      "loss": 1.2619,
+      "step": 27290
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.770417142857143e-06,
+      "loss": 0.979,
+      "step": 27291
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7701485714285715e-06,
+      "loss": 1.1068,
+      "step": 27292
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7698800000000005e-06,
+      "loss": 1.6465,
+      "step": 27293
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.769611428571429e-06,
+      "loss": 1.2632,
+      "step": 27294
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.769342857142857e-06,
+      "loss": 0.9273,
+      "step": 27295
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.769074285714286e-06,
+      "loss": 1.2193,
+      "step": 27296
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.768805714285714e-06,
+      "loss": 1.3772,
+      "step": 27297
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.768537142857143e-06,
+      "loss": 1.6238,
+      "step": 27298
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.768268571428572e-06,
+      "loss": 1.0235,
+      "step": 27299
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.768e-06,
+      "loss": 1.0774,
+      "step": 27300
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.767731428571429e-06,
+      "loss": 1.0085,
+      "step": 27301
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7674628571428575e-06,
+      "loss": 0.5839,
+      "step": 27302
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.767194285714286e-06,
+      "loss": 1.0414,
+      "step": 27303
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.766925714285715e-06,
+      "loss": 1.3523,
+      "step": 27304
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.766657142857143e-06,
+      "loss": 0.736,
+      "step": 27305
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.766388571428572e-06,
+      "loss": 0.8886,
+      "step": 27306
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.76612e-06,
+      "loss": 0.9681,
+      "step": 27307
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.765851428571428e-06,
+      "loss": 1.1288,
+      "step": 27308
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.765582857142857e-06,
+      "loss": 1.2552,
+      "step": 27309
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.765314285714285e-06,
+      "loss": 1.0428,
+      "step": 27310
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7650457142857144e-06,
+      "loss": 1.2413,
+      "step": 27311
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7647771428571434e-06,
+      "loss": 1.347,
+      "step": 27312
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.764508571428572e-06,
+      "loss": 1.12,
+      "step": 27313
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.764240000000001e-06,
+      "loss": 0.9996,
+      "step": 27314
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.763971428571429e-06,
+      "loss": 1.3933,
+      "step": 27315
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.763702857142858e-06,
+      "loss": 1.1092,
+      "step": 27316
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.763434285714286e-06,
+      "loss": 1.292,
+      "step": 27317
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.763165714285714e-06,
+      "loss": 0.9188,
+      "step": 27318
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.762897142857143e-06,
+      "loss": 1.1515,
+      "step": 27319
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.762628571428571e-06,
+      "loss": 1.0386,
+      "step": 27320
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.76236e-06,
+      "loss": 1.3458,
+      "step": 27321
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.762091428571429e-06,
+      "loss": 1.4329,
+      "step": 27322
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.761822857142857e-06,
+      "loss": 1.1941,
+      "step": 27323
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.761554285714286e-06,
+      "loss": 1.5964,
+      "step": 27324
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.761285714285715e-06,
+      "loss": 1.4193,
+      "step": 27325
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.761017142857143e-06,
+      "loss": 1.2379,
+      "step": 27326
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.760748571428572e-06,
+      "loss": 1.0103,
+      "step": 27327
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.76048e-06,
+      "loss": 0.8692,
+      "step": 27328
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.760211428571429e-06,
+      "loss": 1.2736,
+      "step": 27329
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.759942857142857e-06,
+      "loss": 1.24,
+      "step": 27330
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7596742857142856e-06,
+      "loss": 0.8668,
+      "step": 27331
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.759405714285715e-06,
+      "loss": 1.4361,
+      "step": 27332
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.759137142857143e-06,
+      "loss": 1.7446,
+      "step": 27333
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.758868571428572e-06,
+      "loss": 0.7791,
+      "step": 27334
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7586e-06,
+      "loss": 1.0886,
+      "step": 27335
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.758331428571428e-06,
+      "loss": 1.0925,
+      "step": 27336
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.758062857142857e-06,
+      "loss": 1.3042,
+      "step": 27337
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.757794285714286e-06,
+      "loss": 1.004,
+      "step": 27338
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.757525714285715e-06,
+      "loss": 1.1257,
+      "step": 27339
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.757257142857143e-06,
+      "loss": 0.7189,
+      "step": 27340
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7569885714285716e-06,
+      "loss": 0.9598,
+      "step": 27341
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.756720000000001e-06,
+      "loss": 1.3254,
+      "step": 27342
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.756451428571429e-06,
+      "loss": 1.2845,
+      "step": 27343
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.756182857142858e-06,
+      "loss": 1.171,
+      "step": 27344
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.755914285714286e-06,
+      "loss": 1.2087,
+      "step": 27345
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.755645714285714e-06,
+      "loss": 1.1926,
+      "step": 27346
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.755377142857143e-06,
+      "loss": 1.2999,
+      "step": 27347
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.755108571428571e-06,
+      "loss": 0.6215,
+      "step": 27348
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7548399999999995e-06,
+      "loss": 1.4669,
+      "step": 27349
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7545714285714285e-06,
+      "loss": 1.605,
+      "step": 27350
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7543028571428576e-06,
+      "loss": 1.5149,
+      "step": 27351
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.754034285714287e-06,
+      "loss": 1.4406,
+      "step": 27352
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.753765714285715e-06,
+      "loss": 1.1521,
+      "step": 27353
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.753497142857143e-06,
+      "loss": 1.2239,
+      "step": 27354
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.753228571428572e-06,
+      "loss": 1.3033,
+      "step": 27355
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.75296e-06,
+      "loss": 1.0686,
+      "step": 27356
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.752691428571429e-06,
+      "loss": 1.0316,
+      "step": 27357
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.752422857142857e-06,
+      "loss": 1.2579,
+      "step": 27358
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7521542857142855e-06,
+      "loss": 0.9596,
+      "step": 27359
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7518857142857145e-06,
+      "loss": 0.9956,
+      "step": 27360
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.751617142857143e-06,
+      "loss": 1.22,
+      "step": 27361
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.751348571428572e-06,
+      "loss": 0.6685,
+      "step": 27362
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.75108e-06,
+      "loss": 1.069,
+      "step": 27363
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.750811428571429e-06,
+      "loss": 0.9632,
+      "step": 27364
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.750542857142858e-06,
+      "loss": 1.1116,
+      "step": 27365
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.750274285714286e-06,
+      "loss": 1.365,
+      "step": 27366
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.750005714285715e-06,
+      "loss": 1.4697,
+      "step": 27367
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.749737142857143e-06,
+      "loss": 1.287,
+      "step": 27368
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7494685714285715e-06,
+      "loss": 1.0977,
+      "step": 27369
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7492000000000005e-06,
+      "loss": 1.3777,
+      "step": 27370
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.748931428571429e-06,
+      "loss": 0.7871,
+      "step": 27371
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.748662857142857e-06,
+      "loss": 0.9694,
+      "step": 27372
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.748394285714286e-06,
+      "loss": 1.1391,
+      "step": 27373
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.748125714285714e-06,
+      "loss": 1.1481,
+      "step": 27374
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.747857142857143e-06,
+      "loss": 1.2392,
+      "step": 27375
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.747588571428571e-06,
+      "loss": 1.1139,
+      "step": 27376
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.74732e-06,
+      "loss": 1.0283,
+      "step": 27377
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.747051428571429e-06,
+      "loss": 1.1979,
+      "step": 27378
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7467828571428575e-06,
+      "loss": 1.0985,
+      "step": 27379
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7465142857142865e-06,
+      "loss": 0.9281,
+      "step": 27380
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.746245714285715e-06,
+      "loss": 0.886,
+      "step": 27381
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.745977142857143e-06,
+      "loss": 1.3339,
+      "step": 27382
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.745708571428572e-06,
+      "loss": 0.7813,
+      "step": 27383
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.74544e-06,
+      "loss": 1.226,
+      "step": 27384
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.745171428571429e-06,
+      "loss": 1.0798,
+      "step": 27385
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.744902857142857e-06,
+      "loss": 0.7057,
+      "step": 27386
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.7446342857142855e-06,
+      "loss": 1.3795,
+      "step": 27387
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7443657142857145e-06,
+      "loss": 1.2528,
+      "step": 27388
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.744097142857143e-06,
+      "loss": 0.849,
+      "step": 27389
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.743828571428572e-06,
+      "loss": 0.9976,
+      "step": 27390
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.743560000000001e-06,
+      "loss": 0.8743,
+      "step": 27391
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.743291428571429e-06,
+      "loss": 1.5835,
+      "step": 27392
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.743022857142858e-06,
+      "loss": 1.2556,
+      "step": 27393
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.742754285714286e-06,
+      "loss": 0.978,
+      "step": 27394
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.742485714285714e-06,
+      "loss": 1.4728,
+      "step": 27395
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.742217142857143e-06,
+      "loss": 1.2831,
+      "step": 27396
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7419485714285715e-06,
+      "loss": 1.1738,
+      "step": 27397
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7416800000000005e-06,
+      "loss": 1.1984,
+      "step": 27398
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.741411428571429e-06,
+      "loss": 0.7221,
+      "step": 27399
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.741142857142857e-06,
+      "loss": 1.3123,
+      "step": 27400
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.740874285714286e-06,
+      "loss": 0.9022,
+      "step": 27401
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.740605714285714e-06,
+      "loss": 1.3359,
+      "step": 27402
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.740337142857143e-06,
+      "loss": 1.2159,
+      "step": 27403
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.740068571428572e-06,
+      "loss": 0.8058,
+      "step": 27404
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7398e-06,
+      "loss": 1.3338,
+      "step": 27405
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.739531428571429e-06,
+      "loss": 1.246,
+      "step": 27406
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7392628571428575e-06,
+      "loss": 1.06,
+      "step": 27407
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7389942857142865e-06,
+      "loss": 0.915,
+      "step": 27408
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.738725714285715e-06,
+      "loss": 0.7779,
+      "step": 27409
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.738457142857143e-06,
+      "loss": 1.5516,
+      "step": 27410
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.738188571428572e-06,
+      "loss": 1.0662,
+      "step": 27411
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.73792e-06,
+      "loss": 1.313,
+      "step": 27412
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.737651428571428e-06,
+      "loss": 1.3426,
+      "step": 27413
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.737382857142857e-06,
+      "loss": 1.4577,
+      "step": 27414
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.737114285714285e-06,
+      "loss": 1.0861,
+      "step": 27415
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7368457142857144e-06,
+      "loss": 0.6251,
+      "step": 27416
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7365771428571434e-06,
+      "loss": 0.946,
+      "step": 27417
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.736308571428572e-06,
+      "loss": 1.4135,
+      "step": 27418
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.736040000000001e-06,
+      "loss": 1.4855,
+      "step": 27419
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.735771428571429e-06,
+      "loss": 1.3158,
+      "step": 27420
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.735502857142858e-06,
+      "loss": 0.5226,
+      "step": 27421
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.735234285714286e-06,
+      "loss": 0.9853,
+      "step": 27422
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.734965714285714e-06,
+      "loss": 1.5929,
+      "step": 27423
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.734697142857143e-06,
+      "loss": 0.95,
+      "step": 27424
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.734428571428571e-06,
+      "loss": 1.0161,
+      "step": 27425
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.73416e-06,
+      "loss": 0.9674,
+      "step": 27426
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.733891428571429e-06,
+      "loss": 1.3297,
+      "step": 27427
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.733622857142857e-06,
+      "loss": 0.9523,
+      "step": 27428
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.733354285714286e-06,
+      "loss": 1.4123,
+      "step": 27429
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.733085714285715e-06,
+      "loss": 1.0239,
+      "step": 27430
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.732817142857144e-06,
+      "loss": 1.3055,
+      "step": 27431
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.732548571428572e-06,
+      "loss": 1.2138,
+      "step": 27432
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.73228e-06,
+      "loss": 1.0253,
+      "step": 27433
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.732011428571429e-06,
+      "loss": 1.0921,
+      "step": 27434
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.731742857142857e-06,
+      "loss": 1.0168,
+      "step": 27435
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7314742857142856e-06,
+      "loss": 1.1079,
+      "step": 27436
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.731205714285715e-06,
+      "loss": 0.9171,
+      "step": 27437
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.730937142857143e-06,
+      "loss": 1.0033,
+      "step": 27438
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.730668571428572e-06,
+      "loss": 0.9848,
+      "step": 27439
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7304e-06,
+      "loss": 1.3409,
+      "step": 27440
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.730131428571428e-06,
+      "loss": 0.7729,
+      "step": 27441
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.729862857142857e-06,
+      "loss": 1.1491,
+      "step": 27442
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.729594285714286e-06,
+      "loss": 1.2636,
+      "step": 27443
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.729325714285715e-06,
+      "loss": 1.1305,
+      "step": 27444
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.729057142857143e-06,
+      "loss": 1.044,
+      "step": 27445
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7287885714285716e-06,
+      "loss": 1.1878,
+      "step": 27446
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.728520000000001e-06,
+      "loss": 1.3855,
+      "step": 27447
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.728251428571429e-06,
+      "loss": 1.26,
+      "step": 27448
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.727982857142858e-06,
+      "loss": 1.1286,
+      "step": 27449
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.727714285714286e-06,
+      "loss": 1.1642,
+      "step": 27450
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.727445714285714e-06,
+      "loss": 0.8386,
+      "step": 27451
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.727177142857143e-06,
+      "loss": 1.1972,
+      "step": 27452
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.726908571428571e-06,
+      "loss": 1.1227,
+      "step": 27453
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.72664e-06,
+      "loss": 1.1156,
+      "step": 27454
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7263714285714285e-06,
+      "loss": 0.8675,
+      "step": 27455
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7261028571428576e-06,
+      "loss": 1.1533,
+      "step": 27456
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.725834285714287e-06,
+      "loss": 1.2014,
+      "step": 27457
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.725565714285715e-06,
+      "loss": 0.9501,
+      "step": 27458
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.725297142857143e-06,
+      "loss": 1.0505,
+      "step": 27459
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.725028571428572e-06,
+      "loss": 1.4679,
+      "step": 27460
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.72476e-06,
+      "loss": 1.2853,
+      "step": 27461
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.724491428571429e-06,
+      "loss": 1.2028,
+      "step": 27462
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.724222857142857e-06,
+      "loss": 0.9142,
+      "step": 27463
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7239542857142855e-06,
+      "loss": 1.1073,
+      "step": 27464
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7236857142857145e-06,
+      "loss": 0.9742,
+      "step": 27465
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.723417142857143e-06,
+      "loss": 1.1507,
+      "step": 27466
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.723148571428572e-06,
+      "loss": 1.0839,
+      "step": 27467
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.72288e-06,
+      "loss": 1.9462,
+      "step": 27468
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.722611428571429e-06,
+      "loss": 1.0613,
+      "step": 27469
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.722342857142858e-06,
+      "loss": 1.1184,
+      "step": 27470
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.722074285714286e-06,
+      "loss": 1.1327,
+      "step": 27471
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.721805714285715e-06,
+      "loss": 1.017,
+      "step": 27472
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.721537142857143e-06,
+      "loss": 1.2392,
+      "step": 27473
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7212685714285715e-06,
+      "loss": 1.0579,
+      "step": 27474
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7210000000000005e-06,
+      "loss": 1.1163,
+      "step": 27475
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.720731428571429e-06,
+      "loss": 0.8739,
+      "step": 27476
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.720462857142857e-06,
+      "loss": 0.8299,
+      "step": 27477
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.720194285714286e-06,
+      "loss": 1.309,
+      "step": 27478
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.719925714285714e-06,
+      "loss": 0.857,
+      "step": 27479
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.719657142857143e-06,
+      "loss": 1.0816,
+      "step": 27480
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.719388571428571e-06,
+      "loss": 0.8773,
+      "step": 27481
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.71912e-06,
+      "loss": 1.059,
+      "step": 27482
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.718851428571429e-06,
+      "loss": 1.0409,
+      "step": 27483
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7185828571428575e-06,
+      "loss": 1.4182,
+      "step": 27484
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7183142857142865e-06,
+      "loss": 1.282,
+      "step": 27485
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.718045714285715e-06,
+      "loss": 1.1817,
+      "step": 27486
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.717777142857143e-06,
+      "loss": 0.9499,
+      "step": 27487
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.717508571428572e-06,
+      "loss": 1.2469,
+      "step": 27488
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.71724e-06,
+      "loss": 0.7526,
+      "step": 27489
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.716971428571429e-06,
+      "loss": 1.253,
+      "step": 27490
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.716702857142857e-06,
+      "loss": 0.6022,
+      "step": 27491
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7164342857142855e-06,
+      "loss": 1.4081,
+      "step": 27492
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7161657142857145e-06,
+      "loss": 1.0815,
+      "step": 27493
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.715897142857143e-06,
+      "loss": 0.9779,
+      "step": 27494
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.715628571428572e-06,
+      "loss": 0.7456,
+      "step": 27495
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.715360000000001e-06,
+      "loss": 1.1447,
+      "step": 27496
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.715091428571429e-06,
+      "loss": 1.4114,
+      "step": 27497
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.714822857142858e-06,
+      "loss": 1.3381,
+      "step": 27498
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.714554285714286e-06,
+      "loss": 0.9343,
+      "step": 27499
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.714285714285714e-06,
+      "loss": 0.8831,
+      "step": 27500
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.714017142857143e-06,
+      "loss": 0.7754,
+      "step": 27501
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7137485714285715e-06,
+      "loss": 0.9991,
+      "step": 27502
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7134800000000005e-06,
+      "loss": 1.2358,
+      "step": 27503
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.713211428571429e-06,
+      "loss": 1.1271,
+      "step": 27504
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.712942857142857e-06,
+      "loss": 1.2931,
+      "step": 27505
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.712674285714286e-06,
+      "loss": 0.8484,
+      "step": 27506
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.712405714285714e-06,
+      "loss": 1.1735,
+      "step": 27507
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.712137142857143e-06,
+      "loss": 1.0934,
+      "step": 27508
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.711868571428572e-06,
+      "loss": 0.8326,
+      "step": 27509
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7116e-06,
+      "loss": 0.7587,
+      "step": 27510
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.711331428571429e-06,
+      "loss": 1.0069,
+      "step": 27511
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7110628571428574e-06,
+      "loss": 0.8694,
+      "step": 27512
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7107942857142865e-06,
+      "loss": 1.2086,
+      "step": 27513
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.710525714285715e-06,
+      "loss": 1.2806,
+      "step": 27514
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.710257142857143e-06,
+      "loss": 0.8703,
+      "step": 27515
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.709988571428572e-06,
+      "loss": 0.9195,
+      "step": 27516
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.70972e-06,
+      "loss": 1.6883,
+      "step": 27517
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.709451428571429e-06,
+      "loss": 0.9754,
+      "step": 27518
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.709182857142857e-06,
+      "loss": 1.1513,
+      "step": 27519
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.708914285714285e-06,
+      "loss": 0.8378,
+      "step": 27520
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.708645714285714e-06,
+      "loss": 1.1672,
+      "step": 27521
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7083771428571434e-06,
+      "loss": 1.3346,
+      "step": 27522
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.708108571428572e-06,
+      "loss": 0.9327,
+      "step": 27523
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.707840000000001e-06,
+      "loss": 1.5733,
+      "step": 27524
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.707571428571429e-06,
+      "loss": 1.0704,
+      "step": 27525
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.707302857142858e-06,
+      "loss": 1.0679,
+      "step": 27526
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.707034285714286e-06,
+      "loss": 1.2513,
+      "step": 27527
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.706765714285714e-06,
+      "loss": 0.7224,
+      "step": 27528
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.706497142857143e-06,
+      "loss": 1.1457,
+      "step": 27529
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.706228571428571e-06,
+      "loss": 0.7632,
+      "step": 27530
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.70596e-06,
+      "loss": 1.5298,
+      "step": 27531
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.705691428571429e-06,
+      "loss": 1.033,
+      "step": 27532
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.705422857142857e-06,
+      "loss": 1.2458,
+      "step": 27533
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.705154285714286e-06,
+      "loss": 1.0588,
+      "step": 27534
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.704885714285715e-06,
+      "loss": 1.0387,
+      "step": 27535
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.704617142857144e-06,
+      "loss": 1.4718,
+      "step": 27536
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.704348571428572e-06,
+      "loss": 0.9112,
+      "step": 27537
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.70408e-06,
+      "loss": 1.4254,
+      "step": 27538
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.703811428571429e-06,
+      "loss": 0.9348,
+      "step": 27539
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.703542857142857e-06,
+      "loss": 1.2252,
+      "step": 27540
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7032742857142856e-06,
+      "loss": 0.9477,
+      "step": 27541
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.703005714285715e-06,
+      "loss": 1.2917,
+      "step": 27542
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.702737142857143e-06,
+      "loss": 1.1471,
+      "step": 27543
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.702468571428572e-06,
+      "loss": 0.9767,
+      "step": 27544
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7022e-06,
+      "loss": 1.1541,
+      "step": 27545
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.701931428571428e-06,
+      "loss": 1.0337,
+      "step": 27546
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.701662857142857e-06,
+      "loss": 1.0356,
+      "step": 27547
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.701394285714286e-06,
+      "loss": 1.434,
+      "step": 27548
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.701125714285715e-06,
+      "loss": 0.9884,
+      "step": 27549
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.700857142857143e-06,
+      "loss": 1.3661,
+      "step": 27550
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.7005885714285716e-06,
+      "loss": 0.616,
+      "step": 27551
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.700320000000001e-06,
+      "loss": 1.6453,
+      "step": 27552
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.700051428571429e-06,
+      "loss": 1.1323,
+      "step": 27553
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.699782857142858e-06,
+      "loss": 1.0907,
+      "step": 27554
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.699514285714286e-06,
+      "loss": 0.8956,
+      "step": 27555
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.699245714285714e-06,
+      "loss": 1.0045,
+      "step": 27556
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.698977142857143e-06,
+      "loss": 1.0809,
+      "step": 27557
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.698708571428571e-06,
+      "loss": 0.881,
+      "step": 27558
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.69844e-06,
+      "loss": 1.4972,
+      "step": 27559
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.6981714285714285e-06,
+      "loss": 1.0161,
+      "step": 27560
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.6979028571428576e-06,
+      "loss": 1.2411,
+      "step": 27561
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.697634285714287e-06,
+      "loss": 0.974,
+      "step": 27562
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.697365714285715e-06,
+      "loss": 1.1135,
+      "step": 27563
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.697097142857143e-06,
+      "loss": 1.2115,
+      "step": 27564
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.696828571428572e-06,
+      "loss": 0.8706,
+      "step": 27565
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.69656e-06,
+      "loss": 1.0343,
+      "step": 27566
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.696291428571429e-06,
+      "loss": 1.1992,
+      "step": 27567
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.696022857142857e-06,
+      "loss": 1.3135,
+      "step": 27568
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6957542857142855e-06,
+      "loss": 1.2301,
+      "step": 27569
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6954857142857145e-06,
+      "loss": 1.1216,
+      "step": 27570
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.695217142857143e-06,
+      "loss": 1.081,
+      "step": 27571
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.694948571428572e-06,
+      "loss": 1.4764,
+      "step": 27572
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.69468e-06,
+      "loss": 1.1207,
+      "step": 27573
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.694411428571429e-06,
+      "loss": 1.0281,
+      "step": 27574
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.694142857142858e-06,
+      "loss": 1.0364,
+      "step": 27575
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.693874285714286e-06,
+      "loss": 1.183,
+      "step": 27576
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.693605714285715e-06,
+      "loss": 1.0969,
+      "step": 27577
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.693337142857143e-06,
+      "loss": 1.4912,
+      "step": 27578
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6930685714285715e-06,
+      "loss": 1.1121,
+      "step": 27579
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6928000000000005e-06,
+      "loss": 0.9723,
+      "step": 27580
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.692531428571429e-06,
+      "loss": 1.1392,
+      "step": 27581
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.692262857142858e-06,
+      "loss": 0.9197,
+      "step": 27582
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.691994285714286e-06,
+      "loss": 1.1688,
+      "step": 27583
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.691725714285714e-06,
+      "loss": 0.9425,
+      "step": 27584
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.691457142857143e-06,
+      "loss": 1.4964,
+      "step": 27585
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.691188571428571e-06,
+      "loss": 1.2792,
+      "step": 27586
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6909199999999995e-06,
+      "loss": 1.8464,
+      "step": 27587
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.690651428571429e-06,
+      "loss": 1.0274,
+      "step": 27588
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6903828571428575e-06,
+      "loss": 1.6029,
+      "step": 27589
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6901142857142865e-06,
+      "loss": 1.2526,
+      "step": 27590
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.689845714285715e-06,
+      "loss": 1.0927,
+      "step": 27591
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.689577142857143e-06,
+      "loss": 1.1797,
+      "step": 27592
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.689308571428572e-06,
+      "loss": 1.1622,
+      "step": 27593
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.68904e-06,
+      "loss": 1.0942,
+      "step": 27594
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.688771428571429e-06,
+      "loss": 1.4322,
+      "step": 27595
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.688502857142857e-06,
+      "loss": 1.2306,
+      "step": 27596
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6882342857142855e-06,
+      "loss": 1.1351,
+      "step": 27597
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6879657142857145e-06,
+      "loss": 1.0631,
+      "step": 27598
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.687697142857143e-06,
+      "loss": 1.3403,
+      "step": 27599
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.687428571428572e-06,
+      "loss": 1.2903,
+      "step": 27600
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.687160000000001e-06,
+      "loss": 0.7555,
+      "step": 27601
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.686891428571429e-06,
+      "loss": 1.1429,
+      "step": 27602
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.686622857142858e-06,
+      "loss": 1.4128,
+      "step": 27603
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.686354285714286e-06,
+      "loss": 1.2808,
+      "step": 27604
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.686085714285715e-06,
+      "loss": 1.0251,
+      "step": 27605
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.685817142857143e-06,
+      "loss": 1.2092,
+      "step": 27606
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6855485714285715e-06,
+      "loss": 1.183,
+      "step": 27607
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6852800000000005e-06,
+      "loss": 0.738,
+      "step": 27608
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.685011428571429e-06,
+      "loss": 1.4111,
+      "step": 27609
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.684742857142857e-06,
+      "loss": 1.252,
+      "step": 27610
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.684474285714286e-06,
+      "loss": 1.4314,
+      "step": 27611
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.684205714285714e-06,
+      "loss": 0.5136,
+      "step": 27612
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.683937142857143e-06,
+      "loss": 0.906,
+      "step": 27613
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.683668571428572e-06,
+      "loss": 1.0883,
+      "step": 27614
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6834e-06,
+      "loss": 0.9021,
+      "step": 27615
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.683131428571429e-06,
+      "loss": 1.5784,
+      "step": 27616
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6828628571428574e-06,
+      "loss": 1.4304,
+      "step": 27617
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6825942857142865e-06,
+      "loss": 1.4695,
+      "step": 27618
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.682325714285715e-06,
+      "loss": 0.9294,
+      "step": 27619
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.682057142857143e-06,
+      "loss": 1.4136,
+      "step": 27620
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.681788571428572e-06,
+      "loss": 1.0234,
+      "step": 27621
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.68152e-06,
+      "loss": 0.7451,
+      "step": 27622
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.681251428571429e-06,
+      "loss": 1.2891,
+      "step": 27623
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.680982857142857e-06,
+      "loss": 1.3679,
+      "step": 27624
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.680714285714285e-06,
+      "loss": 1.1068,
+      "step": 27625
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.680445714285714e-06,
+      "loss": 1.18,
+      "step": 27626
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6801771428571434e-06,
+      "loss": 0.9946,
+      "step": 27627
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.679908571428572e-06,
+      "loss": 1.0635,
+      "step": 27628
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.679640000000001e-06,
+      "loss": 0.8357,
+      "step": 27629
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.679371428571429e-06,
+      "loss": 1.1167,
+      "step": 27630
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.679102857142858e-06,
+      "loss": 1.1448,
+      "step": 27631
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.678834285714286e-06,
+      "loss": 1.3852,
+      "step": 27632
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.678565714285714e-06,
+      "loss": 0.8181,
+      "step": 27633
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.678297142857143e-06,
+      "loss": 0.6392,
+      "step": 27634
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.678028571428571e-06,
+      "loss": 0.796,
+      "step": 27635
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.67776e-06,
+      "loss": 1.0414,
+      "step": 27636
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.677491428571429e-06,
+      "loss": 1.008,
+      "step": 27637
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.677222857142857e-06,
+      "loss": 1.354,
+      "step": 27638
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.676954285714286e-06,
+      "loss": 0.8172,
+      "step": 27639
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.676685714285714e-06,
+      "loss": 0.9613,
+      "step": 27640
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.676417142857144e-06,
+      "loss": 1.158,
+      "step": 27641
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.676148571428572e-06,
+      "loss": 1.0478,
+      "step": 27642
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.67588e-06,
+      "loss": 0.7161,
+      "step": 27643
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.675611428571429e-06,
+      "loss": 1.3576,
+      "step": 27644
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.675342857142857e-06,
+      "loss": 1.1539,
+      "step": 27645
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.675074285714286e-06,
+      "loss": 1.2245,
+      "step": 27646
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.674805714285715e-06,
+      "loss": 1.1785,
+      "step": 27647
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.674537142857143e-06,
+      "loss": 1.3253,
+      "step": 27648
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.674268571428572e-06,
+      "loss": 0.6559,
+      "step": 27649
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.674e-06,
+      "loss": 0.5653,
+      "step": 27650
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.673731428571428e-06,
+      "loss": 0.8517,
+      "step": 27651
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.673462857142857e-06,
+      "loss": 1.4982,
+      "step": 27652
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.673194285714285e-06,
+      "loss": 1.2958,
+      "step": 27653
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.672925714285715e-06,
+      "loss": 1.239,
+      "step": 27654
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.672657142857143e-06,
+      "loss": 1.3625,
+      "step": 27655
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6723885714285716e-06,
+      "loss": 1.239,
+      "step": 27656
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.672120000000001e-06,
+      "loss": 1.582,
+      "step": 27657
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.671851428571429e-06,
+      "loss": 1.4385,
+      "step": 27658
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.671582857142858e-06,
+      "loss": 1.212,
+      "step": 27659
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.671314285714286e-06,
+      "loss": 0.9879,
+      "step": 27660
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.671045714285714e-06,
+      "loss": 1.1189,
+      "step": 27661
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.670777142857143e-06,
+      "loss": 1.2935,
+      "step": 27662
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.670508571428571e-06,
+      "loss": 1.1,
+      "step": 27663
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.67024e-06,
+      "loss": 1.0069,
+      "step": 27664
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6699714285714285e-06,
+      "loss": 1.0851,
+      "step": 27665
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.669702857142857e-06,
+      "loss": 0.7767,
+      "step": 27666
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.669434285714287e-06,
+      "loss": 1.1271,
+      "step": 27667
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.669165714285715e-06,
+      "loss": 1.3635,
+      "step": 27668
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.668897142857144e-06,
+      "loss": 1.2706,
+      "step": 27669
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.668628571428572e-06,
+      "loss": 1.1692,
+      "step": 27670
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.66836e-06,
+      "loss": 1.1009,
+      "step": 27671
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.668091428571429e-06,
+      "loss": 0.9361,
+      "step": 27672
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.667822857142857e-06,
+      "loss": 1.31,
+      "step": 27673
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6675542857142855e-06,
+      "loss": 1.3703,
+      "step": 27674
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6672857142857145e-06,
+      "loss": 1.2434,
+      "step": 27675
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.667017142857143e-06,
+      "loss": 1.064,
+      "step": 27676
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.666748571428572e-06,
+      "loss": 0.9684,
+      "step": 27677
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.66648e-06,
+      "loss": 0.7519,
+      "step": 27678
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.666211428571428e-06,
+      "loss": 1.0692,
+      "step": 27679
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.665942857142858e-06,
+      "loss": 1.4734,
+      "step": 27680
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.665674285714286e-06,
+      "loss": 1.1732,
+      "step": 27681
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.665405714285715e-06,
+      "loss": 0.9098,
+      "step": 27682
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.665137142857143e-06,
+      "loss": 0.9274,
+      "step": 27683
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6648685714285715e-06,
+      "loss": 1.2377,
+      "step": 27684
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6646000000000005e-06,
+      "loss": 0.8401,
+      "step": 27685
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.664331428571429e-06,
+      "loss": 1.2071,
+      "step": 27686
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.664062857142858e-06,
+      "loss": 1.4224,
+      "step": 27687
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.663794285714286e-06,
+      "loss": 0.9888,
+      "step": 27688
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.663525714285714e-06,
+      "loss": 0.8178,
+      "step": 27689
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.663257142857143e-06,
+      "loss": 1.0454,
+      "step": 27690
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.662988571428571e-06,
+      "loss": 1.3506,
+      "step": 27691
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.66272e-06,
+      "loss": 1.2068,
+      "step": 27692
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6624514285714285e-06,
+      "loss": 1.1553,
+      "step": 27693
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6621828571428575e-06,
+      "loss": 0.7448,
+      "step": 27694
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6619142857142865e-06,
+      "loss": 0.7458,
+      "step": 27695
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.661645714285715e-06,
+      "loss": 0.9297,
+      "step": 27696
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.661377142857143e-06,
+      "loss": 0.8822,
+      "step": 27697
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.661108571428572e-06,
+      "loss": 1.157,
+      "step": 27698
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.66084e-06,
+      "loss": 0.9809,
+      "step": 27699
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.660571428571429e-06,
+      "loss": 0.8836,
+      "step": 27700
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.660302857142857e-06,
+      "loss": 0.9357,
+      "step": 27701
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6600342857142855e-06,
+      "loss": 1.3032,
+      "step": 27702
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6597657142857145e-06,
+      "loss": 1.1407,
+      "step": 27703
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.659497142857143e-06,
+      "loss": 1.2551,
+      "step": 27704
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.659228571428572e-06,
+      "loss": 0.9904,
+      "step": 27705
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.65896e-06,
+      "loss": 1.15,
+      "step": 27706
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.658691428571429e-06,
+      "loss": 1.1542,
+      "step": 27707
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.658422857142858e-06,
+      "loss": 1.6036,
+      "step": 27708
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.658154285714286e-06,
+      "loss": 1.7208,
+      "step": 27709
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.657885714285715e-06,
+      "loss": 1.0804,
+      "step": 27710
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.657617142857143e-06,
+      "loss": 1.2387,
+      "step": 27711
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6573485714285715e-06,
+      "loss": 1.4496,
+      "step": 27712
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6570800000000005e-06,
+      "loss": 1.4425,
+      "step": 27713
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.656811428571429e-06,
+      "loss": 1.4383,
+      "step": 27714
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.656542857142857e-06,
+      "loss": 0.866,
+      "step": 27715
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.656274285714286e-06,
+      "loss": 1.4925,
+      "step": 27716
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.656005714285714e-06,
+      "loss": 1.4098,
+      "step": 27717
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.655737142857143e-06,
+      "loss": 1.2423,
+      "step": 27718
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.655468571428571e-06,
+      "loss": 1.1898,
+      "step": 27719
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6552e-06,
+      "loss": 1.0371,
+      "step": 27720
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.654931428571429e-06,
+      "loss": 1.37,
+      "step": 27721
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6546628571428574e-06,
+      "loss": 1.4524,
+      "step": 27722
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6543942857142865e-06,
+      "loss": 1.1454,
+      "step": 27723
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.654125714285715e-06,
+      "loss": 1.1749,
+      "step": 27724
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.653857142857143e-06,
+      "loss": 0.8178,
+      "step": 27725
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.653588571428572e-06,
+      "loss": 1.025,
+      "step": 27726
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.65332e-06,
+      "loss": 1.1087,
+      "step": 27727
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.653051428571429e-06,
+      "loss": 1.1074,
+      "step": 27728
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.652782857142857e-06,
+      "loss": 1.5603,
+      "step": 27729
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.652514285714285e-06,
+      "loss": 1.2313,
+      "step": 27730
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.652245714285714e-06,
+      "loss": 0.9756,
+      "step": 27731
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.651977142857143e-06,
+      "loss": 1.2719,
+      "step": 27732
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.6517085714285725e-06,
+      "loss": 1.5004,
+      "step": 27733
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.651440000000001e-06,
+      "loss": 1.06,
+      "step": 27734
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.651171428571429e-06,
+      "loss": 1.3175,
+      "step": 27735
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.650902857142858e-06,
+      "loss": 1.0483,
+      "step": 27736
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.650634285714286e-06,
+      "loss": 0.9605,
+      "step": 27737
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.650365714285714e-06,
+      "loss": 1.2096,
+      "step": 27738
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.650097142857143e-06,
+      "loss": 0.9955,
+      "step": 27739
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.649828571428571e-06,
+      "loss": 1.3899,
+      "step": 27740
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.64956e-06,
+      "loss": 1.5712,
+      "step": 27741
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.649291428571429e-06,
+      "loss": 1.498,
+      "step": 27742
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.649022857142857e-06,
+      "loss": 0.6926,
+      "step": 27743
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.648754285714286e-06,
+      "loss": 1.0267,
+      "step": 27744
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.648485714285714e-06,
+      "loss": 1.1594,
+      "step": 27745
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.648217142857143e-06,
+      "loss": 1.3051,
+      "step": 27746
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.647948571428572e-06,
+      "loss": 1.1932,
+      "step": 27747
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.64768e-06,
+      "loss": 1.48,
+      "step": 27748
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.647411428571429e-06,
+      "loss": 1.2716,
+      "step": 27749
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.647142857142857e-06,
+      "loss": 0.8922,
+      "step": 27750
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.646874285714286e-06,
+      "loss": 1.0579,
+      "step": 27751
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.646605714285715e-06,
+      "loss": 1.1786,
+      "step": 27752
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.646337142857143e-06,
+      "loss": 0.8831,
+      "step": 27753
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.646068571428572e-06,
+      "loss": 0.8132,
+      "step": 27754
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6458e-06,
+      "loss": 1.4562,
+      "step": 27755
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.645531428571429e-06,
+      "loss": 1.1634,
+      "step": 27756
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.645262857142857e-06,
+      "loss": 1.3384,
+      "step": 27757
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.644994285714285e-06,
+      "loss": 1.2442,
+      "step": 27758
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.644725714285714e-06,
+      "loss": 1.3143,
+      "step": 27759
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.644457142857143e-06,
+      "loss": 1.2537,
+      "step": 27760
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6441885714285716e-06,
+      "loss": 1.2114,
+      "step": 27761
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.643920000000001e-06,
+      "loss": 1.4781,
+      "step": 27762
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.643651428571429e-06,
+      "loss": 0.9996,
+      "step": 27763
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.643382857142858e-06,
+      "loss": 1.0506,
+      "step": 27764
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.643114285714286e-06,
+      "loss": 0.8329,
+      "step": 27765
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.642845714285714e-06,
+      "loss": 1.0952,
+      "step": 27766
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.642577142857143e-06,
+      "loss": 1.2678,
+      "step": 27767
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.642308571428571e-06,
+      "loss": 1.2977,
+      "step": 27768
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.64204e-06,
+      "loss": 1.2264,
+      "step": 27769
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6417714285714285e-06,
+      "loss": 1.1036,
+      "step": 27770
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.641502857142857e-06,
+      "loss": 1.4021,
+      "step": 27771
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.641234285714286e-06,
+      "loss": 1.2288,
+      "step": 27772
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.640965714285715e-06,
+      "loss": 1.1004,
+      "step": 27773
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.640697142857144e-06,
+      "loss": 1.0451,
+      "step": 27774
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.640428571428572e-06,
+      "loss": 0.6749,
+      "step": 27775
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.64016e-06,
+      "loss": 1.3239,
+      "step": 27776
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.639891428571429e-06,
+      "loss": 1.0905,
+      "step": 27777
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.639622857142857e-06,
+      "loss": 1.353,
+      "step": 27778
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.639354285714286e-06,
+      "loss": 0.7438,
+      "step": 27779
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6390857142857145e-06,
+      "loss": 1.3776,
+      "step": 27780
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.638817142857143e-06,
+      "loss": 1.2295,
+      "step": 27781
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.638548571428572e-06,
+      "loss": 1.1143,
+      "step": 27782
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.63828e-06,
+      "loss": 1.0575,
+      "step": 27783
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.638011428571428e-06,
+      "loss": 0.8866,
+      "step": 27784
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.637742857142857e-06,
+      "loss": 0.7353,
+      "step": 27785
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.637474285714286e-06,
+      "loss": 1.0303,
+      "step": 27786
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.637205714285715e-06,
+      "loss": 0.8555,
+      "step": 27787
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.636937142857143e-06,
+      "loss": 0.8506,
+      "step": 27788
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6366685714285715e-06,
+      "loss": 1.2213,
+      "step": 27789
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6364000000000005e-06,
+      "loss": 0.6624,
+      "step": 27790
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.636131428571429e-06,
+      "loss": 1.0894,
+      "step": 27791
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.635862857142858e-06,
+      "loss": 1.2009,
+      "step": 27792
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.635594285714286e-06,
+      "loss": 1.4155,
+      "step": 27793
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.635325714285714e-06,
+      "loss": 1.0458,
+      "step": 27794
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.635057142857143e-06,
+      "loss": 1.3179,
+      "step": 27795
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.634788571428571e-06,
+      "loss": 1.0023,
+      "step": 27796
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.63452e-06,
+      "loss": 0.978,
+      "step": 27797
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6342514285714285e-06,
+      "loss": 1.4693,
+      "step": 27798
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6339828571428575e-06,
+      "loss": 0.8601,
+      "step": 27799
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6337142857142865e-06,
+      "loss": 1.1235,
+      "step": 27800
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.633445714285715e-06,
+      "loss": 0.9301,
+      "step": 27801
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.633177142857143e-06,
+      "loss": 1.3542,
+      "step": 27802
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.632908571428572e-06,
+      "loss": 1.3883,
+      "step": 27803
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.63264e-06,
+      "loss": 1.1335,
+      "step": 27804
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.632371428571429e-06,
+      "loss": 0.9208,
+      "step": 27805
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.632102857142857e-06,
+      "loss": 1.0174,
+      "step": 27806
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6318342857142855e-06,
+      "loss": 1.353,
+      "step": 27807
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6315657142857145e-06,
+      "loss": 1.0381,
+      "step": 27808
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.631297142857143e-06,
+      "loss": 1.0309,
+      "step": 27809
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.631028571428572e-06,
+      "loss": 1.1603,
+      "step": 27810
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.63076e-06,
+      "loss": 1.0792,
+      "step": 27811
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.630491428571429e-06,
+      "loss": 0.9341,
+      "step": 27812
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.630222857142858e-06,
+      "loss": 0.8811,
+      "step": 27813
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.629954285714286e-06,
+      "loss": 1.3298,
+      "step": 27814
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.629685714285715e-06,
+      "loss": 0.7567,
+      "step": 27815
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.629417142857143e-06,
+      "loss": 0.7916,
+      "step": 27816
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6291485714285715e-06,
+      "loss": 0.9669,
+      "step": 27817
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6288800000000005e-06,
+      "loss": 0.9618,
+      "step": 27818
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.628611428571429e-06,
+      "loss": 1.0915,
+      "step": 27819
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.628342857142858e-06,
+      "loss": 1.3278,
+      "step": 27820
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.628074285714286e-06,
+      "loss": 1.2455,
+      "step": 27821
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.627805714285714e-06,
+      "loss": 0.8623,
+      "step": 27822
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.627537142857143e-06,
+      "loss": 1.2338,
+      "step": 27823
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.627268571428571e-06,
+      "loss": 1.014,
+      "step": 27824
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.627e-06,
+      "loss": 0.8543,
+      "step": 27825
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.626731428571429e-06,
+      "loss": 1.1017,
+      "step": 27826
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6264628571428574e-06,
+      "loss": 0.9184,
+      "step": 27827
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6261942857142865e-06,
+      "loss": 1.1568,
+      "step": 27828
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.625925714285715e-06,
+      "loss": 1.4659,
+      "step": 27829
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.625657142857143e-06,
+      "loss": 1.2636,
+      "step": 27830
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.625388571428572e-06,
+      "loss": 1.2946,
+      "step": 27831
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.62512e-06,
+      "loss": 1.1839,
+      "step": 27832
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.624851428571429e-06,
+      "loss": 1.0539,
+      "step": 27833
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.624582857142857e-06,
+      "loss": 0.931,
+      "step": 27834
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.624314285714285e-06,
+      "loss": 1.0443,
+      "step": 27835
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.624045714285714e-06,
+      "loss": 1.1792,
+      "step": 27836
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.623777142857143e-06,
+      "loss": 0.734,
+      "step": 27837
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.623508571428572e-06,
+      "loss": 0.9575,
+      "step": 27838
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.623240000000001e-06,
+      "loss": 1.2104,
+      "step": 27839
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.622971428571429e-06,
+      "loss": 1.4901,
+      "step": 27840
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.622702857142858e-06,
+      "loss": 0.8749,
+      "step": 27841
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.622434285714286e-06,
+      "loss": 1.1867,
+      "step": 27842
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.622165714285715e-06,
+      "loss": 0.9675,
+      "step": 27843
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.621897142857143e-06,
+      "loss": 1.0737,
+      "step": 27844
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.621628571428571e-06,
+      "loss": 0.9223,
+      "step": 27845
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.62136e-06,
+      "loss": 0.9749,
+      "step": 27846
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.621091428571429e-06,
+      "loss": 1.1829,
+      "step": 27847
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.620822857142857e-06,
+      "loss": 1.4706,
+      "step": 27848
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.620554285714286e-06,
+      "loss": 1.5643,
+      "step": 27849
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.620285714285714e-06,
+      "loss": 1.2907,
+      "step": 27850
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.620017142857143e-06,
+      "loss": 1.1919,
+      "step": 27851
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.619748571428572e-06,
+      "loss": 0.9634,
+      "step": 27852
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.61948e-06,
+      "loss": 0.7809,
+      "step": 27853
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.619211428571429e-06,
+      "loss": 1.1513,
+      "step": 27854
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.618942857142857e-06,
+      "loss": 1.1373,
+      "step": 27855
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.618674285714286e-06,
+      "loss": 1.3773,
+      "step": 27856
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.618405714285715e-06,
+      "loss": 0.8206,
+      "step": 27857
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.618137142857143e-06,
+      "loss": 1.1511,
+      "step": 27858
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.617868571428572e-06,
+      "loss": 1.1658,
+      "step": 27859
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6176e-06,
+      "loss": 0.6602,
+      "step": 27860
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.617331428571429e-06,
+      "loss": 1.3256,
+      "step": 27861
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.617062857142857e-06,
+      "loss": 1.1662,
+      "step": 27862
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.616794285714285e-06,
+      "loss": 0.8744,
+      "step": 27863
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.616525714285714e-06,
+      "loss": 0.9124,
+      "step": 27864
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.616257142857143e-06,
+      "loss": 0.6557,
+      "step": 27865
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.615988571428572e-06,
+      "loss": 0.9636,
+      "step": 27866
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.615720000000001e-06,
+      "loss": 1.3209,
+      "step": 27867
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.615451428571429e-06,
+      "loss": 1.7191,
+      "step": 27868
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.615182857142858e-06,
+      "loss": 1.0349,
+      "step": 27869
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.614914285714286e-06,
+      "loss": 0.5643,
+      "step": 27870
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.614645714285714e-06,
+      "loss": 1.5068,
+      "step": 27871
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.614377142857143e-06,
+      "loss": 1.2203,
+      "step": 27872
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.614108571428571e-06,
+      "loss": 0.8327,
+      "step": 27873
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.61384e-06,
+      "loss": 1.293,
+      "step": 27874
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6135714285714285e-06,
+      "loss": 1.0446,
+      "step": 27875
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.613302857142857e-06,
+      "loss": 1.458,
+      "step": 27876
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.613034285714286e-06,
+      "loss": 0.7379,
+      "step": 27877
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.612765714285715e-06,
+      "loss": 1.1469,
+      "step": 27878
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.612497142857144e-06,
+      "loss": 0.941,
+      "step": 27879
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.612228571428572e-06,
+      "loss": 1.1487,
+      "step": 27880
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.61196e-06,
+      "loss": 1.3948,
+      "step": 27881
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.611691428571429e-06,
+      "loss": 1.5752,
+      "step": 27882
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.611422857142857e-06,
+      "loss": 1.2137,
+      "step": 27883
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.611154285714286e-06,
+      "loss": 1.4163,
+      "step": 27884
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6108857142857145e-06,
+      "loss": 1.2691,
+      "step": 27885
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.610617142857143e-06,
+      "loss": 0.9698,
+      "step": 27886
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.610348571428572e-06,
+      "loss": 0.7693,
+      "step": 27887
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.61008e-06,
+      "loss": 1.5959,
+      "step": 27888
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.609811428571428e-06,
+      "loss": 1.3343,
+      "step": 27889
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.609542857142857e-06,
+      "loss": 1.5978,
+      "step": 27890
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.609274285714286e-06,
+      "loss": 1.0369,
+      "step": 27891
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.609005714285715e-06,
+      "loss": 0.9572,
+      "step": 27892
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.608737142857143e-06,
+      "loss": 0.9571,
+      "step": 27893
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6084685714285715e-06,
+      "loss": 1.0415,
+      "step": 27894
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6082000000000005e-06,
+      "loss": 1.1866,
+      "step": 27895
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.607931428571429e-06,
+      "loss": 0.7356,
+      "step": 27896
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.607662857142858e-06,
+      "loss": 1.3444,
+      "step": 27897
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.607394285714286e-06,
+      "loss": 1.3103,
+      "step": 27898
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.607125714285714e-06,
+      "loss": 1.113,
+      "step": 27899
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.606857142857143e-06,
+      "loss": 1.1702,
+      "step": 27900
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.606588571428571e-06,
+      "loss": 1.0365,
+      "step": 27901
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.60632e-06,
+      "loss": 0.9361,
+      "step": 27902
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6060514285714285e-06,
+      "loss": 1.1475,
+      "step": 27903
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6057828571428575e-06,
+      "loss": 0.9909,
+      "step": 27904
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6055142857142865e-06,
+      "loss": 1.121,
+      "step": 27905
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.605245714285715e-06,
+      "loss": 1.4222,
+      "step": 27906
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.604977142857144e-06,
+      "loss": 1.3103,
+      "step": 27907
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.604708571428572e-06,
+      "loss": 1.2802,
+      "step": 27908
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.60444e-06,
+      "loss": 0.7465,
+      "step": 27909
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.604171428571429e-06,
+      "loss": 1.0323,
+      "step": 27910
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.603902857142857e-06,
+      "loss": 1.1319,
+      "step": 27911
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.6036342857142855e-06,
+      "loss": 1.37,
+      "step": 27912
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.6033657142857145e-06,
+      "loss": 1.3162,
+      "step": 27913
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.603097142857143e-06,
+      "loss": 0.8714,
+      "step": 27914
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.602828571428572e-06,
+      "loss": 1.1418,
+      "step": 27915
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.60256e-06,
+      "loss": 1.184,
+      "step": 27916
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.602291428571429e-06,
+      "loss": 1.3345,
+      "step": 27917
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.602022857142858e-06,
+      "loss": 0.6765,
+      "step": 27918
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.601754285714286e-06,
+      "loss": 0.9039,
+      "step": 27919
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.601485714285715e-06,
+      "loss": 1.4215,
+      "step": 27920
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.601217142857143e-06,
+      "loss": 0.9214,
+      "step": 27921
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.6009485714285714e-06,
+      "loss": 0.9933,
+      "step": 27922
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.6006800000000005e-06,
+      "loss": 0.9927,
+      "step": 27923
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.600411428571429e-06,
+      "loss": 1.0161,
+      "step": 27924
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.600142857142858e-06,
+      "loss": 0.6532,
+      "step": 27925
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.599874285714286e-06,
+      "loss": 1.2279,
+      "step": 27926
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.599605714285714e-06,
+      "loss": 1.4609,
+      "step": 27927
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.599337142857143e-06,
+      "loss": 1.2719,
+      "step": 27928
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.599068571428571e-06,
+      "loss": 0.8301,
+      "step": 27929
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5988e-06,
+      "loss": 1.1196,
+      "step": 27930
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.598531428571429e-06,
+      "loss": 1.2908,
+      "step": 27931
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5982628571428574e-06,
+      "loss": 1.3507,
+      "step": 27932
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5979942857142865e-06,
+      "loss": 1.4305,
+      "step": 27933
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.597725714285715e-06,
+      "loss": 1.3917,
+      "step": 27934
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.597457142857143e-06,
+      "loss": 1.3385,
+      "step": 27935
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.597188571428572e-06,
+      "loss": 1.2603,
+      "step": 27936
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.59692e-06,
+      "loss": 1.1788,
+      "step": 27937
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.596651428571429e-06,
+      "loss": 1.1446,
+      "step": 27938
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.596382857142857e-06,
+      "loss": 1.3955,
+      "step": 27939
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.596114285714285e-06,
+      "loss": 1.172,
+      "step": 27940
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.595845714285714e-06,
+      "loss": 1.3242,
+      "step": 27941
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.595577142857143e-06,
+      "loss": 1.2778,
+      "step": 27942
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.595308571428572e-06,
+      "loss": 1.5541,
+      "step": 27943
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.595040000000001e-06,
+      "loss": 0.912,
+      "step": 27944
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.594771428571429e-06,
+      "loss": 1.3254,
+      "step": 27945
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.594502857142858e-06,
+      "loss": 1.1569,
+      "step": 27946
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.594234285714286e-06,
+      "loss": 1.2719,
+      "step": 27947
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.593965714285715e-06,
+      "loss": 1.3945,
+      "step": 27948
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.593697142857143e-06,
+      "loss": 1.3994,
+      "step": 27949
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.593428571428571e-06,
+      "loss": 0.9683,
+      "step": 27950
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.59316e-06,
+      "loss": 0.8227,
+      "step": 27951
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.592891428571429e-06,
+      "loss": 1.4773,
+      "step": 27952
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.592622857142857e-06,
+      "loss": 1.3138,
+      "step": 27953
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.592354285714286e-06,
+      "loss": 1.1818,
+      "step": 27954
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.592085714285714e-06,
+      "loss": 1.0266,
+      "step": 27955
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.591817142857143e-06,
+      "loss": 1.2406,
+      "step": 27956
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.591548571428572e-06,
+      "loss": 0.8562,
+      "step": 27957
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.59128e-06,
+      "loss": 1.2105,
+      "step": 27958
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.591011428571429e-06,
+      "loss": 1.1339,
+      "step": 27959
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.590742857142857e-06,
+      "loss": 0.825,
+      "step": 27960
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.590474285714286e-06,
+      "loss": 1.2553,
+      "step": 27961
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.590205714285715e-06,
+      "loss": 1.1296,
+      "step": 27962
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.589937142857143e-06,
+      "loss": 1.3479,
+      "step": 27963
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.589668571428572e-06,
+      "loss": 1.1084,
+      "step": 27964
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5894e-06,
+      "loss": 1.0348,
+      "step": 27965
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.589131428571429e-06,
+      "loss": 1.4838,
+      "step": 27966
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.588862857142857e-06,
+      "loss": 1.1439,
+      "step": 27967
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.588594285714285e-06,
+      "loss": 0.5793,
+      "step": 27968
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.588325714285714e-06,
+      "loss": 1.2647,
+      "step": 27969
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.588057142857143e-06,
+      "loss": 1.3556,
+      "step": 27970
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.587788571428572e-06,
+      "loss": 1.3578,
+      "step": 27971
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.587520000000001e-06,
+      "loss": 0.9211,
+      "step": 27972
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.587251428571429e-06,
+      "loss": 1.4194,
+      "step": 27973
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.586982857142858e-06,
+      "loss": 1.2237,
+      "step": 27974
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.586714285714286e-06,
+      "loss": 1.6824,
+      "step": 27975
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.586445714285714e-06,
+      "loss": 1.0189,
+      "step": 27976
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.586177142857143e-06,
+      "loss": 1.1501,
+      "step": 27977
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.585908571428571e-06,
+      "loss": 0.6565,
+      "step": 27978
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.58564e-06,
+      "loss": 1.2354,
+      "step": 27979
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5853714285714285e-06,
+      "loss": 1.1999,
+      "step": 27980
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.585102857142857e-06,
+      "loss": 1.1073,
+      "step": 27981
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.584834285714286e-06,
+      "loss": 1.034,
+      "step": 27982
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.584565714285715e-06,
+      "loss": 0.8216,
+      "step": 27983
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.584297142857144e-06,
+      "loss": 0.8751,
+      "step": 27984
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.584028571428572e-06,
+      "loss": 0.974,
+      "step": 27985
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.58376e-06,
+      "loss": 1.4789,
+      "step": 27986
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.583491428571429e-06,
+      "loss": 1.1042,
+      "step": 27987
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.583222857142857e-06,
+      "loss": 1.0188,
+      "step": 27988
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.582954285714286e-06,
+      "loss": 1.1229,
+      "step": 27989
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5826857142857145e-06,
+      "loss": 1.0697,
+      "step": 27990
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.582417142857143e-06,
+      "loss": 1.0596,
+      "step": 27991
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.582148571428572e-06,
+      "loss": 1.4109,
+      "step": 27992
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.58188e-06,
+      "loss": 1.0725,
+      "step": 27993
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.581611428571429e-06,
+      "loss": 1.2209,
+      "step": 27994
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.581342857142857e-06,
+      "loss": 1.1794,
+      "step": 27995
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.581074285714286e-06,
+      "loss": 1.3917,
+      "step": 27996
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.580805714285715e-06,
+      "loss": 1.3744,
+      "step": 27997
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.580537142857143e-06,
+      "loss": 1.0918,
+      "step": 27998
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5802685714285715e-06,
+      "loss": 1.0929,
+      "step": 27999
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5800000000000005e-06,
+      "loss": 1.4731,
+      "step": 28000
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.579731428571429e-06,
+      "loss": 1.2663,
+      "step": 28001
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.579462857142858e-06,
+      "loss": 1.0887,
+      "step": 28002
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.579194285714286e-06,
+      "loss": 1.305,
+      "step": 28003
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.578925714285714e-06,
+      "loss": 0.7375,
+      "step": 28004
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.578657142857143e-06,
+      "loss": 1.0946,
+      "step": 28005
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.578388571428571e-06,
+      "loss": 1.2656,
+      "step": 28006
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.57812e-06,
+      "loss": 1.3472,
+      "step": 28007
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5778514285714285e-06,
+      "loss": 1.0496,
+      "step": 28008
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5775828571428575e-06,
+      "loss": 1.3073,
+      "step": 28009
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5773142857142865e-06,
+      "loss": 0.9272,
+      "step": 28010
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.577045714285715e-06,
+      "loss": 1.4719,
+      "step": 28011
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.576777142857144e-06,
+      "loss": 1.1285,
+      "step": 28012
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.576508571428572e-06,
+      "loss": 1.0627,
+      "step": 28013
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.57624e-06,
+      "loss": 1.2211,
+      "step": 28014
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.575971428571429e-06,
+      "loss": 0.9914,
+      "step": 28015
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.575702857142857e-06,
+      "loss": 0.7987,
+      "step": 28016
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.575434285714286e-06,
+      "loss": 1.2489,
+      "step": 28017
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5751657142857145e-06,
+      "loss": 1.0819,
+      "step": 28018
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.574897142857143e-06,
+      "loss": 1.102,
+      "step": 28019
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.574628571428572e-06,
+      "loss": 0.8378,
+      "step": 28020
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.57436e-06,
+      "loss": 1.378,
+      "step": 28021
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.574091428571429e-06,
+      "loss": 1.6231,
+      "step": 28022
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.573822857142858e-06,
+      "loss": 1.1405,
+      "step": 28023
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.573554285714286e-06,
+      "loss": 1.1198,
+      "step": 28024
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.573285714285715e-06,
+      "loss": 1.3851,
+      "step": 28025
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.573017142857143e-06,
+      "loss": 1.3314,
+      "step": 28026
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5727485714285714e-06,
+      "loss": 1.3143,
+      "step": 28027
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5724800000000005e-06,
+      "loss": 0.9384,
+      "step": 28028
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.572211428571429e-06,
+      "loss": 1.0455,
+      "step": 28029
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.571942857142858e-06,
+      "loss": 1.1046,
+      "step": 28030
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.571674285714286e-06,
+      "loss": 1.2708,
+      "step": 28031
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.571405714285714e-06,
+      "loss": 1.3044,
+      "step": 28032
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.571137142857143e-06,
+      "loss": 0.8516,
+      "step": 28033
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.570868571428571e-06,
+      "loss": 1.0051,
+      "step": 28034
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5706e-06,
+      "loss": 1.0107,
+      "step": 28035
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.570331428571429e-06,
+      "loss": 1.3856,
+      "step": 28036
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5700628571428574e-06,
+      "loss": 1.1204,
+      "step": 28037
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5697942857142865e-06,
+      "loss": 0.7995,
+      "step": 28038
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.569525714285715e-06,
+      "loss": 1.1393,
+      "step": 28039
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.569257142857143e-06,
+      "loss": 1.3663,
+      "step": 28040
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.568988571428572e-06,
+      "loss": 1.3682,
+      "step": 28041
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.56872e-06,
+      "loss": 1.0635,
+      "step": 28042
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.568451428571429e-06,
+      "loss": 0.9873,
+      "step": 28043
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.568182857142857e-06,
+      "loss": 1.3244,
+      "step": 28044
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.567914285714285e-06,
+      "loss": 0.9365,
+      "step": 28045
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.567645714285714e-06,
+      "loss": 0.9328,
+      "step": 28046
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.567377142857143e-06,
+      "loss": 0.9805,
+      "step": 28047
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.567108571428572e-06,
+      "loss": 0.9699,
+      "step": 28048
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.566840000000001e-06,
+      "loss": 0.9635,
+      "step": 28049
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.566571428571429e-06,
+      "loss": 1.5583,
+      "step": 28050
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.566302857142858e-06,
+      "loss": 0.7811,
+      "step": 28051
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.566034285714286e-06,
+      "loss": 1.211,
+      "step": 28052
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.565765714285715e-06,
+      "loss": 1.0108,
+      "step": 28053
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.565497142857143e-06,
+      "loss": 0.8355,
+      "step": 28054
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.565228571428571e-06,
+      "loss": 0.922,
+      "step": 28055
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.56496e-06,
+      "loss": 1.0497,
+      "step": 28056
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.564691428571429e-06,
+      "loss": 1.3702,
+      "step": 28057
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.564422857142858e-06,
+      "loss": 1.3018,
+      "step": 28058
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.564154285714286e-06,
+      "loss": 1.1626,
+      "step": 28059
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.563885714285714e-06,
+      "loss": 0.893,
+      "step": 28060
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.563617142857143e-06,
+      "loss": 1.2051,
+      "step": 28061
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.563348571428572e-06,
+      "loss": 1.6432,
+      "step": 28062
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.56308e-06,
+      "loss": 1.0851,
+      "step": 28063
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.562811428571429e-06,
+      "loss": 1.3485,
+      "step": 28064
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.562542857142857e-06,
+      "loss": 1.2655,
+      "step": 28065
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.562274285714286e-06,
+      "loss": 1.1227,
+      "step": 28066
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.562005714285715e-06,
+      "loss": 1.2169,
+      "step": 28067
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.561737142857143e-06,
+      "loss": 1.0264,
+      "step": 28068
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.561468571428572e-06,
+      "loss": 0.9738,
+      "step": 28069
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5612e-06,
+      "loss": 0.8421,
+      "step": 28070
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.560931428571429e-06,
+      "loss": 1.2981,
+      "step": 28071
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.560662857142857e-06,
+      "loss": 0.8104,
+      "step": 28072
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.560394285714285e-06,
+      "loss": 1.0601,
+      "step": 28073
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.560125714285714e-06,
+      "loss": 1.4427,
+      "step": 28074
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.559857142857143e-06,
+      "loss": 0.9903,
+      "step": 28075
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.559588571428572e-06,
+      "loss": 1.4988,
+      "step": 28076
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.559320000000001e-06,
+      "loss": 0.9854,
+      "step": 28077
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.559051428571429e-06,
+      "loss": 1.5177,
+      "step": 28078
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.558782857142858e-06,
+      "loss": 0.9689,
+      "step": 28079
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.558514285714286e-06,
+      "loss": 1.4455,
+      "step": 28080
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.558245714285715e-06,
+      "loss": 1.3295,
+      "step": 28081
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.557977142857143e-06,
+      "loss": 0.7925,
+      "step": 28082
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.557708571428571e-06,
+      "loss": 1.5428,
+      "step": 28083
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.55744e-06,
+      "loss": 1.1936,
+      "step": 28084
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.5571714285714285e-06,
+      "loss": 1.3508,
+      "step": 28085
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.556902857142857e-06,
+      "loss": 0.874,
+      "step": 28086
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.556634285714286e-06,
+      "loss": 1.148,
+      "step": 28087
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.556365714285714e-06,
+      "loss": 1.1282,
+      "step": 28088
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.556097142857144e-06,
+      "loss": 0.9458,
+      "step": 28089
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.555828571428572e-06,
+      "loss": 1.3905,
+      "step": 28090
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.55556e-06,
+      "loss": 1.1025,
+      "step": 28091
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.555291428571429e-06,
+      "loss": 1.2489,
+      "step": 28092
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.555022857142857e-06,
+      "loss": 1.0572,
+      "step": 28093
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.554754285714286e-06,
+      "loss": 1.0839,
+      "step": 28094
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5544857142857145e-06,
+      "loss": 0.907,
+      "step": 28095
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.554217142857143e-06,
+      "loss": 1.111,
+      "step": 28096
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.553948571428572e-06,
+      "loss": 1.5302,
+      "step": 28097
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.55368e-06,
+      "loss": 1.1814,
+      "step": 28098
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.553411428571429e-06,
+      "loss": 1.0981,
+      "step": 28099
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.553142857142857e-06,
+      "loss": 0.8497,
+      "step": 28100
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.552874285714285e-06,
+      "loss": 1.1477,
+      "step": 28101
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.552605714285715e-06,
+      "loss": 0.9527,
+      "step": 28102
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.552337142857143e-06,
+      "loss": 1.0562,
+      "step": 28103
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.552068571428572e-06,
+      "loss": 1.3613,
+      "step": 28104
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5518000000000005e-06,
+      "loss": 0.8237,
+      "step": 28105
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.551531428571429e-06,
+      "loss": 0.9232,
+      "step": 28106
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.551262857142858e-06,
+      "loss": 1.3146,
+      "step": 28107
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.550994285714286e-06,
+      "loss": 1.2138,
+      "step": 28108
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.550725714285714e-06,
+      "loss": 0.9708,
+      "step": 28109
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.550457142857143e-06,
+      "loss": 1.379,
+      "step": 28110
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.550188571428571e-06,
+      "loss": 1.4832,
+      "step": 28111
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.54992e-06,
+      "loss": 0.8261,
+      "step": 28112
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5496514285714285e-06,
+      "loss": 0.9543,
+      "step": 28113
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.549382857142857e-06,
+      "loss": 0.9776,
+      "step": 28114
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5491142857142865e-06,
+      "loss": 1.1194,
+      "step": 28115
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.548845714285715e-06,
+      "loss": 1.2464,
+      "step": 28116
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.548577142857144e-06,
+      "loss": 0.8233,
+      "step": 28117
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.548308571428572e-06,
+      "loss": 1.0064,
+      "step": 28118
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.54804e-06,
+      "loss": 1.1196,
+      "step": 28119
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.547771428571429e-06,
+      "loss": 1.2,
+      "step": 28120
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.547502857142857e-06,
+      "loss": 1.3528,
+      "step": 28121
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.547234285714286e-06,
+      "loss": 1.2711,
+      "step": 28122
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5469657142857145e-06,
+      "loss": 1.016,
+      "step": 28123
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.546697142857143e-06,
+      "loss": 1.3994,
+      "step": 28124
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.546428571428572e-06,
+      "loss": 1.1098,
+      "step": 28125
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.54616e-06,
+      "loss": 1.5474,
+      "step": 28126
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.545891428571428e-06,
+      "loss": 1.4807,
+      "step": 28127
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.545622857142858e-06,
+      "loss": 1.1526,
+      "step": 28128
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.545354285714286e-06,
+      "loss": 1.3123,
+      "step": 28129
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.545085714285715e-06,
+      "loss": 0.668,
+      "step": 28130
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.544817142857143e-06,
+      "loss": 0.8652,
+      "step": 28131
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5445485714285714e-06,
+      "loss": 0.9028,
+      "step": 28132
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5442800000000005e-06,
+      "loss": 0.896,
+      "step": 28133
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.544011428571429e-06,
+      "loss": 1.5225,
+      "step": 28134
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.543742857142858e-06,
+      "loss": 1.1323,
+      "step": 28135
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.543474285714286e-06,
+      "loss": 1.0925,
+      "step": 28136
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.543205714285714e-06,
+      "loss": 1.4462,
+      "step": 28137
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.542937142857143e-06,
+      "loss": 1.8266,
+      "step": 28138
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.542668571428571e-06,
+      "loss": 1.2803,
+      "step": 28139
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5424e-06,
+      "loss": 0.8997,
+      "step": 28140
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.542131428571428e-06,
+      "loss": 1.4181,
+      "step": 28141
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5418628571428574e-06,
+      "loss": 0.8964,
+      "step": 28142
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5415942857142865e-06,
+      "loss": 1.3546,
+      "step": 28143
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.541325714285715e-06,
+      "loss": 0.8196,
+      "step": 28144
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.541057142857144e-06,
+      "loss": 0.9087,
+      "step": 28145
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.540788571428572e-06,
+      "loss": 1.4549,
+      "step": 28146
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.54052e-06,
+      "loss": 0.8858,
+      "step": 28147
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.540251428571429e-06,
+      "loss": 0.9454,
+      "step": 28148
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.539982857142857e-06,
+      "loss": 1.4025,
+      "step": 28149
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.539714285714285e-06,
+      "loss": 1.1887,
+      "step": 28150
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.539445714285714e-06,
+      "loss": 1.3469,
+      "step": 28151
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.539177142857143e-06,
+      "loss": 1.1817,
+      "step": 28152
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.538908571428572e-06,
+      "loss": 1.5123,
+      "step": 28153
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.53864e-06,
+      "loss": 1.4116,
+      "step": 28154
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.538371428571429e-06,
+      "loss": 0.9425,
+      "step": 28155
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.538102857142858e-06,
+      "loss": 0.9288,
+      "step": 28156
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.537834285714286e-06,
+      "loss": 1.2817,
+      "step": 28157
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.537565714285715e-06,
+      "loss": 1.1258,
+      "step": 28158
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.537297142857143e-06,
+      "loss": 1.4512,
+      "step": 28159
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.537028571428571e-06,
+      "loss": 0.976,
+      "step": 28160
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.53676e-06,
+      "loss": 1.4818,
+      "step": 28161
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.536491428571429e-06,
+      "loss": 1.1637,
+      "step": 28162
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.536222857142858e-06,
+      "loss": 1.3957,
+      "step": 28163
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.535954285714286e-06,
+      "loss": 1.1488,
+      "step": 28164
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.535685714285714e-06,
+      "loss": 1.1275,
+      "step": 28165
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.535417142857143e-06,
+      "loss": 0.9395,
+      "step": 28166
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.535148571428571e-06,
+      "loss": 1.4799,
+      "step": 28167
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.534880000000001e-06,
+      "loss": 1.2521,
+      "step": 28168
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.534611428571429e-06,
+      "loss": 1.4035,
+      "step": 28169
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.534342857142857e-06,
+      "loss": 1.2557,
+      "step": 28170
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.534074285714286e-06,
+      "loss": 1.9453,
+      "step": 28171
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.533805714285715e-06,
+      "loss": 0.838,
+      "step": 28172
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.533537142857143e-06,
+      "loss": 1.5801,
+      "step": 28173
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.533268571428572e-06,
+      "loss": 1.4233,
+      "step": 28174
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.533e-06,
+      "loss": 1.6407,
+      "step": 28175
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.532731428571429e-06,
+      "loss": 1.625,
+      "step": 28176
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.532462857142857e-06,
+      "loss": 1.1504,
+      "step": 28177
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.532194285714285e-06,
+      "loss": 1.1159,
+      "step": 28178
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.531925714285714e-06,
+      "loss": 1.3333,
+      "step": 28179
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5316571428571425e-06,
+      "loss": 1.2842,
+      "step": 28180
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.531388571428572e-06,
+      "loss": 1.222,
+      "step": 28181
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.531120000000001e-06,
+      "loss": 1.0961,
+      "step": 28182
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.530851428571429e-06,
+      "loss": 1.2803,
+      "step": 28183
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.530582857142858e-06,
+      "loss": 0.9542,
+      "step": 28184
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.530314285714286e-06,
+      "loss": 0.9716,
+      "step": 28185
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.530045714285715e-06,
+      "loss": 1.499,
+      "step": 28186
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.529777142857143e-06,
+      "loss": 1.6406,
+      "step": 28187
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.529508571428571e-06,
+      "loss": 1.048,
+      "step": 28188
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.52924e-06,
+      "loss": 0.9087,
+      "step": 28189
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5289714285714285e-06,
+      "loss": 1.0741,
+      "step": 28190
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5287028571428576e-06,
+      "loss": 1.0932,
+      "step": 28191
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.528434285714286e-06,
+      "loss": 0.8999,
+      "step": 28192
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.528165714285714e-06,
+      "loss": 0.8876,
+      "step": 28193
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.527897142857143e-06,
+      "loss": 1.0032,
+      "step": 28194
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.527628571428572e-06,
+      "loss": 1.1665,
+      "step": 28195
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.52736e-06,
+      "loss": 1.3147,
+      "step": 28196
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.527091428571429e-06,
+      "loss": 1.0284,
+      "step": 28197
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.526822857142857e-06,
+      "loss": 1.4235,
+      "step": 28198
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.526554285714286e-06,
+      "loss": 1.7054,
+      "step": 28199
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5262857142857145e-06,
+      "loss": 1.1872,
+      "step": 28200
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.526017142857143e-06,
+      "loss": 1.4068,
+      "step": 28201
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.525748571428572e-06,
+      "loss": 1.3695,
+      "step": 28202
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.52548e-06,
+      "loss": 1.2305,
+      "step": 28203
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.525211428571429e-06,
+      "loss": 1.1259,
+      "step": 28204
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.524942857142857e-06,
+      "loss": 1.109,
+      "step": 28205
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.524674285714285e-06,
+      "loss": 1.1829,
+      "step": 28206
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.524405714285714e-06,
+      "loss": 1.2972,
+      "step": 28207
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.524137142857143e-06,
+      "loss": 1.0635,
+      "step": 28208
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.523868571428572e-06,
+      "loss": 1.4937,
+      "step": 28209
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5236000000000005e-06,
+      "loss": 1.2717,
+      "step": 28210
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.523331428571429e-06,
+      "loss": 0.7307,
+      "step": 28211
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.523062857142858e-06,
+      "loss": 1.4354,
+      "step": 28212
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.522794285714286e-06,
+      "loss": 1.0059,
+      "step": 28213
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.522525714285714e-06,
+      "loss": 0.7031,
+      "step": 28214
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.522257142857143e-06,
+      "loss": 0.8997,
+      "step": 28215
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.521988571428571e-06,
+      "loss": 0.8795,
+      "step": 28216
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.52172e-06,
+      "loss": 1.3593,
+      "step": 28217
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5214514285714285e-06,
+      "loss": 1.1934,
+      "step": 28218
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.521182857142857e-06,
+      "loss": 1.3299,
+      "step": 28219
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.520914285714286e-06,
+      "loss": 1.0754,
+      "step": 28220
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.520645714285715e-06,
+      "loss": 1.1815,
+      "step": 28221
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.520377142857144e-06,
+      "loss": 0.9717,
+      "step": 28222
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.520108571428572e-06,
+      "loss": 0.8334,
+      "step": 28223
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.51984e-06,
+      "loss": 1.1684,
+      "step": 28224
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.519571428571429e-06,
+      "loss": 1.3658,
+      "step": 28225
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.519302857142857e-06,
+      "loss": 1.2961,
+      "step": 28226
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.519034285714286e-06,
+      "loss": 1.254,
+      "step": 28227
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5187657142857145e-06,
+      "loss": 0.7275,
+      "step": 28228
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.518497142857143e-06,
+      "loss": 0.8331,
+      "step": 28229
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.518228571428572e-06,
+      "loss": 1.1216,
+      "step": 28230
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.51796e-06,
+      "loss": 1.2216,
+      "step": 28231
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.517691428571429e-06,
+      "loss": 1.2891,
+      "step": 28232
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.517422857142857e-06,
+      "loss": 1.2946,
+      "step": 28233
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.517154285714286e-06,
+      "loss": 1.6962,
+      "step": 28234
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.516885714285715e-06,
+      "loss": 1.3311,
+      "step": 28235
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.516617142857143e-06,
+      "loss": 0.7165,
+      "step": 28236
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5163485714285714e-06,
+      "loss": 0.6625,
+      "step": 28237
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5160800000000005e-06,
+      "loss": 0.7664,
+      "step": 28238
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.515811428571429e-06,
+      "loss": 1.1868,
+      "step": 28239
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.515542857142858e-06,
+      "loss": 1.3548,
+      "step": 28240
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.515274285714286e-06,
+      "loss": 1.3365,
+      "step": 28241
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.515005714285714e-06,
+      "loss": 0.3328,
+      "step": 28242
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.514737142857143e-06,
+      "loss": 1.1581,
+      "step": 28243
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.514468571428571e-06,
+      "loss": 1.3352,
+      "step": 28244
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5142e-06,
+      "loss": 0.8582,
+      "step": 28245
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.513931428571428e-06,
+      "loss": 0.8793,
+      "step": 28246
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5136628571428574e-06,
+      "loss": 1.494,
+      "step": 28247
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.5133942857142865e-06,
+      "loss": 0.9314,
+      "step": 28248
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.513125714285715e-06,
+      "loss": 0.8739,
+      "step": 28249
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.512857142857144e-06,
+      "loss": 1.3191,
+      "step": 28250
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.512588571428572e-06,
+      "loss": 1.0215,
+      "step": 28251
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.51232e-06,
+      "loss": 0.8768,
+      "step": 28252
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.512051428571429e-06,
+      "loss": 1.0076,
+      "step": 28253
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.511782857142857e-06,
+      "loss": 0.8973,
+      "step": 28254
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.511514285714286e-06,
+      "loss": 1.1741,
+      "step": 28255
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.511245714285714e-06,
+      "loss": 1.1788,
+      "step": 28256
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.510977142857143e-06,
+      "loss": 0.949,
+      "step": 28257
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.510708571428572e-06,
+      "loss": 1.0655,
+      "step": 28258
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.51044e-06,
+      "loss": 0.8344,
+      "step": 28259
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.510171428571429e-06,
+      "loss": 0.6592,
+      "step": 28260
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.509902857142858e-06,
+      "loss": 1.2529,
+      "step": 28261
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 6.509634285714286e-06,
+      "loss": 1.2824,
+      "step": 28262
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.509365714285715e-06,
+      "loss": 1.3838,
+      "step": 28263
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.509097142857143e-06,
+      "loss": 1.2467,
+      "step": 28264
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.508828571428571e-06,
+      "loss": 1.1991,
+      "step": 28265
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.50856e-06,
+      "loss": 1.1083,
+      "step": 28266
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.508291428571429e-06,
+      "loss": 1.4479,
+      "step": 28267
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.508022857142858e-06,
+      "loss": 1.0111,
+      "step": 28268
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.507754285714286e-06,
+      "loss": 1.372,
+      "step": 28269
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.507485714285714e-06,
+      "loss": 1.2221,
+      "step": 28270
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.507217142857143e-06,
+      "loss": 0.9171,
+      "step": 28271
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.506948571428571e-06,
+      "loss": 0.8263,
+      "step": 28272
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.50668e-06,
+      "loss": 1.4163,
+      "step": 28273
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.506411428571429e-06,
+      "loss": 0.8301,
+      "step": 28274
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.506142857142857e-06,
+      "loss": 1.621,
+      "step": 28275
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.505874285714286e-06,
+      "loss": 0.9631,
+      "step": 28276
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.505605714285715e-06,
+      "loss": 1.2653,
+      "step": 28277
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.505337142857144e-06,
+      "loss": 1.6797,
+      "step": 28278
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.505068571428572e-06,
+      "loss": 1.2205,
+      "step": 28279
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.5048e-06,
+      "loss": 1.0377,
+      "step": 28280
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.504531428571429e-06,
+      "loss": 1.1714,
+      "step": 28281
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.504262857142857e-06,
+      "loss": 1.3332,
+      "step": 28282
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.503994285714285e-06,
+      "loss": 1.0711,
+      "step": 28283
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.503725714285714e-06,
+      "loss": 1.3817,
+      "step": 28284
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.5034571428571425e-06,
+      "loss": 1.276,
+      "step": 28285
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.5031885714285716e-06,
+      "loss": 1.425,
+      "step": 28286
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.502920000000001e-06,
+      "loss": 1.3506,
+      "step": 28287
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.502651428571429e-06,
+      "loss": 1.2328,
+      "step": 28288
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.502382857142858e-06,
+      "loss": 1.3107,
+      "step": 28289
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.502114285714286e-06,
+      "loss": 1.1407,
+      "step": 28290
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.501845714285715e-06,
+      "loss": 1.0225,
+      "step": 28291
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.501577142857143e-06,
+      "loss": 1.3518,
+      "step": 28292
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.501308571428571e-06,
+      "loss": 1.1544,
+      "step": 28293
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.50104e-06,
+      "loss": 1.2837,
+      "step": 28294
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.5007714285714285e-06,
+      "loss": 1.4878,
+      "step": 28295
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.5005028571428576e-06,
+      "loss": 1.2019,
+      "step": 28296
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.500234285714286e-06,
+      "loss": 0.978,
+      "step": 28297
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.499965714285714e-06,
+      "loss": 1.2355,
+      "step": 28298
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.499697142857143e-06,
+      "loss": 1.6333,
+      "step": 28299
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.499428571428572e-06,
+      "loss": 1.2382,
+      "step": 28300
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.49916e-06,
+      "loss": 0.7124,
+      "step": 28301
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.498891428571429e-06,
+      "loss": 0.9311,
+      "step": 28302
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.498622857142857e-06,
+      "loss": 1.4141,
+      "step": 28303
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.498354285714286e-06,
+      "loss": 0.96,
+      "step": 28304
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4980857142857145e-06,
+      "loss": 0.8606,
+      "step": 28305
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.497817142857143e-06,
+      "loss": 1.1983,
+      "step": 28306
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.497548571428572e-06,
+      "loss": 0.9027,
+      "step": 28307
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.49728e-06,
+      "loss": 1.2044,
+      "step": 28308
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.497011428571429e-06,
+      "loss": 1.2856,
+      "step": 28309
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.496742857142857e-06,
+      "loss": 1.417,
+      "step": 28310
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.496474285714285e-06,
+      "loss": 1.0481,
+      "step": 28311
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.496205714285714e-06,
+      "loss": 0.9119,
+      "step": 28312
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.495937142857143e-06,
+      "loss": 1.0179,
+      "step": 28313
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.495668571428572e-06,
+      "loss": 1.3298,
+      "step": 28314
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4954000000000005e-06,
+      "loss": 1.3149,
+      "step": 28315
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.495131428571429e-06,
+      "loss": 1.1335,
+      "step": 28316
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.494862857142858e-06,
+      "loss": 1.1366,
+      "step": 28317
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.494594285714286e-06,
+      "loss": 0.7681,
+      "step": 28318
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.494325714285715e-06,
+      "loss": 1.3355,
+      "step": 28319
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.494057142857143e-06,
+      "loss": 0.9151,
+      "step": 28320
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.493788571428571e-06,
+      "loss": 1.4303,
+      "step": 28321
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.49352e-06,
+      "loss": 1.3349,
+      "step": 28322
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4932514285714285e-06,
+      "loss": 1.1203,
+      "step": 28323
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.492982857142857e-06,
+      "loss": 1.4039,
+      "step": 28324
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.492714285714286e-06,
+      "loss": 0.8491,
+      "step": 28325
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.492445714285715e-06,
+      "loss": 0.868,
+      "step": 28326
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.492177142857144e-06,
+      "loss": 1.0816,
+      "step": 28327
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.491908571428572e-06,
+      "loss": 0.804,
+      "step": 28328
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.49164e-06,
+      "loss": 1.2319,
+      "step": 28329
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.491371428571429e-06,
+      "loss": 1.0414,
+      "step": 28330
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.491102857142857e-06,
+      "loss": 1.052,
+      "step": 28331
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.490834285714286e-06,
+      "loss": 1.1807,
+      "step": 28332
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4905657142857145e-06,
+      "loss": 1.2958,
+      "step": 28333
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.490297142857143e-06,
+      "loss": 1.0726,
+      "step": 28334
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.490028571428572e-06,
+      "loss": 1.0586,
+      "step": 28335
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.48976e-06,
+      "loss": 1.682,
+      "step": 28336
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.489491428571429e-06,
+      "loss": 1.3139,
+      "step": 28337
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.489222857142857e-06,
+      "loss": 0.9539,
+      "step": 28338
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.488954285714286e-06,
+      "loss": 0.8495,
+      "step": 28339
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.488685714285715e-06,
+      "loss": 1.0399,
+      "step": 28340
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.488417142857143e-06,
+      "loss": 1.3697,
+      "step": 28341
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.488148571428572e-06,
+      "loss": 1.3394,
+      "step": 28342
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4878800000000005e-06,
+      "loss": 1.5521,
+      "step": 28343
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.487611428571429e-06,
+      "loss": 1.0485,
+      "step": 28344
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.487342857142858e-06,
+      "loss": 0.7993,
+      "step": 28345
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.487074285714286e-06,
+      "loss": 0.8218,
+      "step": 28346
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.486805714285714e-06,
+      "loss": 1.0172,
+      "step": 28347
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.486537142857143e-06,
+      "loss": 1.2314,
+      "step": 28348
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.486268571428571e-06,
+      "loss": 1.0058,
+      "step": 28349
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.486e-06,
+      "loss": 1.3868,
+      "step": 28350
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.485731428571428e-06,
+      "loss": 1.1628,
+      "step": 28351
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4854628571428574e-06,
+      "loss": 1.0983,
+      "step": 28352
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4851942857142865e-06,
+      "loss": 1.1922,
+      "step": 28353
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.484925714285715e-06,
+      "loss": 1.3914,
+      "step": 28354
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.484657142857144e-06,
+      "loss": 1.023,
+      "step": 28355
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.484388571428572e-06,
+      "loss": 0.9138,
+      "step": 28356
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.48412e-06,
+      "loss": 0.8706,
+      "step": 28357
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.483851428571429e-06,
+      "loss": 1.0676,
+      "step": 28358
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.483582857142857e-06,
+      "loss": 1.1742,
+      "step": 28359
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.483314285714286e-06,
+      "loss": 1.1125,
+      "step": 28360
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.483045714285714e-06,
+      "loss": 0.9118,
+      "step": 28361
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.482777142857143e-06,
+      "loss": 1.4403,
+      "step": 28362
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.482508571428572e-06,
+      "loss": 1.1287,
+      "step": 28363
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.48224e-06,
+      "loss": 1.5395,
+      "step": 28364
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.481971428571429e-06,
+      "loss": 0.988,
+      "step": 28365
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.481702857142858e-06,
+      "loss": 1.0241,
+      "step": 28366
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.481434285714286e-06,
+      "loss": 1.5304,
+      "step": 28367
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.481165714285715e-06,
+      "loss": 1.1189,
+      "step": 28368
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.480897142857143e-06,
+      "loss": 1.1261,
+      "step": 28369
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.480628571428571e-06,
+      "loss": 0.9496,
+      "step": 28370
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.48036e-06,
+      "loss": 1.1827,
+      "step": 28371
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.480091428571429e-06,
+      "loss": 1.1932,
+      "step": 28372
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.479822857142858e-06,
+      "loss": 1.2647,
+      "step": 28373
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.479554285714286e-06,
+      "loss": 1.0872,
+      "step": 28374
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.479285714285714e-06,
+      "loss": 1.1342,
+      "step": 28375
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.479017142857143e-06,
+      "loss": 1.5034,
+      "step": 28376
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.478748571428571e-06,
+      "loss": 1.047,
+      "step": 28377
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.47848e-06,
+      "loss": 1.0049,
+      "step": 28378
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.478211428571429e-06,
+      "loss": 0.9232,
+      "step": 28379
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.477942857142857e-06,
+      "loss": 1.292,
+      "step": 28380
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.477674285714286e-06,
+      "loss": 0.8516,
+      "step": 28381
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.477405714285715e-06,
+      "loss": 1.1272,
+      "step": 28382
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.477137142857144e-06,
+      "loss": 1.0181,
+      "step": 28383
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.476868571428572e-06,
+      "loss": 1.026,
+      "step": 28384
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4766e-06,
+      "loss": 1.0494,
+      "step": 28385
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.476331428571429e-06,
+      "loss": 1.4214,
+      "step": 28386
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.476062857142857e-06,
+      "loss": 1.2488,
+      "step": 28387
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.475794285714285e-06,
+      "loss": 1.3621,
+      "step": 28388
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.475525714285714e-06,
+      "loss": 1.1966,
+      "step": 28389
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4752571428571425e-06,
+      "loss": 1.1688,
+      "step": 28390
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4749885714285716e-06,
+      "loss": 1.3051,
+      "step": 28391
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.474720000000001e-06,
+      "loss": 1.2653,
+      "step": 28392
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.474451428571429e-06,
+      "loss": 1.2379,
+      "step": 28393
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.474182857142858e-06,
+      "loss": 1.1871,
+      "step": 28394
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.473914285714286e-06,
+      "loss": 1.2126,
+      "step": 28395
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.473645714285715e-06,
+      "loss": 1.4188,
+      "step": 28396
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.473377142857143e-06,
+      "loss": 1.3076,
+      "step": 28397
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.473108571428571e-06,
+      "loss": 1.1781,
+      "step": 28398
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.47284e-06,
+      "loss": 1.0536,
+      "step": 28399
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4725714285714285e-06,
+      "loss": 1.0862,
+      "step": 28400
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4723028571428576e-06,
+      "loss": 1.2201,
+      "step": 28401
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.472034285714286e-06,
+      "loss": 1.4435,
+      "step": 28402
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.471765714285714e-06,
+      "loss": 0.9028,
+      "step": 28403
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.471497142857143e-06,
+      "loss": 1.1875,
+      "step": 28404
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.471228571428572e-06,
+      "loss": 0.9539,
+      "step": 28405
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.470960000000001e-06,
+      "loss": 0.9942,
+      "step": 28406
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.470691428571429e-06,
+      "loss": 1.2601,
+      "step": 28407
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.470422857142857e-06,
+      "loss": 1.3448,
+      "step": 28408
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.470154285714286e-06,
+      "loss": 1.1013,
+      "step": 28409
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4698857142857145e-06,
+      "loss": 1.4847,
+      "step": 28410
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.469617142857143e-06,
+      "loss": 1.5113,
+      "step": 28411
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.469348571428572e-06,
+      "loss": 1.0838,
+      "step": 28412
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.46908e-06,
+      "loss": 0.9302,
+      "step": 28413
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.468811428571429e-06,
+      "loss": 1.3357,
+      "step": 28414
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.468542857142857e-06,
+      "loss": 1.1936,
+      "step": 28415
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.468274285714285e-06,
+      "loss": 1.1489,
+      "step": 28416
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.468005714285714e-06,
+      "loss": 1.3264,
+      "step": 28417
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.467737142857143e-06,
+      "loss": 0.9257,
+      "step": 28418
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.467468571428572e-06,
+      "loss": 1.2553,
+      "step": 28419
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4672000000000005e-06,
+      "loss": 0.6127,
+      "step": 28420
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.466931428571429e-06,
+      "loss": 1.6239,
+      "step": 28421
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.466662857142858e-06,
+      "loss": 1.3548,
+      "step": 28422
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.466394285714286e-06,
+      "loss": 1.1742,
+      "step": 28423
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.466125714285715e-06,
+      "loss": 0.9881,
+      "step": 28424
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.465857142857143e-06,
+      "loss": 1.2971,
+      "step": 28425
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.465588571428571e-06,
+      "loss": 1.1599,
+      "step": 28426
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.46532e-06,
+      "loss": 1.3585,
+      "step": 28427
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4650514285714285e-06,
+      "loss": 1.2652,
+      "step": 28428
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.4647828571428575e-06,
+      "loss": 1.1167,
+      "step": 28429
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.464514285714286e-06,
+      "loss": 1.3092,
+      "step": 28430
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.464245714285715e-06,
+      "loss": 1.2293,
+      "step": 28431
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.463977142857144e-06,
+      "loss": 1.2688,
+      "step": 28432
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.463708571428572e-06,
+      "loss": 1.2243,
+      "step": 28433
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.46344e-06,
+      "loss": 1.1783,
+      "step": 28434
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.463171428571429e-06,
+      "loss": 0.6908,
+      "step": 28435
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.462902857142857e-06,
+      "loss": 1.1824,
+      "step": 28436
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 6.462634285714286e-06,
+      "loss": 1.1639,
+      "step": 28437
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4623657142857145e-06,
+      "loss": 0.928,
+      "step": 28438
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.462097142857143e-06,
+      "loss": 1.4174,
+      "step": 28439
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.461828571428572e-06,
+      "loss": 1.1265,
+      "step": 28440
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.46156e-06,
+      "loss": 1.2809,
+      "step": 28441
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.461291428571429e-06,
+      "loss": 0.6655,
+      "step": 28442
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.461022857142857e-06,
+      "loss": 0.7273,
+      "step": 28443
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.460754285714286e-06,
+      "loss": 1.238,
+      "step": 28444
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.460485714285715e-06,
+      "loss": 1.0479,
+      "step": 28445
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.460217142857143e-06,
+      "loss": 1.4636,
+      "step": 28446
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.459948571428572e-06,
+      "loss": 0.9694,
+      "step": 28447
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4596800000000005e-06,
+      "loss": 1.3501,
+      "step": 28448
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.459411428571429e-06,
+      "loss": 0.8931,
+      "step": 28449
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.459142857142858e-06,
+      "loss": 0.9636,
+      "step": 28450
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.458874285714286e-06,
+      "loss": 0.9187,
+      "step": 28451
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.458605714285714e-06,
+      "loss": 1.3107,
+      "step": 28452
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.458337142857143e-06,
+      "loss": 1.2675,
+      "step": 28453
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.458068571428571e-06,
+      "loss": 1.5698,
+      "step": 28454
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4578e-06,
+      "loss": 1.4796,
+      "step": 28455
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.457531428571428e-06,
+      "loss": 1.0124,
+      "step": 28456
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4572628571428574e-06,
+      "loss": 0.923,
+      "step": 28457
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4569942857142865e-06,
+      "loss": 1.3004,
+      "step": 28458
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.456725714285715e-06,
+      "loss": 1.1669,
+      "step": 28459
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.456457142857144e-06,
+      "loss": 0.9061,
+      "step": 28460
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.456188571428572e-06,
+      "loss": 1.2818,
+      "step": 28461
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.45592e-06,
+      "loss": 1.0678,
+      "step": 28462
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.455651428571429e-06,
+      "loss": 1.375,
+      "step": 28463
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.455382857142857e-06,
+      "loss": 1.2976,
+      "step": 28464
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.455114285714286e-06,
+      "loss": 0.7736,
+      "step": 28465
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.454845714285714e-06,
+      "loss": 1.3947,
+      "step": 28466
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.454577142857143e-06,
+      "loss": 1.3183,
+      "step": 28467
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.454308571428572e-06,
+      "loss": 1.4869,
+      "step": 28468
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.45404e-06,
+      "loss": 0.9398,
+      "step": 28469
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.453771428571429e-06,
+      "loss": 0.9266,
+      "step": 28470
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.453502857142858e-06,
+      "loss": 0.836,
+      "step": 28471
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.453234285714286e-06,
+      "loss": 1.203,
+      "step": 28472
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.452965714285715e-06,
+      "loss": 1.1212,
+      "step": 28473
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.452697142857143e-06,
+      "loss": 1.2226,
+      "step": 28474
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.452428571428571e-06,
+      "loss": 1.2549,
+      "step": 28475
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.45216e-06,
+      "loss": 1.1205,
+      "step": 28476
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.451891428571429e-06,
+      "loss": 1.0644,
+      "step": 28477
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.451622857142858e-06,
+      "loss": 1.6886,
+      "step": 28478
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.451354285714286e-06,
+      "loss": 1.4587,
+      "step": 28479
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.451085714285714e-06,
+      "loss": 1.2712,
+      "step": 28480
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.450817142857143e-06,
+      "loss": 1.0473,
+      "step": 28481
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.450548571428571e-06,
+      "loss": 1.2864,
+      "step": 28482
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.45028e-06,
+      "loss": 1.0922,
+      "step": 28483
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.450011428571429e-06,
+      "loss": 1.2393,
+      "step": 28484
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.449742857142857e-06,
+      "loss": 1.1092,
+      "step": 28485
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.449474285714286e-06,
+      "loss": 0.9503,
+      "step": 28486
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.449205714285715e-06,
+      "loss": 1.2003,
+      "step": 28487
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.448937142857144e-06,
+      "loss": 0.9247,
+      "step": 28488
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.448668571428572e-06,
+      "loss": 0.9615,
+      "step": 28489
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4484e-06,
+      "loss": 1.3773,
+      "step": 28490
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.448131428571429e-06,
+      "loss": 0.8999,
+      "step": 28491
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.447862857142857e-06,
+      "loss": 0.7311,
+      "step": 28492
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.447594285714286e-06,
+      "loss": 1.3279,
+      "step": 28493
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.447325714285714e-06,
+      "loss": 1.2797,
+      "step": 28494
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4470571428571425e-06,
+      "loss": 1.597,
+      "step": 28495
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4467885714285716e-06,
+      "loss": 1.356,
+      "step": 28496
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.446520000000001e-06,
+      "loss": 1.4642,
+      "step": 28497
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.446251428571429e-06,
+      "loss": 0.7982,
+      "step": 28498
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.445982857142858e-06,
+      "loss": 1.5533,
+      "step": 28499
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.445714285714286e-06,
+      "loss": 1.1278,
+      "step": 28500
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.445445714285715e-06,
+      "loss": 0.9654,
+      "step": 28501
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.445177142857143e-06,
+      "loss": 1.7163,
+      "step": 28502
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.444908571428571e-06,
+      "loss": 1.0447,
+      "step": 28503
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.44464e-06,
+      "loss": 0.895,
+      "step": 28504
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4443714285714285e-06,
+      "loss": 1.3029,
+      "step": 28505
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4441028571428576e-06,
+      "loss": 1.1565,
+      "step": 28506
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.443834285714286e-06,
+      "loss": 0.8683,
+      "step": 28507
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.443565714285714e-06,
+      "loss": 0.9773,
+      "step": 28508
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.443297142857143e-06,
+      "loss": 1.2648,
+      "step": 28509
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.443028571428572e-06,
+      "loss": 1.1908,
+      "step": 28510
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.442760000000001e-06,
+      "loss": 0.9023,
+      "step": 28511
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.442491428571429e-06,
+      "loss": 1.4883,
+      "step": 28512
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.442222857142857e-06,
+      "loss": 1.0341,
+      "step": 28513
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.441954285714286e-06,
+      "loss": 0.7732,
+      "step": 28514
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4416857142857145e-06,
+      "loss": 1.0042,
+      "step": 28515
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4414171428571435e-06,
+      "loss": 0.8793,
+      "step": 28516
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.441148571428572e-06,
+      "loss": 0.8948,
+      "step": 28517
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.44088e-06,
+      "loss": 0.9402,
+      "step": 28518
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.440611428571429e-06,
+      "loss": 1.0505,
+      "step": 28519
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.440342857142857e-06,
+      "loss": 1.2722,
+      "step": 28520
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.440074285714285e-06,
+      "loss": 1.2001,
+      "step": 28521
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.439805714285714e-06,
+      "loss": 1.1922,
+      "step": 28522
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.439537142857143e-06,
+      "loss": 1.422,
+      "step": 28523
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.439268571428572e-06,
+      "loss": 1.2352,
+      "step": 28524
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4390000000000005e-06,
+      "loss": 1.1894,
+      "step": 28525
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.438731428571429e-06,
+      "loss": 1.3436,
+      "step": 28526
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.438462857142858e-06,
+      "loss": 1.2033,
+      "step": 28527
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.438194285714286e-06,
+      "loss": 1.3565,
+      "step": 28528
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.437925714285715e-06,
+      "loss": 1.3102,
+      "step": 28529
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.437657142857143e-06,
+      "loss": 1.172,
+      "step": 28530
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.437388571428571e-06,
+      "loss": 1.1022,
+      "step": 28531
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.43712e-06,
+      "loss": 1.4055,
+      "step": 28532
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4368514285714285e-06,
+      "loss": 1.3026,
+      "step": 28533
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4365828571428575e-06,
+      "loss": 1.1801,
+      "step": 28534
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.436314285714286e-06,
+      "loss": 0.9871,
+      "step": 28535
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.436045714285714e-06,
+      "loss": 1.3067,
+      "step": 28536
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.435777142857144e-06,
+      "loss": 0.7893,
+      "step": 28537
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.435508571428572e-06,
+      "loss": 1.1752,
+      "step": 28538
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.43524e-06,
+      "loss": 0.8553,
+      "step": 28539
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.434971428571429e-06,
+      "loss": 1.4643,
+      "step": 28540
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.434702857142857e-06,
+      "loss": 1.4085,
+      "step": 28541
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.434434285714286e-06,
+      "loss": 1.1665,
+      "step": 28542
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4341657142857145e-06,
+      "loss": 0.7177,
+      "step": 28543
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.433897142857143e-06,
+      "loss": 1.0562,
+      "step": 28544
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.433628571428572e-06,
+      "loss": 1.3491,
+      "step": 28545
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.43336e-06,
+      "loss": 0.8439,
+      "step": 28546
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.433091428571429e-06,
+      "loss": 1.569,
+      "step": 28547
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.432822857142857e-06,
+      "loss": 1.6853,
+      "step": 28548
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.432554285714285e-06,
+      "loss": 1.1278,
+      "step": 28549
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.432285714285715e-06,
+      "loss": 0.9947,
+      "step": 28550
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.432017142857143e-06,
+      "loss": 1.2559,
+      "step": 28551
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.431748571428572e-06,
+      "loss": 0.8958,
+      "step": 28552
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4314800000000005e-06,
+      "loss": 0.9245,
+      "step": 28553
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.431211428571429e-06,
+      "loss": 1.671,
+      "step": 28554
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.430942857142858e-06,
+      "loss": 1.3022,
+      "step": 28555
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.430674285714286e-06,
+      "loss": 1.6105,
+      "step": 28556
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.430405714285715e-06,
+      "loss": 1.3323,
+      "step": 28557
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.430137142857143e-06,
+      "loss": 1.5027,
+      "step": 28558
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.429868571428571e-06,
+      "loss": 1.0437,
+      "step": 28559
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4296e-06,
+      "loss": 0.7516,
+      "step": 28560
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.429331428571428e-06,
+      "loss": 1.6436,
+      "step": 28561
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.429062857142857e-06,
+      "loss": 1.539,
+      "step": 28562
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4287942857142865e-06,
+      "loss": 1.2366,
+      "step": 28563
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.428525714285715e-06,
+      "loss": 0.9998,
+      "step": 28564
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.428257142857144e-06,
+      "loss": 0.9172,
+      "step": 28565
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.427988571428572e-06,
+      "loss": 1.2207,
+      "step": 28566
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.42772e-06,
+      "loss": 0.9538,
+      "step": 28567
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.427451428571429e-06,
+      "loss": 1.1024,
+      "step": 28568
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.427182857142857e-06,
+      "loss": 1.1352,
+      "step": 28569
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.426914285714286e-06,
+      "loss": 1.2919,
+      "step": 28570
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.426645714285714e-06,
+      "loss": 1.1539,
+      "step": 28571
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.426377142857143e-06,
+      "loss": 1.0343,
+      "step": 28572
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.426108571428572e-06,
+      "loss": 1.3134,
+      "step": 28573
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.42584e-06,
+      "loss": 1.0405,
+      "step": 28574
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.425571428571429e-06,
+      "loss": 1.2088,
+      "step": 28575
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.425302857142858e-06,
+      "loss": 1.4739,
+      "step": 28576
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.425034285714286e-06,
+      "loss": 0.8371,
+      "step": 28577
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.424765714285715e-06,
+      "loss": 0.8715,
+      "step": 28578
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.424497142857143e-06,
+      "loss": 0.8948,
+      "step": 28579
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.424228571428572e-06,
+      "loss": 1.2129,
+      "step": 28580
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.42396e-06,
+      "loss": 1.229,
+      "step": 28581
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.423691428571429e-06,
+      "loss": 1.1999,
+      "step": 28582
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.423422857142858e-06,
+      "loss": 1.4007,
+      "step": 28583
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.423154285714286e-06,
+      "loss": 1.0702,
+      "step": 28584
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.422885714285714e-06,
+      "loss": 0.9244,
+      "step": 28585
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.422617142857143e-06,
+      "loss": 1.4356,
+      "step": 28586
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.422348571428571e-06,
+      "loss": 0.8001,
+      "step": 28587
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.42208e-06,
+      "loss": 0.8849,
+      "step": 28588
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.421811428571428e-06,
+      "loss": 1.0993,
+      "step": 28589
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.421542857142857e-06,
+      "loss": 1.5853,
+      "step": 28590
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.421274285714286e-06,
+      "loss": 1.542,
+      "step": 28591
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.421005714285715e-06,
+      "loss": 1.0606,
+      "step": 28592
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.420737142857144e-06,
+      "loss": 1.0078,
+      "step": 28593
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.420468571428572e-06,
+      "loss": 1.038,
+      "step": 28594
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4202e-06,
+      "loss": 1.2186,
+      "step": 28595
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.419931428571429e-06,
+      "loss": 1.1961,
+      "step": 28596
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.419662857142857e-06,
+      "loss": 0.9342,
+      "step": 28597
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.419394285714286e-06,
+      "loss": 0.871,
+      "step": 28598
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.419125714285714e-06,
+      "loss": 0.8879,
+      "step": 28599
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4188571428571425e-06,
+      "loss": 1.473,
+      "step": 28600
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4185885714285716e-06,
+      "loss": 1.0946,
+      "step": 28601
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.41832e-06,
+      "loss": 1.0942,
+      "step": 28602
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.41805142857143e-06,
+      "loss": 0.8029,
+      "step": 28603
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.417782857142858e-06,
+      "loss": 0.791,
+      "step": 28604
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.417514285714286e-06,
+      "loss": 1.2593,
+      "step": 28605
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.417245714285715e-06,
+      "loss": 1.2636,
+      "step": 28606
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.416977142857143e-06,
+      "loss": 0.8081,
+      "step": 28607
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.416708571428571e-06,
+      "loss": 1.0376,
+      "step": 28608
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.41644e-06,
+      "loss": 1.3936,
+      "step": 28609
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4161714285714285e-06,
+      "loss": 1.1264,
+      "step": 28610
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.4159028571428576e-06,
+      "loss": 1.0449,
+      "step": 28611
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 6.415634285714286e-06,
+      "loss": 1.0128,
+      "step": 28612
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.415365714285714e-06,
+      "loss": 1.332,
+      "step": 28613
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.415097142857143e-06,
+      "loss": 0.9002,
+      "step": 28614
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.414828571428571e-06,
+      "loss": 1.3382,
+      "step": 28615
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.414560000000001e-06,
+      "loss": 1.0864,
+      "step": 28616
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.414291428571429e-06,
+      "loss": 1.2898,
+      "step": 28617
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.414022857142857e-06,
+      "loss": 1.3863,
+      "step": 28618
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.413754285714286e-06,
+      "loss": 1.627,
+      "step": 28619
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4134857142857145e-06,
+      "loss": 1.1669,
+      "step": 28620
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4132171428571435e-06,
+      "loss": 1.2113,
+      "step": 28621
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.412948571428572e-06,
+      "loss": 1.3045,
+      "step": 28622
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.41268e-06,
+      "loss": 1.6454,
+      "step": 28623
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.412411428571429e-06,
+      "loss": 0.8649,
+      "step": 28624
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.412142857142857e-06,
+      "loss": 1.1571,
+      "step": 28625
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.411874285714285e-06,
+      "loss": 1.2696,
+      "step": 28626
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.411605714285714e-06,
+      "loss": 1.0187,
+      "step": 28627
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4113371428571425e-06,
+      "loss": 1.0838,
+      "step": 28628
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.411068571428572e-06,
+      "loss": 1.4934,
+      "step": 28629
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4108000000000005e-06,
+      "loss": 1.1167,
+      "step": 28630
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.410531428571429e-06,
+      "loss": 1.2129,
+      "step": 28631
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.410262857142858e-06,
+      "loss": 1.0507,
+      "step": 28632
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.409994285714286e-06,
+      "loss": 1.0504,
+      "step": 28633
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.409725714285715e-06,
+      "loss": 1.3295,
+      "step": 28634
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.409457142857143e-06,
+      "loss": 0.9622,
+      "step": 28635
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.409188571428571e-06,
+      "loss": 1.5807,
+      "step": 28636
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.40892e-06,
+      "loss": 0.9072,
+      "step": 28637
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4086514285714285e-06,
+      "loss": 0.9623,
+      "step": 28638
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4083828571428575e-06,
+      "loss": 1.507,
+      "step": 28639
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.408114285714286e-06,
+      "loss": 0.8405,
+      "step": 28640
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.407845714285714e-06,
+      "loss": 1.2261,
+      "step": 28641
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.407577142857143e-06,
+      "loss": 1.4015,
+      "step": 28642
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.407308571428572e-06,
+      "loss": 1.259,
+      "step": 28643
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.407040000000001e-06,
+      "loss": 0.9636,
+      "step": 28644
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.406771428571429e-06,
+      "loss": 1.3437,
+      "step": 28645
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.406502857142857e-06,
+      "loss": 1.2013,
+      "step": 28646
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.406234285714286e-06,
+      "loss": 0.9546,
+      "step": 28647
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4059657142857145e-06,
+      "loss": 1.2525,
+      "step": 28648
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.405697142857143e-06,
+      "loss": 0.9149,
+      "step": 28649
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.405428571428572e-06,
+      "loss": 1.1158,
+      "step": 28650
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.40516e-06,
+      "loss": 1.2698,
+      "step": 28651
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.404891428571429e-06,
+      "loss": 1.0341,
+      "step": 28652
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.404622857142857e-06,
+      "loss": 1.1985,
+      "step": 28653
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.404354285714285e-06,
+      "loss": 1.3393,
+      "step": 28654
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.404085714285714e-06,
+      "loss": 1.3484,
+      "step": 28655
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.403817142857143e-06,
+      "loss": 1.0081,
+      "step": 28656
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.403548571428572e-06,
+      "loss": 1.09,
+      "step": 28657
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4032800000000005e-06,
+      "loss": 0.7907,
+      "step": 28658
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.403011428571429e-06,
+      "loss": 1.2766,
+      "step": 28659
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.402742857142858e-06,
+      "loss": 1.0333,
+      "step": 28660
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.402474285714286e-06,
+      "loss": 1.1712,
+      "step": 28661
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.402205714285715e-06,
+      "loss": 1.2412,
+      "step": 28662
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.401937142857143e-06,
+      "loss": 1.0243,
+      "step": 28663
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.401668571428571e-06,
+      "loss": 1.2847,
+      "step": 28664
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4014e-06,
+      "loss": 1.0453,
+      "step": 28665
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.401131428571428e-06,
+      "loss": 0.9897,
+      "step": 28666
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.4008628571428574e-06,
+      "loss": 1.4326,
+      "step": 28667
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.400594285714286e-06,
+      "loss": 0.8558,
+      "step": 28668
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.400325714285715e-06,
+      "loss": 1.3714,
+      "step": 28669
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.400057142857144e-06,
+      "loss": 0.9599,
+      "step": 28670
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.399788571428572e-06,
+      "loss": 1.3687,
+      "step": 28671
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.39952e-06,
+      "loss": 1.4912,
+      "step": 28672
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.399251428571429e-06,
+      "loss": 1.0238,
+      "step": 28673
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.398982857142857e-06,
+      "loss": 0.8492,
+      "step": 28674
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.398714285714286e-06,
+      "loss": 1.6835,
+      "step": 28675
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.398445714285714e-06,
+      "loss": 1.0798,
+      "step": 28676
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.398177142857143e-06,
+      "loss": 1.1224,
+      "step": 28677
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.397908571428572e-06,
+      "loss": 1.2239,
+      "step": 28678
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.39764e-06,
+      "loss": 0.8714,
+      "step": 28679
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.397371428571429e-06,
+      "loss": 1.2572,
+      "step": 28680
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.397102857142857e-06,
+      "loss": 0.8028,
+      "step": 28681
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.396834285714286e-06,
+      "loss": 0.9899,
+      "step": 28682
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.396565714285715e-06,
+      "loss": 1.2583,
+      "step": 28683
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.396297142857143e-06,
+      "loss": 1.2244,
+      "step": 28684
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.396028571428572e-06,
+      "loss": 1.3887,
+      "step": 28685
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.39576e-06,
+      "loss": 0.9949,
+      "step": 28686
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.395491428571429e-06,
+      "loss": 0.8361,
+      "step": 28687
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.395222857142858e-06,
+      "loss": 0.9757,
+      "step": 28688
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.394954285714286e-06,
+      "loss": 1.3237,
+      "step": 28689
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.394685714285715e-06,
+      "loss": 0.7252,
+      "step": 28690
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.394417142857143e-06,
+      "loss": 1.1288,
+      "step": 28691
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.394148571428571e-06,
+      "loss": 1.079,
+      "step": 28692
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.39388e-06,
+      "loss": 1.8548,
+      "step": 28693
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.393611428571428e-06,
+      "loss": 0.8954,
+      "step": 28694
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.393342857142857e-06,
+      "loss": 1.3615,
+      "step": 28695
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.393074285714286e-06,
+      "loss": 0.723,
+      "step": 28696
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.392805714285715e-06,
+      "loss": 0.7149,
+      "step": 28697
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.392537142857144e-06,
+      "loss": 1.07,
+      "step": 28698
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.392268571428572e-06,
+      "loss": 0.7866,
+      "step": 28699
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.392e-06,
+      "loss": 1.0671,
+      "step": 28700
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.391731428571429e-06,
+      "loss": 1.043,
+      "step": 28701
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.391462857142857e-06,
+      "loss": 0.9805,
+      "step": 28702
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.391194285714286e-06,
+      "loss": 1.3689,
+      "step": 28703
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.390925714285714e-06,
+      "loss": 1.1127,
+      "step": 28704
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3906571428571425e-06,
+      "loss": 0.8929,
+      "step": 28705
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3903885714285716e-06,
+      "loss": 1.4194,
+      "step": 28706
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.39012e-06,
+      "loss": 0.9391,
+      "step": 28707
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.389851428571429e-06,
+      "loss": 1.1737,
+      "step": 28708
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.389582857142858e-06,
+      "loss": 1.1405,
+      "step": 28709
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.389314285714286e-06,
+      "loss": 1.2717,
+      "step": 28710
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.389045714285715e-06,
+      "loss": 1.2136,
+      "step": 28711
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.388777142857143e-06,
+      "loss": 1.2255,
+      "step": 28712
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.388508571428571e-06,
+      "loss": 1.0086,
+      "step": 28713
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.38824e-06,
+      "loss": 1.1286,
+      "step": 28714
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3879714285714285e-06,
+      "loss": 0.9074,
+      "step": 28715
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3877028571428576e-06,
+      "loss": 1.1535,
+      "step": 28716
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.387434285714286e-06,
+      "loss": 0.9669,
+      "step": 28717
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.387165714285714e-06,
+      "loss": 1.1444,
+      "step": 28718
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.386897142857143e-06,
+      "loss": 1.236,
+      "step": 28719
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.386628571428571e-06,
+      "loss": 1.1419,
+      "step": 28720
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.38636e-06,
+      "loss": 0.8966,
+      "step": 28721
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.386091428571429e-06,
+      "loss": 0.988,
+      "step": 28722
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.385822857142857e-06,
+      "loss": 0.8385,
+      "step": 28723
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.385554285714286e-06,
+      "loss": 1.3212,
+      "step": 28724
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3852857142857145e-06,
+      "loss": 1.3032,
+      "step": 28725
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3850171428571435e-06,
+      "loss": 1.0725,
+      "step": 28726
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.384748571428572e-06,
+      "loss": 1.119,
+      "step": 28727
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.38448e-06,
+      "loss": 1.4856,
+      "step": 28728
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.384211428571429e-06,
+      "loss": 1.3405,
+      "step": 28729
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.383942857142857e-06,
+      "loss": 0.7573,
+      "step": 28730
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.383674285714286e-06,
+      "loss": 1.3458,
+      "step": 28731
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.383405714285714e-06,
+      "loss": 1.0096,
+      "step": 28732
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3831371428571425e-06,
+      "loss": 1.715,
+      "step": 28733
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3828685714285715e-06,
+      "loss": 1.3136,
+      "step": 28734
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3826000000000005e-06,
+      "loss": 1.0731,
+      "step": 28735
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.382331428571429e-06,
+      "loss": 1.1746,
+      "step": 28736
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.382062857142858e-06,
+      "loss": 1.624,
+      "step": 28737
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.381794285714286e-06,
+      "loss": 1.329,
+      "step": 28738
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.381525714285715e-06,
+      "loss": 1.0325,
+      "step": 28739
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.381257142857143e-06,
+      "loss": 1.0104,
+      "step": 28740
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.380988571428571e-06,
+      "loss": 1.0311,
+      "step": 28741
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.38072e-06,
+      "loss": 0.9175,
+      "step": 28742
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3804514285714285e-06,
+      "loss": 1.1599,
+      "step": 28743
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3801828571428575e-06,
+      "loss": 0.7117,
+      "step": 28744
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.379914285714286e-06,
+      "loss": 1.1993,
+      "step": 28745
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.379645714285714e-06,
+      "loss": 1.5502,
+      "step": 28746
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.379377142857143e-06,
+      "loss": 0.7881,
+      "step": 28747
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.379108571428572e-06,
+      "loss": 1.1942,
+      "step": 28748
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.378840000000001e-06,
+      "loss": 1.0015,
+      "step": 28749
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.378571428571429e-06,
+      "loss": 1.2119,
+      "step": 28750
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.378302857142857e-06,
+      "loss": 0.9792,
+      "step": 28751
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.378034285714286e-06,
+      "loss": 1.083,
+      "step": 28752
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3777657142857145e-06,
+      "loss": 1.0987,
+      "step": 28753
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3774971428571435e-06,
+      "loss": 1.35,
+      "step": 28754
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.377228571428572e-06,
+      "loss": 1.4601,
+      "step": 28755
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.37696e-06,
+      "loss": 1.5815,
+      "step": 28756
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.376691428571429e-06,
+      "loss": 1.0471,
+      "step": 28757
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.376422857142857e-06,
+      "loss": 1.2271,
+      "step": 28758
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.376154285714285e-06,
+      "loss": 1.4215,
+      "step": 28759
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.375885714285714e-06,
+      "loss": 1.3124,
+      "step": 28760
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.375617142857143e-06,
+      "loss": 1.3086,
+      "step": 28761
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.375348571428572e-06,
+      "loss": 0.898,
+      "step": 28762
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3750800000000005e-06,
+      "loss": 1.0713,
+      "step": 28763
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.374811428571429e-06,
+      "loss": 0.8247,
+      "step": 28764
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.374542857142858e-06,
+      "loss": 1.424,
+      "step": 28765
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.374274285714286e-06,
+      "loss": 1.1345,
+      "step": 28766
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.374005714285715e-06,
+      "loss": 0.9957,
+      "step": 28767
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.373737142857143e-06,
+      "loss": 1.2644,
+      "step": 28768
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.373468571428571e-06,
+      "loss": 1.2074,
+      "step": 28769
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3732e-06,
+      "loss": 1.3639,
+      "step": 28770
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.372931428571428e-06,
+      "loss": 1.1092,
+      "step": 28771
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.3726628571428574e-06,
+      "loss": 0.86,
+      "step": 28772
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.372394285714286e-06,
+      "loss": 1.1119,
+      "step": 28773
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.372125714285715e-06,
+      "loss": 1.3493,
+      "step": 28774
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.371857142857144e-06,
+      "loss": 0.9447,
+      "step": 28775
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.371588571428572e-06,
+      "loss": 0.9602,
+      "step": 28776
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.371320000000001e-06,
+      "loss": 1.0308,
+      "step": 28777
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.371051428571429e-06,
+      "loss": 0.9285,
+      "step": 28778
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.370782857142857e-06,
+      "loss": 1.1253,
+      "step": 28779
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.370514285714286e-06,
+      "loss": 1.0961,
+      "step": 28780
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.370245714285714e-06,
+      "loss": 1.5113,
+      "step": 28781
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.369977142857143e-06,
+      "loss": 0.6473,
+      "step": 28782
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.369708571428572e-06,
+      "loss": 1.3171,
+      "step": 28783
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.36944e-06,
+      "loss": 0.9641,
+      "step": 28784
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.369171428571429e-06,
+      "loss": 1.1194,
+      "step": 28785
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.368902857142857e-06,
+      "loss": 1.3327,
+      "step": 28786
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 6.368634285714286e-06,
+      "loss": 0.9604,
+      "step": 28787
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.368365714285715e-06,
+      "loss": 1.0479,
+      "step": 28788
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.368097142857143e-06,
+      "loss": 1.317,
+      "step": 28789
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.367828571428572e-06,
+      "loss": 1.2025,
+      "step": 28790
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.36756e-06,
+      "loss": 1.6863,
+      "step": 28791
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.367291428571429e-06,
+      "loss": 1.0507,
+      "step": 28792
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.367022857142858e-06,
+      "loss": 1.0394,
+      "step": 28793
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.366754285714286e-06,
+      "loss": 0.9219,
+      "step": 28794
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.366485714285715e-06,
+      "loss": 1.2117,
+      "step": 28795
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.366217142857143e-06,
+      "loss": 1.0512,
+      "step": 28796
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.365948571428571e-06,
+      "loss": 0.8193,
+      "step": 28797
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.36568e-06,
+      "loss": 1.262,
+      "step": 28798
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.365411428571428e-06,
+      "loss": 0.9713,
+      "step": 28799
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.365142857142857e-06,
+      "loss": 1.3162,
+      "step": 28800
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.364874285714286e-06,
+      "loss": 1.2878,
+      "step": 28801
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.364605714285715e-06,
+      "loss": 1.2662,
+      "step": 28802
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.364337142857144e-06,
+      "loss": 1.0645,
+      "step": 28803
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.364068571428572e-06,
+      "loss": 0.8464,
+      "step": 28804
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3638e-06,
+      "loss": 1.1736,
+      "step": 28805
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.363531428571429e-06,
+      "loss": 1.5025,
+      "step": 28806
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.363262857142857e-06,
+      "loss": 1.2746,
+      "step": 28807
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.362994285714286e-06,
+      "loss": 1.2373,
+      "step": 28808
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.362725714285714e-06,
+      "loss": 0.981,
+      "step": 28809
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3624571428571425e-06,
+      "loss": 1.4577,
+      "step": 28810
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3621885714285716e-06,
+      "loss": 0.8348,
+      "step": 28811
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.36192e-06,
+      "loss": 1.3913,
+      "step": 28812
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.361651428571429e-06,
+      "loss": 1.1004,
+      "step": 28813
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.361382857142858e-06,
+      "loss": 0.6633,
+      "step": 28814
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.361114285714286e-06,
+      "loss": 1.018,
+      "step": 28815
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.360845714285715e-06,
+      "loss": 0.9374,
+      "step": 28816
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.360577142857143e-06,
+      "loss": 0.9223,
+      "step": 28817
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.360308571428572e-06,
+      "loss": 0.6991,
+      "step": 28818
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.36004e-06,
+      "loss": 1.3395,
+      "step": 28819
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3597714285714285e-06,
+      "loss": 1.0972,
+      "step": 28820
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3595028571428575e-06,
+      "loss": 1.1187,
+      "step": 28821
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.359234285714286e-06,
+      "loss": 0.9269,
+      "step": 28822
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.358965714285714e-06,
+      "loss": 1.02,
+      "step": 28823
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.358697142857143e-06,
+      "loss": 1.2037,
+      "step": 28824
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.358428571428571e-06,
+      "loss": 1.2843,
+      "step": 28825
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.35816e-06,
+      "loss": 1.0055,
+      "step": 28826
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.357891428571429e-06,
+      "loss": 1.09,
+      "step": 28827
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.357622857142857e-06,
+      "loss": 1.1727,
+      "step": 28828
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.357354285714286e-06,
+      "loss": 1.0109,
+      "step": 28829
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3570857142857145e-06,
+      "loss": 1.3845,
+      "step": 28830
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3568171428571435e-06,
+      "loss": 0.9652,
+      "step": 28831
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.356548571428572e-06,
+      "loss": 1.0829,
+      "step": 28832
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.35628e-06,
+      "loss": 1.2747,
+      "step": 28833
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.356011428571429e-06,
+      "loss": 1.0793,
+      "step": 28834
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.355742857142857e-06,
+      "loss": 1.3606,
+      "step": 28835
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.355474285714286e-06,
+      "loss": 1.42,
+      "step": 28836
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.355205714285714e-06,
+      "loss": 1.0209,
+      "step": 28837
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3549371428571425e-06,
+      "loss": 1.1422,
+      "step": 28838
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3546685714285715e-06,
+      "loss": 1.6177,
+      "step": 28839
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3544000000000005e-06,
+      "loss": 1.5043,
+      "step": 28840
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3541314285714295e-06,
+      "loss": 1.4967,
+      "step": 28841
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.353862857142858e-06,
+      "loss": 1.2649,
+      "step": 28842
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.353594285714286e-06,
+      "loss": 1.22,
+      "step": 28843
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.353325714285715e-06,
+      "loss": 0.9303,
+      "step": 28844
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.353057142857143e-06,
+      "loss": 1.2766,
+      "step": 28845
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.352788571428571e-06,
+      "loss": 1.0725,
+      "step": 28846
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.35252e-06,
+      "loss": 0.8908,
+      "step": 28847
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3522514285714285e-06,
+      "loss": 1.0701,
+      "step": 28848
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3519828571428575e-06,
+      "loss": 0.9417,
+      "step": 28849
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.351714285714286e-06,
+      "loss": 1.1216,
+      "step": 28850
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.351445714285714e-06,
+      "loss": 1.4891,
+      "step": 28851
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.351177142857143e-06,
+      "loss": 0.9073,
+      "step": 28852
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.350908571428572e-06,
+      "loss": 0.7427,
+      "step": 28853
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.350640000000001e-06,
+      "loss": 1.0898,
+      "step": 28854
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.350371428571429e-06,
+      "loss": 0.8913,
+      "step": 28855
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.350102857142857e-06,
+      "loss": 0.9803,
+      "step": 28856
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.349834285714286e-06,
+      "loss": 1.4345,
+      "step": 28857
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3495657142857145e-06,
+      "loss": 0.9891,
+      "step": 28858
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3492971428571435e-06,
+      "loss": 1.0475,
+      "step": 28859
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.349028571428572e-06,
+      "loss": 1.014,
+      "step": 28860
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.34876e-06,
+      "loss": 1.2386,
+      "step": 28861
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.348491428571429e-06,
+      "loss": 1.3885,
+      "step": 28862
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.348222857142857e-06,
+      "loss": 0.7106,
+      "step": 28863
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.347954285714286e-06,
+      "loss": 1.0843,
+      "step": 28864
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.347685714285714e-06,
+      "loss": 1.332,
+      "step": 28865
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.347417142857143e-06,
+      "loss": 1.2881,
+      "step": 28866
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.347148571428572e-06,
+      "loss": 0.7669,
+      "step": 28867
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3468800000000005e-06,
+      "loss": 0.9251,
+      "step": 28868
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.346611428571429e-06,
+      "loss": 0.8362,
+      "step": 28869
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.346342857142858e-06,
+      "loss": 1.2247,
+      "step": 28870
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.346074285714286e-06,
+      "loss": 0.886,
+      "step": 28871
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.345805714285715e-06,
+      "loss": 0.7223,
+      "step": 28872
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.345537142857143e-06,
+      "loss": 1.3147,
+      "step": 28873
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.345268571428571e-06,
+      "loss": 1.0966,
+      "step": 28874
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.345e-06,
+      "loss": 0.8244,
+      "step": 28875
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.344731428571428e-06,
+      "loss": 0.7835,
+      "step": 28876
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3444628571428574e-06,
+      "loss": 0.8883,
+      "step": 28877
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.344194285714286e-06,
+      "loss": 0.9387,
+      "step": 28878
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.343925714285715e-06,
+      "loss": 1.139,
+      "step": 28879
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.343657142857144e-06,
+      "loss": 1.2814,
+      "step": 28880
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.343388571428572e-06,
+      "loss": 1.1831,
+      "step": 28881
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.343120000000001e-06,
+      "loss": 1.1595,
+      "step": 28882
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.342851428571429e-06,
+      "loss": 1.5791,
+      "step": 28883
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.342582857142857e-06,
+      "loss": 0.8798,
+      "step": 28884
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.342314285714286e-06,
+      "loss": 1.1269,
+      "step": 28885
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.342045714285714e-06,
+      "loss": 1.1672,
+      "step": 28886
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.341777142857143e-06,
+      "loss": 1.1872,
+      "step": 28887
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.341508571428572e-06,
+      "loss": 1.4363,
+      "step": 28888
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.34124e-06,
+      "loss": 1.2965,
+      "step": 28889
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.340971428571429e-06,
+      "loss": 0.9243,
+      "step": 28890
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.340702857142857e-06,
+      "loss": 1.0204,
+      "step": 28891
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.340434285714286e-06,
+      "loss": 1.0414,
+      "step": 28892
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.340165714285715e-06,
+      "loss": 1.1358,
+      "step": 28893
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.339897142857143e-06,
+      "loss": 0.975,
+      "step": 28894
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.339628571428572e-06,
+      "loss": 1.0415,
+      "step": 28895
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.33936e-06,
+      "loss": 0.8516,
+      "step": 28896
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.339091428571429e-06,
+      "loss": 1.3158,
+      "step": 28897
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.338822857142858e-06,
+      "loss": 1.4999,
+      "step": 28898
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.338554285714286e-06,
+      "loss": 1.0218,
+      "step": 28899
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.338285714285715e-06,
+      "loss": 0.9994,
+      "step": 28900
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.338017142857143e-06,
+      "loss": 0.8251,
+      "step": 28901
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.337748571428571e-06,
+      "loss": 0.8788,
+      "step": 28902
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.33748e-06,
+      "loss": 1.1364,
+      "step": 28903
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.337211428571428e-06,
+      "loss": 1.0136,
+      "step": 28904
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.336942857142857e-06,
+      "loss": 1.516,
+      "step": 28905
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.336674285714286e-06,
+      "loss": 0.7123,
+      "step": 28906
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.336405714285715e-06,
+      "loss": 1.491,
+      "step": 28907
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.336137142857144e-06,
+      "loss": 1.6558,
+      "step": 28908
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.335868571428572e-06,
+      "loss": 1.3425,
+      "step": 28909
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3356e-06,
+      "loss": 1.2004,
+      "step": 28910
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.335331428571429e-06,
+      "loss": 1.6149,
+      "step": 28911
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.335062857142857e-06,
+      "loss": 1.196,
+      "step": 28912
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.334794285714286e-06,
+      "loss": 1.1413,
+      "step": 28913
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.334525714285714e-06,
+      "loss": 1.2509,
+      "step": 28914
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3342571428571425e-06,
+      "loss": 1.1554,
+      "step": 28915
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3339885714285716e-06,
+      "loss": 1.1606,
+      "step": 28916
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.33372e-06,
+      "loss": 1.3924,
+      "step": 28917
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.333451428571429e-06,
+      "loss": 0.8988,
+      "step": 28918
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.333182857142858e-06,
+      "loss": 1.1098,
+      "step": 28919
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.332914285714286e-06,
+      "loss": 0.6895,
+      "step": 28920
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.332645714285715e-06,
+      "loss": 1.4815,
+      "step": 28921
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.332377142857143e-06,
+      "loss": 1.5346,
+      "step": 28922
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.332108571428572e-06,
+      "loss": 1.3451,
+      "step": 28923
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.33184e-06,
+      "loss": 1.3066,
+      "step": 28924
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3315714285714285e-06,
+      "loss": 1.4452,
+      "step": 28925
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3313028571428575e-06,
+      "loss": 0.8829,
+      "step": 28926
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.331034285714286e-06,
+      "loss": 1.3608,
+      "step": 28927
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.330765714285715e-06,
+      "loss": 1.0416,
+      "step": 28928
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.330497142857143e-06,
+      "loss": 1.0457,
+      "step": 28929
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.330228571428571e-06,
+      "loss": 0.5769,
+      "step": 28930
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.32996e-06,
+      "loss": 0.9325,
+      "step": 28931
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.329691428571429e-06,
+      "loss": 1.3573,
+      "step": 28932
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.329422857142857e-06,
+      "loss": 1.0488,
+      "step": 28933
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.329154285714286e-06,
+      "loss": 0.7464,
+      "step": 28934
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3288857142857145e-06,
+      "loss": 1.3004,
+      "step": 28935
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3286171428571435e-06,
+      "loss": 1.0591,
+      "step": 28936
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.328348571428572e-06,
+      "loss": 1.1841,
+      "step": 28937
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.32808e-06,
+      "loss": 0.7428,
+      "step": 28938
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.327811428571429e-06,
+      "loss": 1.203,
+      "step": 28939
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.327542857142857e-06,
+      "loss": 1.209,
+      "step": 28940
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.327274285714286e-06,
+      "loss": 1.1354,
+      "step": 28941
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.327005714285714e-06,
+      "loss": 1.5285,
+      "step": 28942
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3267371428571425e-06,
+      "loss": 1.1559,
+      "step": 28943
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3264685714285715e-06,
+      "loss": 1.0835,
+      "step": 28944
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3262000000000005e-06,
+      "loss": 1.4037,
+      "step": 28945
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3259314285714295e-06,
+      "loss": 1.1108,
+      "step": 28946
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.325662857142858e-06,
+      "loss": 1.3125,
+      "step": 28947
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.325394285714286e-06,
+      "loss": 1.143,
+      "step": 28948
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.325125714285715e-06,
+      "loss": 1.2093,
+      "step": 28949
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.324857142857143e-06,
+      "loss": 1.0513,
+      "step": 28950
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.324588571428571e-06,
+      "loss": 1.1722,
+      "step": 28951
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.32432e-06,
+      "loss": 0.9872,
+      "step": 28952
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3240514285714285e-06,
+      "loss": 0.9084,
+      "step": 28953
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.3237828571428575e-06,
+      "loss": 0.7926,
+      "step": 28954
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.323514285714286e-06,
+      "loss": 1.2445,
+      "step": 28955
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.323245714285714e-06,
+      "loss": 1.0403,
+      "step": 28956
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.322977142857143e-06,
+      "loss": 1.3106,
+      "step": 28957
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.322708571428572e-06,
+      "loss": 1.3497,
+      "step": 28958
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.322440000000001e-06,
+      "loss": 1.129,
+      "step": 28959
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.322171428571429e-06,
+      "loss": 1.2726,
+      "step": 28960
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.321902857142857e-06,
+      "loss": 1.5678,
+      "step": 28961
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 6.321634285714286e-06,
+      "loss": 1.4062,
+      "step": 28962
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3213657142857145e-06,
+      "loss": 1.033,
+      "step": 28963
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3210971428571435e-06,
+      "loss": 0.864,
+      "step": 28964
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.320828571428572e-06,
+      "loss": 1.0433,
+      "step": 28965
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.32056e-06,
+      "loss": 1.4025,
+      "step": 28966
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.320291428571429e-06,
+      "loss": 0.9423,
+      "step": 28967
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.320022857142857e-06,
+      "loss": 1.1115,
+      "step": 28968
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.319754285714286e-06,
+      "loss": 0.9742,
+      "step": 28969
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.319485714285714e-06,
+      "loss": 1.5543,
+      "step": 28970
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.319217142857143e-06,
+      "loss": 0.8611,
+      "step": 28971
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.318948571428572e-06,
+      "loss": 1.4992,
+      "step": 28972
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3186800000000005e-06,
+      "loss": 0.9573,
+      "step": 28973
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.318411428571429e-06,
+      "loss": 0.8043,
+      "step": 28974
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.318142857142858e-06,
+      "loss": 1.415,
+      "step": 28975
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.317874285714286e-06,
+      "loss": 1.1061,
+      "step": 28976
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.317605714285715e-06,
+      "loss": 1.3074,
+      "step": 28977
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.317337142857143e-06,
+      "loss": 1.0402,
+      "step": 28978
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.317068571428571e-06,
+      "loss": 1.2491,
+      "step": 28979
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3168e-06,
+      "loss": 1.5892,
+      "step": 28980
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.316531428571428e-06,
+      "loss": 1.3073,
+      "step": 28981
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3162628571428574e-06,
+      "loss": 1.1289,
+      "step": 28982
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.315994285714286e-06,
+      "loss": 0.817,
+      "step": 28983
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.315725714285715e-06,
+      "loss": 0.872,
+      "step": 28984
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.315457142857144e-06,
+      "loss": 1.4255,
+      "step": 28985
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.315188571428572e-06,
+      "loss": 1.5826,
+      "step": 28986
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.314920000000001e-06,
+      "loss": 1.3154,
+      "step": 28987
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.314651428571429e-06,
+      "loss": 1.3644,
+      "step": 28988
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.314382857142857e-06,
+      "loss": 1.3149,
+      "step": 28989
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.314114285714286e-06,
+      "loss": 1.4643,
+      "step": 28990
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.313845714285714e-06,
+      "loss": 0.973,
+      "step": 28991
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3135771428571434e-06,
+      "loss": 0.9347,
+      "step": 28992
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.313308571428572e-06,
+      "loss": 1.3413,
+      "step": 28993
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.31304e-06,
+      "loss": 1.0817,
+      "step": 28994
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.312771428571429e-06,
+      "loss": 1.1492,
+      "step": 28995
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.312502857142857e-06,
+      "loss": 1.4941,
+      "step": 28996
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.312234285714285e-06,
+      "loss": 1.0576,
+      "step": 28997
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.311965714285715e-06,
+      "loss": 1.345,
+      "step": 28998
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.311697142857143e-06,
+      "loss": 1.0924,
+      "step": 28999
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.311428571428572e-06,
+      "loss": 1.0658,
+      "step": 29000
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.31116e-06,
+      "loss": 1.2733,
+      "step": 29001
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.310891428571429e-06,
+      "loss": 1.3518,
+      "step": 29002
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.310622857142858e-06,
+      "loss": 1.1918,
+      "step": 29003
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.310354285714286e-06,
+      "loss": 1.0883,
+      "step": 29004
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.310085714285715e-06,
+      "loss": 1.3001,
+      "step": 29005
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.309817142857143e-06,
+      "loss": 0.8883,
+      "step": 29006
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.309548571428571e-06,
+      "loss": 1.064,
+      "step": 29007
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.30928e-06,
+      "loss": 1.0911,
+      "step": 29008
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.309011428571428e-06,
+      "loss": 1.2266,
+      "step": 29009
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.308742857142857e-06,
+      "loss": 1.227,
+      "step": 29010
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.308474285714286e-06,
+      "loss": 1.6951,
+      "step": 29011
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.308205714285715e-06,
+      "loss": 0.9356,
+      "step": 29012
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.307937142857144e-06,
+      "loss": 1.3027,
+      "step": 29013
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.307668571428572e-06,
+      "loss": 1.4916,
+      "step": 29014
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.307400000000001e-06,
+      "loss": 1.0864,
+      "step": 29015
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.307131428571429e-06,
+      "loss": 0.8424,
+      "step": 29016
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.306862857142857e-06,
+      "loss": 1.1818,
+      "step": 29017
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.306594285714286e-06,
+      "loss": 1.2975,
+      "step": 29018
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.306325714285714e-06,
+      "loss": 1.2671,
+      "step": 29019
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3060571428571425e-06,
+      "loss": 1.1676,
+      "step": 29020
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3057885714285716e-06,
+      "loss": 1.0113,
+      "step": 29021
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.30552e-06,
+      "loss": 1.3553,
+      "step": 29022
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.305251428571429e-06,
+      "loss": 1.349,
+      "step": 29023
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.304982857142858e-06,
+      "loss": 0.9958,
+      "step": 29024
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.304714285714286e-06,
+      "loss": 1.3101,
+      "step": 29025
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.304445714285715e-06,
+      "loss": 1.3734,
+      "step": 29026
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.304177142857143e-06,
+      "loss": 0.9742,
+      "step": 29027
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.303908571428572e-06,
+      "loss": 1.6665,
+      "step": 29028
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.30364e-06,
+      "loss": 1.5499,
+      "step": 29029
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3033714285714285e-06,
+      "loss": 1.4595,
+      "step": 29030
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3031028571428575e-06,
+      "loss": 1.0677,
+      "step": 29031
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.302834285714286e-06,
+      "loss": 1.5896,
+      "step": 29032
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.302565714285715e-06,
+      "loss": 0.7448,
+      "step": 29033
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.302297142857143e-06,
+      "loss": 1.2623,
+      "step": 29034
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.302028571428571e-06,
+      "loss": 0.9683,
+      "step": 29035
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.30176e-06,
+      "loss": 1.6504,
+      "step": 29036
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.301491428571429e-06,
+      "loss": 1.3939,
+      "step": 29037
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.301222857142857e-06,
+      "loss": 1.278,
+      "step": 29038
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.300954285714286e-06,
+      "loss": 1.0213,
+      "step": 29039
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3006857142857145e-06,
+      "loss": 0.9203,
+      "step": 29040
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.3004171428571435e-06,
+      "loss": 1.0995,
+      "step": 29041
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.300148571428572e-06,
+      "loss": 1.182,
+      "step": 29042
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.29988e-06,
+      "loss": 1.7097,
+      "step": 29043
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.299611428571429e-06,
+      "loss": 0.8261,
+      "step": 29044
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.299342857142857e-06,
+      "loss": 1.3869,
+      "step": 29045
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.299074285714286e-06,
+      "loss": 0.9251,
+      "step": 29046
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.298805714285714e-06,
+      "loss": 1.0544,
+      "step": 29047
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2985371428571425e-06,
+      "loss": 1.0482,
+      "step": 29048
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2982685714285715e-06,
+      "loss": 0.7889,
+      "step": 29049
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.298e-06,
+      "loss": 1.0973,
+      "step": 29050
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2977314285714295e-06,
+      "loss": 1.0175,
+      "step": 29051
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.297462857142858e-06,
+      "loss": 0.9494,
+      "step": 29052
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.297194285714286e-06,
+      "loss": 1.0212,
+      "step": 29053
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.296925714285715e-06,
+      "loss": 1.161,
+      "step": 29054
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.296657142857143e-06,
+      "loss": 1.1263,
+      "step": 29055
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.296388571428572e-06,
+      "loss": 1.2074,
+      "step": 29056
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.29612e-06,
+      "loss": 1.0813,
+      "step": 29057
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2958514285714285e-06,
+      "loss": 1.0173,
+      "step": 29058
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2955828571428575e-06,
+      "loss": 1.0121,
+      "step": 29059
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.295314285714286e-06,
+      "loss": 0.9773,
+      "step": 29060
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.295045714285714e-06,
+      "loss": 1.0434,
+      "step": 29061
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.294777142857143e-06,
+      "loss": 1.1865,
+      "step": 29062
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.294508571428571e-06,
+      "loss": 1.6342,
+      "step": 29063
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.294240000000001e-06,
+      "loss": 1.1428,
+      "step": 29064
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.293971428571429e-06,
+      "loss": 1.365,
+      "step": 29065
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.293702857142857e-06,
+      "loss": 1.1952,
+      "step": 29066
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.293434285714286e-06,
+      "loss": 1.0969,
+      "step": 29067
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2931657142857145e-06,
+      "loss": 0.6467,
+      "step": 29068
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2928971428571435e-06,
+      "loss": 1.1108,
+      "step": 29069
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.292628571428572e-06,
+      "loss": 0.9474,
+      "step": 29070
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.29236e-06,
+      "loss": 1.1874,
+      "step": 29071
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.292091428571429e-06,
+      "loss": 0.9641,
+      "step": 29072
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.291822857142857e-06,
+      "loss": 0.8027,
+      "step": 29073
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.291554285714286e-06,
+      "loss": 1.1202,
+      "step": 29074
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.291285714285714e-06,
+      "loss": 1.4068,
+      "step": 29075
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.291017142857142e-06,
+      "loss": 1.0523,
+      "step": 29076
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.290748571428572e-06,
+      "loss": 1.3479,
+      "step": 29077
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2904800000000005e-06,
+      "loss": 0.9912,
+      "step": 29078
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2902114285714295e-06,
+      "loss": 0.9529,
+      "step": 29079
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.289942857142858e-06,
+      "loss": 1.6274,
+      "step": 29080
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.289674285714286e-06,
+      "loss": 0.756,
+      "step": 29081
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.289405714285715e-06,
+      "loss": 1.2097,
+      "step": 29082
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.289137142857143e-06,
+      "loss": 0.7141,
+      "step": 29083
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.288868571428571e-06,
+      "loss": 1.1091,
+      "step": 29084
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2886e-06,
+      "loss": 1.0338,
+      "step": 29085
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.288331428571428e-06,
+      "loss": 1.1963,
+      "step": 29086
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2880628571428574e-06,
+      "loss": 0.992,
+      "step": 29087
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.287794285714286e-06,
+      "loss": 0.5239,
+      "step": 29088
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.287525714285714e-06,
+      "loss": 0.7105,
+      "step": 29089
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.287257142857144e-06,
+      "loss": 1.182,
+      "step": 29090
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.286988571428572e-06,
+      "loss": 0.9833,
+      "step": 29091
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.286720000000001e-06,
+      "loss": 1.1412,
+      "step": 29092
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.286451428571429e-06,
+      "loss": 1.0233,
+      "step": 29093
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.286182857142857e-06,
+      "loss": 0.8176,
+      "step": 29094
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.285914285714286e-06,
+      "loss": 1.3191,
+      "step": 29095
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.285645714285714e-06,
+      "loss": 0.7412,
+      "step": 29096
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2853771428571434e-06,
+      "loss": 0.7833,
+      "step": 29097
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.285108571428572e-06,
+      "loss": 1.1976,
+      "step": 29098
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.28484e-06,
+      "loss": 1.4067,
+      "step": 29099
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.284571428571429e-06,
+      "loss": 0.9331,
+      "step": 29100
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.284302857142857e-06,
+      "loss": 1.1957,
+      "step": 29101
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.284034285714286e-06,
+      "loss": 1.0355,
+      "step": 29102
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.283765714285714e-06,
+      "loss": 1.0562,
+      "step": 29103
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.283497142857143e-06,
+      "loss": 1.3122,
+      "step": 29104
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.283228571428572e-06,
+      "loss": 0.8612,
+      "step": 29105
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.28296e-06,
+      "loss": 0.9474,
+      "step": 29106
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.282691428571429e-06,
+      "loss": 1.458,
+      "step": 29107
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.282422857142858e-06,
+      "loss": 1.3737,
+      "step": 29108
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.282154285714286e-06,
+      "loss": 1.5172,
+      "step": 29109
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.281885714285715e-06,
+      "loss": 1.1936,
+      "step": 29110
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.281617142857143e-06,
+      "loss": 1.0597,
+      "step": 29111
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.281348571428571e-06,
+      "loss": 1.4054,
+      "step": 29112
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.28108e-06,
+      "loss": 1.2121,
+      "step": 29113
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.280811428571428e-06,
+      "loss": 1.056,
+      "step": 29114
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.280542857142857e-06,
+      "loss": 0.8356,
+      "step": 29115
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2802742857142856e-06,
+      "loss": 1.4043,
+      "step": 29116
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.280005714285715e-06,
+      "loss": 0.8409,
+      "step": 29117
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.279737142857144e-06,
+      "loss": 1.0302,
+      "step": 29118
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.279468571428572e-06,
+      "loss": 1.1549,
+      "step": 29119
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.279200000000001e-06,
+      "loss": 1.1954,
+      "step": 29120
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.278931428571429e-06,
+      "loss": 0.9049,
+      "step": 29121
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.278662857142857e-06,
+      "loss": 1.2161,
+      "step": 29122
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.278394285714286e-06,
+      "loss": 0.784,
+      "step": 29123
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.278125714285714e-06,
+      "loss": 1.0387,
+      "step": 29124
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2778571428571425e-06,
+      "loss": 1.3425,
+      "step": 29125
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2775885714285716e-06,
+      "loss": 1.2766,
+      "step": 29126
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.27732e-06,
+      "loss": 1.1416,
+      "step": 29127
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.277051428571429e-06,
+      "loss": 1.0004,
+      "step": 29128
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.276782857142857e-06,
+      "loss": 0.899,
+      "step": 29129
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.276514285714286e-06,
+      "loss": 0.9235,
+      "step": 29130
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.276245714285715e-06,
+      "loss": 1.0604,
+      "step": 29131
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.275977142857143e-06,
+      "loss": 0.8839,
+      "step": 29132
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.275708571428572e-06,
+      "loss": 1.2416,
+      "step": 29133
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.27544e-06,
+      "loss": 1.3577,
+      "step": 29134
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2751714285714285e-06,
+      "loss": 0.933,
+      "step": 29135
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.2749028571428575e-06,
+      "loss": 0.5087,
+      "step": 29136
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 6.274634285714286e-06,
+      "loss": 0.9832,
+      "step": 29137
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.274365714285715e-06,
+      "loss": 1.4566,
+      "step": 29138
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.274097142857143e-06,
+      "loss": 0.8641,
+      "step": 29139
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.273828571428571e-06,
+      "loss": 1.1328,
+      "step": 29140
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.27356e-06,
+      "loss": 1.1571,
+      "step": 29141
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.273291428571428e-06,
+      "loss": 1.0071,
+      "step": 29142
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.273022857142858e-06,
+      "loss": 1.0165,
+      "step": 29143
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.272754285714286e-06,
+      "loss": 1.3225,
+      "step": 29144
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2724857142857145e-06,
+      "loss": 0.9797,
+      "step": 29145
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2722171428571435e-06,
+      "loss": 0.798,
+      "step": 29146
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.271948571428572e-06,
+      "loss": 1.4607,
+      "step": 29147
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.27168e-06,
+      "loss": 1.3485,
+      "step": 29148
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.271411428571429e-06,
+      "loss": 1.0106,
+      "step": 29149
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.271142857142857e-06,
+      "loss": 1.221,
+      "step": 29150
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.270874285714286e-06,
+      "loss": 0.8706,
+      "step": 29151
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.270605714285714e-06,
+      "loss": 1.1195,
+      "step": 29152
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2703371428571425e-06,
+      "loss": 1.5063,
+      "step": 29153
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2700685714285715e-06,
+      "loss": 0.982,
+      "step": 29154
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2698e-06,
+      "loss": 1.2702,
+      "step": 29155
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.269531428571429e-06,
+      "loss": 1.5544,
+      "step": 29156
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.269262857142858e-06,
+      "loss": 1.2755,
+      "step": 29157
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.268994285714286e-06,
+      "loss": 1.1232,
+      "step": 29158
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.268725714285715e-06,
+      "loss": 1.1458,
+      "step": 29159
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.268457142857143e-06,
+      "loss": 1.4668,
+      "step": 29160
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.268188571428572e-06,
+      "loss": 1.18,
+      "step": 29161
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.26792e-06,
+      "loss": 1.0129,
+      "step": 29162
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2676514285714285e-06,
+      "loss": 1.4868,
+      "step": 29163
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2673828571428575e-06,
+      "loss": 1.5419,
+      "step": 29164
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.267114285714286e-06,
+      "loss": 0.8844,
+      "step": 29165
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.266845714285715e-06,
+      "loss": 1.1308,
+      "step": 29166
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.266577142857143e-06,
+      "loss": 0.7945,
+      "step": 29167
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.266308571428571e-06,
+      "loss": 1.0337,
+      "step": 29168
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.26604e-06,
+      "loss": 1.5008,
+      "step": 29169
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.265771428571429e-06,
+      "loss": 0.9415,
+      "step": 29170
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.265502857142857e-06,
+      "loss": 1.0862,
+      "step": 29171
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.265234285714286e-06,
+      "loss": 1.5642,
+      "step": 29172
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2649657142857145e-06,
+      "loss": 1.2074,
+      "step": 29173
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2646971428571435e-06,
+      "loss": 1.2322,
+      "step": 29174
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.264428571428572e-06,
+      "loss": 1.4194,
+      "step": 29175
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.26416e-06,
+      "loss": 0.8645,
+      "step": 29176
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.263891428571429e-06,
+      "loss": 1.2789,
+      "step": 29177
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.263622857142857e-06,
+      "loss": 0.7331,
+      "step": 29178
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.263354285714286e-06,
+      "loss": 1.2132,
+      "step": 29179
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.263085714285714e-06,
+      "loss": 1.1933,
+      "step": 29180
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.262817142857142e-06,
+      "loss": 1.0131,
+      "step": 29181
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2625485714285714e-06,
+      "loss": 0.7067,
+      "step": 29182
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2622800000000005e-06,
+      "loss": 1.299,
+      "step": 29183
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2620114285714295e-06,
+      "loss": 1.4711,
+      "step": 29184
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.261742857142858e-06,
+      "loss": 1.4381,
+      "step": 29185
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.261474285714286e-06,
+      "loss": 0.9035,
+      "step": 29186
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.261205714285715e-06,
+      "loss": 1.0721,
+      "step": 29187
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.260937142857143e-06,
+      "loss": 0.7439,
+      "step": 29188
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.260668571428572e-06,
+      "loss": 1.1872,
+      "step": 29189
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2604e-06,
+      "loss": 1.3634,
+      "step": 29190
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.260131428571428e-06,
+      "loss": 1.3447,
+      "step": 29191
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2598628571428574e-06,
+      "loss": 0.9961,
+      "step": 29192
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.259594285714286e-06,
+      "loss": 0.7564,
+      "step": 29193
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.259325714285714e-06,
+      "loss": 1.2238,
+      "step": 29194
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.259057142857143e-06,
+      "loss": 1.1172,
+      "step": 29195
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.258788571428572e-06,
+      "loss": 1.1347,
+      "step": 29196
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.258520000000001e-06,
+      "loss": 1.2406,
+      "step": 29197
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.258251428571429e-06,
+      "loss": 1.3474,
+      "step": 29198
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.257982857142857e-06,
+      "loss": 1.4839,
+      "step": 29199
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.257714285714286e-06,
+      "loss": 0.7929,
+      "step": 29200
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.257445714285714e-06,
+      "loss": 1.7934,
+      "step": 29201
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2571771428571434e-06,
+      "loss": 1.2497,
+      "step": 29202
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.256908571428572e-06,
+      "loss": 1.1915,
+      "step": 29203
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.25664e-06,
+      "loss": 1.3675,
+      "step": 29204
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.256371428571429e-06,
+      "loss": 1.0467,
+      "step": 29205
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.256102857142857e-06,
+      "loss": 1.1203,
+      "step": 29206
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.255834285714286e-06,
+      "loss": 1.0495,
+      "step": 29207
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.255565714285714e-06,
+      "loss": 1.0108,
+      "step": 29208
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.255297142857143e-06,
+      "loss": 1.0626,
+      "step": 29209
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.255028571428572e-06,
+      "loss": 1.0549,
+      "step": 29210
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.25476e-06,
+      "loss": 0.8397,
+      "step": 29211
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.254491428571429e-06,
+      "loss": 1.2394,
+      "step": 29212
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.254222857142858e-06,
+      "loss": 1.2877,
+      "step": 29213
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.253954285714286e-06,
+      "loss": 0.857,
+      "step": 29214
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.253685714285715e-06,
+      "loss": 1.1176,
+      "step": 29215
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.253417142857143e-06,
+      "loss": 0.7528,
+      "step": 29216
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.253148571428571e-06,
+      "loss": 1.0477,
+      "step": 29217
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.25288e-06,
+      "loss": 1.3663,
+      "step": 29218
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.252611428571428e-06,
+      "loss": 1.3226,
+      "step": 29219
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.252342857142857e-06,
+      "loss": 1.1407,
+      "step": 29220
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2520742857142856e-06,
+      "loss": 1.4134,
+      "step": 29221
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2518057142857146e-06,
+      "loss": 0.9617,
+      "step": 29222
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.251537142857144e-06,
+      "loss": 1.0684,
+      "step": 29223
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.251268571428572e-06,
+      "loss": 1.125,
+      "step": 29224
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.251000000000001e-06,
+      "loss": 1.1289,
+      "step": 29225
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.250731428571429e-06,
+      "loss": 1.0292,
+      "step": 29226
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.250462857142857e-06,
+      "loss": 1.3733,
+      "step": 29227
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.250194285714286e-06,
+      "loss": 1.0055,
+      "step": 29228
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.249925714285714e-06,
+      "loss": 1.5135,
+      "step": 29229
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.249657142857143e-06,
+      "loss": 1.1067,
+      "step": 29230
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2493885714285715e-06,
+      "loss": 1.4658,
+      "step": 29231
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.24912e-06,
+      "loss": 0.7879,
+      "step": 29232
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.248851428571429e-06,
+      "loss": 1.3341,
+      "step": 29233
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.248582857142857e-06,
+      "loss": 1.0896,
+      "step": 29234
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.248314285714286e-06,
+      "loss": 1.4105,
+      "step": 29235
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.248045714285715e-06,
+      "loss": 1.1656,
+      "step": 29236
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.247777142857143e-06,
+      "loss": 1.0099,
+      "step": 29237
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.247508571428572e-06,
+      "loss": 1.0275,
+      "step": 29238
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.24724e-06,
+      "loss": 0.7458,
+      "step": 29239
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2469714285714285e-06,
+      "loss": 1.3827,
+      "step": 29240
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2467028571428575e-06,
+      "loss": 1.2686,
+      "step": 29241
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.246434285714286e-06,
+      "loss": 1.219,
+      "step": 29242
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.246165714285715e-06,
+      "loss": 1.1497,
+      "step": 29243
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.245897142857143e-06,
+      "loss": 1.052,
+      "step": 29244
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.245628571428571e-06,
+      "loss": 0.9766,
+      "step": 29245
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.24536e-06,
+      "loss": 1.097,
+      "step": 29246
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.245091428571428e-06,
+      "loss": 1.5086,
+      "step": 29247
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.244822857142857e-06,
+      "loss": 0.9541,
+      "step": 29248
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.244554285714286e-06,
+      "loss": 1.5993,
+      "step": 29249
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2442857142857145e-06,
+      "loss": 1.0979,
+      "step": 29250
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2440171428571435e-06,
+      "loss": 0.9463,
+      "step": 29251
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.243748571428572e-06,
+      "loss": 1.0188,
+      "step": 29252
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.243480000000001e-06,
+      "loss": 0.8939,
+      "step": 29253
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.243211428571429e-06,
+      "loss": 0.9626,
+      "step": 29254
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.242942857142857e-06,
+      "loss": 1.3732,
+      "step": 29255
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.242674285714286e-06,
+      "loss": 0.8327,
+      "step": 29256
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.242405714285714e-06,
+      "loss": 1.6028,
+      "step": 29257
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2421371428571425e-06,
+      "loss": 0.9549,
+      "step": 29258
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2418685714285715e-06,
+      "loss": 1.5392,
+      "step": 29259
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2416e-06,
+      "loss": 1.2367,
+      "step": 29260
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.241331428571429e-06,
+      "loss": 1.3056,
+      "step": 29261
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.241062857142858e-06,
+      "loss": 0.9816,
+      "step": 29262
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.240794285714286e-06,
+      "loss": 1.0461,
+      "step": 29263
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.240525714285715e-06,
+      "loss": 0.8631,
+      "step": 29264
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.240257142857143e-06,
+      "loss": 1.3203,
+      "step": 29265
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.239988571428572e-06,
+      "loss": 1.0413,
+      "step": 29266
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.23972e-06,
+      "loss": 1.1399,
+      "step": 29267
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2394514285714285e-06,
+      "loss": 1.2702,
+      "step": 29268
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2391828571428575e-06,
+      "loss": 0.9499,
+      "step": 29269
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.238914285714286e-06,
+      "loss": 1.182,
+      "step": 29270
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.238645714285715e-06,
+      "loss": 1.1781,
+      "step": 29271
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.238377142857143e-06,
+      "loss": 0.9794,
+      "step": 29272
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.238108571428571e-06,
+      "loss": 1.6456,
+      "step": 29273
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.23784e-06,
+      "loss": 0.5898,
+      "step": 29274
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.237571428571429e-06,
+      "loss": 0.9648,
+      "step": 29275
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.237302857142858e-06,
+      "loss": 1.222,
+      "step": 29276
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.237034285714286e-06,
+      "loss": 1.0367,
+      "step": 29277
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2367657142857145e-06,
+      "loss": 1.1796,
+      "step": 29278
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2364971428571435e-06,
+      "loss": 1.0336,
+      "step": 29279
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.236228571428572e-06,
+      "loss": 1.1664,
+      "step": 29280
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.23596e-06,
+      "loss": 0.9755,
+      "step": 29281
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.235691428571429e-06,
+      "loss": 1.4418,
+      "step": 29282
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.235422857142857e-06,
+      "loss": 1.0304,
+      "step": 29283
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.235154285714286e-06,
+      "loss": 1.048,
+      "step": 29284
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.234885714285714e-06,
+      "loss": 1.0487,
+      "step": 29285
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.234617142857142e-06,
+      "loss": 1.2748,
+      "step": 29286
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2343485714285714e-06,
+      "loss": 0.9178,
+      "step": 29287
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2340800000000005e-06,
+      "loss": 1.0488,
+      "step": 29288
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2338114285714295e-06,
+      "loss": 1.4989,
+      "step": 29289
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.233542857142858e-06,
+      "loss": 0.7844,
+      "step": 29290
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.233274285714286e-06,
+      "loss": 0.8122,
+      "step": 29291
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.233005714285715e-06,
+      "loss": 1.1219,
+      "step": 29292
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.232737142857143e-06,
+      "loss": 1.1536,
+      "step": 29293
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.232468571428572e-06,
+      "loss": 1.1936,
+      "step": 29294
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2322e-06,
+      "loss": 1.0556,
+      "step": 29295
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.231931428571428e-06,
+      "loss": 1.2031,
+      "step": 29296
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2316628571428574e-06,
+      "loss": 1.3552,
+      "step": 29297
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.231394285714286e-06,
+      "loss": 1.1015,
+      "step": 29298
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.231125714285714e-06,
+      "loss": 1.096,
+      "step": 29299
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.230857142857143e-06,
+      "loss": 1.0173,
+      "step": 29300
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.230588571428572e-06,
+      "loss": 1.5786,
+      "step": 29301
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.230320000000001e-06,
+      "loss": 1.2114,
+      "step": 29302
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.230051428571429e-06,
+      "loss": 1.2654,
+      "step": 29303
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.229782857142857e-06,
+      "loss": 1.4058,
+      "step": 29304
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.229514285714286e-06,
+      "loss": 1.0866,
+      "step": 29305
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.229245714285714e-06,
+      "loss": 1.3764,
+      "step": 29306
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.2289771428571434e-06,
+      "loss": 1.2273,
+      "step": 29307
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.228708571428572e-06,
+      "loss": 1.0749,
+      "step": 29308
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.22844e-06,
+      "loss": 1.1201,
+      "step": 29309
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.228171428571429e-06,
+      "loss": 0.9581,
+      "step": 29310
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.227902857142857e-06,
+      "loss": 1.1105,
+      "step": 29311
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 6.227634285714286e-06,
+      "loss": 1.539,
+      "step": 29312
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.227365714285714e-06,
+      "loss": 1.2536,
+      "step": 29313
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.227097142857143e-06,
+      "loss": 1.2744,
+      "step": 29314
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.226828571428572e-06,
+      "loss": 1.1807,
+      "step": 29315
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.22656e-06,
+      "loss": 1.3626,
+      "step": 29316
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2262914285714294e-06,
+      "loss": 1.5182,
+      "step": 29317
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.226022857142858e-06,
+      "loss": 1.1953,
+      "step": 29318
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.225754285714286e-06,
+      "loss": 0.8029,
+      "step": 29319
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.225485714285715e-06,
+      "loss": 1.1647,
+      "step": 29320
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.225217142857143e-06,
+      "loss": 1.0495,
+      "step": 29321
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.224948571428571e-06,
+      "loss": 1.4917,
+      "step": 29322
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.22468e-06,
+      "loss": 1.0788,
+      "step": 29323
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.224411428571428e-06,
+      "loss": 1.2283,
+      "step": 29324
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.224142857142857e-06,
+      "loss": 1.0067,
+      "step": 29325
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2238742857142856e-06,
+      "loss": 0.9352,
+      "step": 29326
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2236057142857146e-06,
+      "loss": 0.6879,
+      "step": 29327
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.223337142857144e-06,
+      "loss": 1.3988,
+      "step": 29328
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.223068571428572e-06,
+      "loss": 1.035,
+      "step": 29329
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.222800000000001e-06,
+      "loss": 1.2391,
+      "step": 29330
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.222531428571429e-06,
+      "loss": 0.9779,
+      "step": 29331
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.222262857142857e-06,
+      "loss": 0.9925,
+      "step": 29332
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.221994285714286e-06,
+      "loss": 1.5664,
+      "step": 29333
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.221725714285714e-06,
+      "loss": 1.2017,
+      "step": 29334
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.221457142857143e-06,
+      "loss": 0.5885,
+      "step": 29335
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2211885714285715e-06,
+      "loss": 1.1558,
+      "step": 29336
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.22092e-06,
+      "loss": 1.5018,
+      "step": 29337
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.220651428571429e-06,
+      "loss": 1.0656,
+      "step": 29338
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.220382857142857e-06,
+      "loss": 0.8172,
+      "step": 29339
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.220114285714286e-06,
+      "loss": 1.2085,
+      "step": 29340
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.219845714285715e-06,
+      "loss": 0.7407,
+      "step": 29341
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.219577142857143e-06,
+      "loss": 1.0465,
+      "step": 29342
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.219308571428572e-06,
+      "loss": 1.0408,
+      "step": 29343
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.21904e-06,
+      "loss": 0.8931,
+      "step": 29344
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2187714285714285e-06,
+      "loss": 0.9654,
+      "step": 29345
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2185028571428575e-06,
+      "loss": 1.0614,
+      "step": 29346
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.218234285714286e-06,
+      "loss": 0.8263,
+      "step": 29347
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.217965714285715e-06,
+      "loss": 0.9551,
+      "step": 29348
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.217697142857143e-06,
+      "loss": 1.2353,
+      "step": 29349
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.217428571428571e-06,
+      "loss": 1.1099,
+      "step": 29350
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.21716e-06,
+      "loss": 1.3109,
+      "step": 29351
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.216891428571428e-06,
+      "loss": 1.4844,
+      "step": 29352
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.216622857142857e-06,
+      "loss": 1.346,
+      "step": 29353
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.216354285714286e-06,
+      "loss": 0.8786,
+      "step": 29354
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2160857142857145e-06,
+      "loss": 1.3051,
+      "step": 29355
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2158171428571435e-06,
+      "loss": 0.7713,
+      "step": 29356
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.215548571428572e-06,
+      "loss": 1.142,
+      "step": 29357
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.215280000000001e-06,
+      "loss": 1.1602,
+      "step": 29358
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.215011428571429e-06,
+      "loss": 1.6162,
+      "step": 29359
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.214742857142857e-06,
+      "loss": 1.1486,
+      "step": 29360
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.214474285714286e-06,
+      "loss": 1.2988,
+      "step": 29361
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.214205714285714e-06,
+      "loss": 1.1375,
+      "step": 29362
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.213937142857143e-06,
+      "loss": 1.2899,
+      "step": 29363
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2136685714285715e-06,
+      "loss": 1.3909,
+      "step": 29364
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2134e-06,
+      "loss": 0.9322,
+      "step": 29365
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.213131428571429e-06,
+      "loss": 1.4986,
+      "step": 29366
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.212862857142858e-06,
+      "loss": 1.3059,
+      "step": 29367
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.212594285714286e-06,
+      "loss": 1.1526,
+      "step": 29368
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.212325714285715e-06,
+      "loss": 1.3382,
+      "step": 29369
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.212057142857143e-06,
+      "loss": 0.8896,
+      "step": 29370
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.211788571428572e-06,
+      "loss": 1.079,
+      "step": 29371
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.21152e-06,
+      "loss": 1.2977,
+      "step": 29372
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2112514285714285e-06,
+      "loss": 1.2773,
+      "step": 29373
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2109828571428575e-06,
+      "loss": 1.2292,
+      "step": 29374
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.210714285714286e-06,
+      "loss": 0.9773,
+      "step": 29375
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.210445714285715e-06,
+      "loss": 1.0393,
+      "step": 29376
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.210177142857143e-06,
+      "loss": 1.1362,
+      "step": 29377
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.209908571428571e-06,
+      "loss": 0.9681,
+      "step": 29378
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.20964e-06,
+      "loss": 1.3041,
+      "step": 29379
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.209371428571429e-06,
+      "loss": 0.6716,
+      "step": 29380
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.209102857142858e-06,
+      "loss": 1.5494,
+      "step": 29381
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.208834285714286e-06,
+      "loss": 0.6207,
+      "step": 29382
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2085657142857145e-06,
+      "loss": 0.977,
+      "step": 29383
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2082971428571435e-06,
+      "loss": 1.0631,
+      "step": 29384
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.208028571428572e-06,
+      "loss": 1.1613,
+      "step": 29385
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.20776e-06,
+      "loss": 0.9519,
+      "step": 29386
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.207491428571429e-06,
+      "loss": 1.1186,
+      "step": 29387
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.207222857142857e-06,
+      "loss": 0.9478,
+      "step": 29388
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.206954285714286e-06,
+      "loss": 1.6451,
+      "step": 29389
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.206685714285714e-06,
+      "loss": 1.2413,
+      "step": 29390
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.206417142857142e-06,
+      "loss": 1.3117,
+      "step": 29391
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2061485714285714e-06,
+      "loss": 1.2529,
+      "step": 29392
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2058800000000005e-06,
+      "loss": 1.1004,
+      "step": 29393
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2056114285714295e-06,
+      "loss": 1.4455,
+      "step": 29394
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.205342857142858e-06,
+      "loss": 1.3568,
+      "step": 29395
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.205074285714286e-06,
+      "loss": 1.2252,
+      "step": 29396
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.204805714285715e-06,
+      "loss": 1.3921,
+      "step": 29397
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.204537142857143e-06,
+      "loss": 0.9745,
+      "step": 29398
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.204268571428572e-06,
+      "loss": 1.2959,
+      "step": 29399
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.204e-06,
+      "loss": 0.8463,
+      "step": 29400
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.203731428571428e-06,
+      "loss": 0.9043,
+      "step": 29401
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2034628571428574e-06,
+      "loss": 1.5607,
+      "step": 29402
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.203194285714286e-06,
+      "loss": 1.3171,
+      "step": 29403
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.202925714285715e-06,
+      "loss": 0.9123,
+      "step": 29404
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.202657142857143e-06,
+      "loss": 0.633,
+      "step": 29405
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.202388571428572e-06,
+      "loss": 1.0483,
+      "step": 29406
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.202120000000001e-06,
+      "loss": 1.2569,
+      "step": 29407
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.201851428571429e-06,
+      "loss": 0.922,
+      "step": 29408
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.201582857142857e-06,
+      "loss": 1.1873,
+      "step": 29409
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.201314285714286e-06,
+      "loss": 0.9362,
+      "step": 29410
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.201045714285714e-06,
+      "loss": 1.4683,
+      "step": 29411
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.2007771428571434e-06,
+      "loss": 1.119,
+      "step": 29412
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.200508571428572e-06,
+      "loss": 1.8498,
+      "step": 29413
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.20024e-06,
+      "loss": 1.2409,
+      "step": 29414
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.199971428571429e-06,
+      "loss": 1.4766,
+      "step": 29415
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.199702857142857e-06,
+      "loss": 1.3884,
+      "step": 29416
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.199434285714286e-06,
+      "loss": 0.9441,
+      "step": 29417
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.199165714285714e-06,
+      "loss": 1.2753,
+      "step": 29418
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.198897142857143e-06,
+      "loss": 1.0537,
+      "step": 29419
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.198628571428572e-06,
+      "loss": 1.2353,
+      "step": 29420
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.19836e-06,
+      "loss": 1.1941,
+      "step": 29421
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1980914285714294e-06,
+      "loss": 1.0731,
+      "step": 29422
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.197822857142858e-06,
+      "loss": 0.9324,
+      "step": 29423
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.197554285714286e-06,
+      "loss": 1.0619,
+      "step": 29424
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.197285714285715e-06,
+      "loss": 0.7811,
+      "step": 29425
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.197017142857143e-06,
+      "loss": 1.1139,
+      "step": 29426
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.196748571428572e-06,
+      "loss": 1.2842,
+      "step": 29427
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.19648e-06,
+      "loss": 1.0808,
+      "step": 29428
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.196211428571428e-06,
+      "loss": 1.6966,
+      "step": 29429
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.195942857142857e-06,
+      "loss": 0.8466,
+      "step": 29430
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1956742857142856e-06,
+      "loss": 1.2938,
+      "step": 29431
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1954057142857146e-06,
+      "loss": 1.1485,
+      "step": 29432
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.195137142857144e-06,
+      "loss": 0.8075,
+      "step": 29433
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.194868571428572e-06,
+      "loss": 0.8187,
+      "step": 29434
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.194600000000001e-06,
+      "loss": 1.3953,
+      "step": 29435
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.194331428571429e-06,
+      "loss": 1.0328,
+      "step": 29436
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.194062857142857e-06,
+      "loss": 1.0991,
+      "step": 29437
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.193794285714286e-06,
+      "loss": 1.7857,
+      "step": 29438
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.193525714285714e-06,
+      "loss": 1.1651,
+      "step": 29439
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.193257142857143e-06,
+      "loss": 1.0735,
+      "step": 29440
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1929885714285715e-06,
+      "loss": 0.9697,
+      "step": 29441
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.19272e-06,
+      "loss": 0.9005,
+      "step": 29442
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.192451428571429e-06,
+      "loss": 1.3131,
+      "step": 29443
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.192182857142857e-06,
+      "loss": 1.364,
+      "step": 29444
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.191914285714286e-06,
+      "loss": 0.7727,
+      "step": 29445
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.191645714285715e-06,
+      "loss": 1.1155,
+      "step": 29446
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.191377142857143e-06,
+      "loss": 1.2194,
+      "step": 29447
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.191108571428572e-06,
+      "loss": 1.3371,
+      "step": 29448
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.19084e-06,
+      "loss": 0.9669,
+      "step": 29449
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1905714285714285e-06,
+      "loss": 1.1255,
+      "step": 29450
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1903028571428575e-06,
+      "loss": 1.1686,
+      "step": 29451
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.190034285714286e-06,
+      "loss": 1.2763,
+      "step": 29452
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.189765714285715e-06,
+      "loss": 1.3435,
+      "step": 29453
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.189497142857143e-06,
+      "loss": 1.0466,
+      "step": 29454
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.189228571428571e-06,
+      "loss": 1.8008,
+      "step": 29455
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.18896e-06,
+      "loss": 0.9015,
+      "step": 29456
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.188691428571428e-06,
+      "loss": 0.9975,
+      "step": 29457
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.188422857142857e-06,
+      "loss": 1.1021,
+      "step": 29458
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.188154285714286e-06,
+      "loss": 0.9233,
+      "step": 29459
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1878857142857145e-06,
+      "loss": 1.0661,
+      "step": 29460
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1876171428571435e-06,
+      "loss": 1.052,
+      "step": 29461
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.187348571428572e-06,
+      "loss": 0.7297,
+      "step": 29462
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.187080000000001e-06,
+      "loss": 1.2705,
+      "step": 29463
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.186811428571429e-06,
+      "loss": 0.9456,
+      "step": 29464
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.186542857142857e-06,
+      "loss": 0.7786,
+      "step": 29465
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.186274285714286e-06,
+      "loss": 1.4647,
+      "step": 29466
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.186005714285714e-06,
+      "loss": 1.2282,
+      "step": 29467
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.185737142857143e-06,
+      "loss": 0.9972,
+      "step": 29468
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1854685714285715e-06,
+      "loss": 1.1642,
+      "step": 29469
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1852e-06,
+      "loss": 0.9593,
+      "step": 29470
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.184931428571429e-06,
+      "loss": 1.0459,
+      "step": 29471
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.184662857142858e-06,
+      "loss": 0.8414,
+      "step": 29472
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.184394285714286e-06,
+      "loss": 1.196,
+      "step": 29473
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.184125714285715e-06,
+      "loss": 1.1855,
+      "step": 29474
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.183857142857143e-06,
+      "loss": 1.2588,
+      "step": 29475
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.183588571428572e-06,
+      "loss": 1.107,
+      "step": 29476
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.18332e-06,
+      "loss": 1.3245,
+      "step": 29477
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1830514285714285e-06,
+      "loss": 1.1096,
+      "step": 29478
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.1827828571428575e-06,
+      "loss": 1.6285,
+      "step": 29479
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.182514285714286e-06,
+      "loss": 1.1052,
+      "step": 29480
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.182245714285715e-06,
+      "loss": 1.1653,
+      "step": 29481
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.181977142857143e-06,
+      "loss": 1.0264,
+      "step": 29482
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.181708571428571e-06,
+      "loss": 1.3886,
+      "step": 29483
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.18144e-06,
+      "loss": 1.5466,
+      "step": 29484
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.181171428571429e-06,
+      "loss": 0.6852,
+      "step": 29485
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.180902857142858e-06,
+      "loss": 1.1572,
+      "step": 29486
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 6.180634285714286e-06,
+      "loss": 0.9427,
+      "step": 29487
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1803657142857145e-06,
+      "loss": 1.2071,
+      "step": 29488
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1800971428571435e-06,
+      "loss": 1.2933,
+      "step": 29489
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.179828571428572e-06,
+      "loss": 1.1237,
+      "step": 29490
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.179560000000001e-06,
+      "loss": 1.5014,
+      "step": 29491
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.179291428571429e-06,
+      "loss": 1.1676,
+      "step": 29492
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.179022857142857e-06,
+      "loss": 1.3493,
+      "step": 29493
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.178754285714286e-06,
+      "loss": 0.7306,
+      "step": 29494
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.178485714285714e-06,
+      "loss": 1.2372,
+      "step": 29495
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.178217142857142e-06,
+      "loss": 0.9403,
+      "step": 29496
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1779485714285714e-06,
+      "loss": 1.3353,
+      "step": 29497
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.17768e-06,
+      "loss": 1.0691,
+      "step": 29498
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1774114285714295e-06,
+      "loss": 1.1407,
+      "step": 29499
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.177142857142858e-06,
+      "loss": 1.0626,
+      "step": 29500
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.176874285714286e-06,
+      "loss": 1.2445,
+      "step": 29501
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.176605714285715e-06,
+      "loss": 1.2365,
+      "step": 29502
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.176337142857143e-06,
+      "loss": 1.4542,
+      "step": 29503
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.176068571428572e-06,
+      "loss": 1.0857,
+      "step": 29504
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1758e-06,
+      "loss": 1.0395,
+      "step": 29505
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.175531428571428e-06,
+      "loss": 1.0024,
+      "step": 29506
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1752628571428574e-06,
+      "loss": 1.33,
+      "step": 29507
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.174994285714286e-06,
+      "loss": 0.9205,
+      "step": 29508
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.174725714285715e-06,
+      "loss": 1.272,
+      "step": 29509
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.174457142857143e-06,
+      "loss": 1.5619,
+      "step": 29510
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.174188571428571e-06,
+      "loss": 0.7574,
+      "step": 29511
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.173920000000001e-06,
+      "loss": 0.888,
+      "step": 29512
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.173651428571429e-06,
+      "loss": 1.1567,
+      "step": 29513
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.173382857142858e-06,
+      "loss": 1.1129,
+      "step": 29514
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.173114285714286e-06,
+      "loss": 1.6577,
+      "step": 29515
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.172845714285714e-06,
+      "loss": 1.0945,
+      "step": 29516
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1725771428571434e-06,
+      "loss": 1.3791,
+      "step": 29517
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.172308571428572e-06,
+      "loss": 1.1951,
+      "step": 29518
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.17204e-06,
+      "loss": 1.227,
+      "step": 29519
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.171771428571429e-06,
+      "loss": 0.6381,
+      "step": 29520
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.171502857142857e-06,
+      "loss": 0.9831,
+      "step": 29521
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.171234285714286e-06,
+      "loss": 1.4621,
+      "step": 29522
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.170965714285714e-06,
+      "loss": 0.6081,
+      "step": 29523
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.170697142857142e-06,
+      "loss": 1.032,
+      "step": 29524
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.170428571428572e-06,
+      "loss": 1.0364,
+      "step": 29525
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.17016e-06,
+      "loss": 1.4846,
+      "step": 29526
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.169891428571429e-06,
+      "loss": 1.2207,
+      "step": 29527
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.169622857142858e-06,
+      "loss": 1.3109,
+      "step": 29528
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.169354285714286e-06,
+      "loss": 1.0034,
+      "step": 29529
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.169085714285715e-06,
+      "loss": 1.328,
+      "step": 29530
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.168817142857143e-06,
+      "loss": 0.7788,
+      "step": 29531
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.168548571428572e-06,
+      "loss": 0.9847,
+      "step": 29532
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.16828e-06,
+      "loss": 0.953,
+      "step": 29533
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.168011428571428e-06,
+      "loss": 1.5677,
+      "step": 29534
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.167742857142857e-06,
+      "loss": 1.2693,
+      "step": 29535
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1674742857142855e-06,
+      "loss": 1.6673,
+      "step": 29536
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.167205714285714e-06,
+      "loss": 1.1996,
+      "step": 29537
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.166937142857144e-06,
+      "loss": 1.4255,
+      "step": 29538
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.166668571428572e-06,
+      "loss": 0.7115,
+      "step": 29539
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.166400000000001e-06,
+      "loss": 1.2003,
+      "step": 29540
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.166131428571429e-06,
+      "loss": 1.2026,
+      "step": 29541
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.165862857142857e-06,
+      "loss": 1.4379,
+      "step": 29542
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.165594285714286e-06,
+      "loss": 1.3525,
+      "step": 29543
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.165325714285714e-06,
+      "loss": 1.3915,
+      "step": 29544
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.165057142857143e-06,
+      "loss": 0.7494,
+      "step": 29545
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1647885714285715e-06,
+      "loss": 1.0705,
+      "step": 29546
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.16452e-06,
+      "loss": 1.1981,
+      "step": 29547
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.164251428571429e-06,
+      "loss": 1.1566,
+      "step": 29548
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.163982857142857e-06,
+      "loss": 0.9497,
+      "step": 29549
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.163714285714286e-06,
+      "loss": 1.0557,
+      "step": 29550
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.163445714285714e-06,
+      "loss": 0.976,
+      "step": 29551
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.163177142857143e-06,
+      "loss": 1.0569,
+      "step": 29552
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.162908571428572e-06,
+      "loss": 1.3406,
+      "step": 29553
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.16264e-06,
+      "loss": 1.3224,
+      "step": 29554
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.162371428571429e-06,
+      "loss": 1.1936,
+      "step": 29555
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1621028571428575e-06,
+      "loss": 1.0687,
+      "step": 29556
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.161834285714286e-06,
+      "loss": 0.8434,
+      "step": 29557
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.161565714285715e-06,
+      "loss": 1.3209,
+      "step": 29558
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.161297142857143e-06,
+      "loss": 0.8662,
+      "step": 29559
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.161028571428571e-06,
+      "loss": 1.2093,
+      "step": 29560
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.16076e-06,
+      "loss": 1.089,
+      "step": 29561
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.160491428571428e-06,
+      "loss": 1.3085,
+      "step": 29562
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.160222857142857e-06,
+      "loss": 1.203,
+      "step": 29563
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1599542857142855e-06,
+      "loss": 0.9809,
+      "step": 29564
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1596857142857145e-06,
+      "loss": 0.979,
+      "step": 29565
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1594171428571435e-06,
+      "loss": 0.9176,
+      "step": 29566
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.159148571428572e-06,
+      "loss": 1.2925,
+      "step": 29567
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.158880000000001e-06,
+      "loss": 1.1796,
+      "step": 29568
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.158611428571429e-06,
+      "loss": 1.203,
+      "step": 29569
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.158342857142857e-06,
+      "loss": 0.9256,
+      "step": 29570
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.158074285714286e-06,
+      "loss": 1.2413,
+      "step": 29571
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.157805714285714e-06,
+      "loss": 1.1871,
+      "step": 29572
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.157537142857143e-06,
+      "loss": 0.8645,
+      "step": 29573
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1572685714285715e-06,
+      "loss": 1.2888,
+      "step": 29574
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.157e-06,
+      "loss": 1.1599,
+      "step": 29575
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.156731428571429e-06,
+      "loss": 1.3482,
+      "step": 29576
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.156462857142857e-06,
+      "loss": 0.9549,
+      "step": 29577
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.156194285714287e-06,
+      "loss": 0.7378,
+      "step": 29578
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.155925714285715e-06,
+      "loss": 0.8517,
+      "step": 29579
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.155657142857143e-06,
+      "loss": 1.1574,
+      "step": 29580
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.155388571428572e-06,
+      "loss": 1.2658,
+      "step": 29581
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.15512e-06,
+      "loss": 1.3638,
+      "step": 29582
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1548514285714285e-06,
+      "loss": 1.1709,
+      "step": 29583
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1545828571428575e-06,
+      "loss": 1.4674,
+      "step": 29584
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.154314285714286e-06,
+      "loss": 1.3905,
+      "step": 29585
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.154045714285715e-06,
+      "loss": 1.0376,
+      "step": 29586
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.153777142857143e-06,
+      "loss": 0.7376,
+      "step": 29587
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.153508571428571e-06,
+      "loss": 1.0871,
+      "step": 29588
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.15324e-06,
+      "loss": 0.8797,
+      "step": 29589
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.152971428571428e-06,
+      "loss": 1.5322,
+      "step": 29590
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.152702857142858e-06,
+      "loss": 1.0021,
+      "step": 29591
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.152434285714286e-06,
+      "loss": 1.2932,
+      "step": 29592
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1521657142857145e-06,
+      "loss": 1.2588,
+      "step": 29593
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1518971428571435e-06,
+      "loss": 1.1904,
+      "step": 29594
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.151628571428572e-06,
+      "loss": 1.4565,
+      "step": 29595
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.151360000000001e-06,
+      "loss": 1.2283,
+      "step": 29596
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.151091428571429e-06,
+      "loss": 1.4029,
+      "step": 29597
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.150822857142857e-06,
+      "loss": 1.7632,
+      "step": 29598
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.150554285714286e-06,
+      "loss": 1.2364,
+      "step": 29599
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.150285714285714e-06,
+      "loss": 0.7182,
+      "step": 29600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.150017142857143e-06,
+      "loss": 1.0068,
+      "step": 29601
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1497485714285714e-06,
+      "loss": 1.1707,
+      "step": 29602
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.14948e-06,
+      "loss": 0.8906,
+      "step": 29603
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.149211428571429e-06,
+      "loss": 0.8866,
+      "step": 29604
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.148942857142858e-06,
+      "loss": 0.8547,
+      "step": 29605
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.148674285714286e-06,
+      "loss": 1.1358,
+      "step": 29606
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.148405714285715e-06,
+      "loss": 1.5107,
+      "step": 29607
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.148137142857143e-06,
+      "loss": 0.8899,
+      "step": 29608
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.147868571428572e-06,
+      "loss": 1.4838,
+      "step": 29609
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1476e-06,
+      "loss": 1.396,
+      "step": 29610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.147331428571428e-06,
+      "loss": 1.3226,
+      "step": 29611
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1470628571428574e-06,
+      "loss": 1.4592,
+      "step": 29612
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.146794285714286e-06,
+      "loss": 1.2828,
+      "step": 29613
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.146525714285715e-06,
+      "loss": 1.3657,
+      "step": 29614
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.146257142857143e-06,
+      "loss": 1.132,
+      "step": 29615
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.145988571428571e-06,
+      "loss": 1.1407,
+      "step": 29616
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.14572e-06,
+      "loss": 0.9067,
+      "step": 29617
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.145451428571429e-06,
+      "loss": 1.4429,
+      "step": 29618
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.145182857142858e-06,
+      "loss": 1.3983,
+      "step": 29619
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.144914285714286e-06,
+      "loss": 1.3859,
+      "step": 29620
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.144645714285714e-06,
+      "loss": 1.1598,
+      "step": 29621
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1443771428571434e-06,
+      "loss": 0.9757,
+      "step": 29622
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.144108571428572e-06,
+      "loss": 1.6314,
+      "step": 29623
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.14384e-06,
+      "loss": 1.2551,
+      "step": 29624
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.143571428571429e-06,
+      "loss": 0.8233,
+      "step": 29625
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.143302857142857e-06,
+      "loss": 0.9094,
+      "step": 29626
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.143034285714286e-06,
+      "loss": 1.129,
+      "step": 29627
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.142765714285714e-06,
+      "loss": 1.5258,
+      "step": 29628
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.142497142857142e-06,
+      "loss": 0.6384,
+      "step": 29629
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.142228571428571e-06,
+      "loss": 1.3126,
+      "step": 29630
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.14196e-06,
+      "loss": 1.371,
+      "step": 29631
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.141691428571429e-06,
+      "loss": 1.0712,
+      "step": 29632
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.141422857142858e-06,
+      "loss": 1.2122,
+      "step": 29633
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.141154285714286e-06,
+      "loss": 1.5067,
+      "step": 29634
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.140885714285715e-06,
+      "loss": 0.6657,
+      "step": 29635
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.140617142857143e-06,
+      "loss": 1.4054,
+      "step": 29636
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.140348571428572e-06,
+      "loss": 1.1323,
+      "step": 29637
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.14008e-06,
+      "loss": 1.2658,
+      "step": 29638
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.139811428571428e-06,
+      "loss": 0.8935,
+      "step": 29639
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.139542857142857e-06,
+      "loss": 1.0524,
+      "step": 29640
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1392742857142855e-06,
+      "loss": 0.8967,
+      "step": 29641
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1390057142857146e-06,
+      "loss": 0.9355,
+      "step": 29642
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.138737142857143e-06,
+      "loss": 1.5129,
+      "step": 29643
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.138468571428572e-06,
+      "loss": 1.0469,
+      "step": 29644
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.138200000000001e-06,
+      "loss": 1.1695,
+      "step": 29645
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.137931428571429e-06,
+      "loss": 1.0958,
+      "step": 29646
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.137662857142857e-06,
+      "loss": 1.2155,
+      "step": 29647
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.137394285714286e-06,
+      "loss": 1.0292,
+      "step": 29648
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.137125714285714e-06,
+      "loss": 1.1169,
+      "step": 29649
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.136857142857143e-06,
+      "loss": 1.2455,
+      "step": 29650
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1365885714285715e-06,
+      "loss": 1.3619,
+      "step": 29651
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.13632e-06,
+      "loss": 0.8064,
+      "step": 29652
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.136051428571429e-06,
+      "loss": 1.3631,
+      "step": 29653
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.135782857142857e-06,
+      "loss": 1.2829,
+      "step": 29654
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.135514285714286e-06,
+      "loss": 1.0023,
+      "step": 29655
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.135245714285714e-06,
+      "loss": 1.3441,
+      "step": 29656
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.134977142857143e-06,
+      "loss": 1.3261,
+      "step": 29657
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.134708571428572e-06,
+      "loss": 1.3426,
+      "step": 29658
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.13444e-06,
+      "loss": 0.9732,
+      "step": 29659
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.134171428571429e-06,
+      "loss": 1.4843,
+      "step": 29660
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.1339028571428575e-06,
+      "loss": 1.2812,
+      "step": 29661
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 6.133634285714286e-06,
+      "loss": 0.9795,
+      "step": 29662
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.133365714285715e-06,
+      "loss": 1.1118,
+      "step": 29663
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.133097142857143e-06,
+      "loss": 1.2966,
+      "step": 29664
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.132828571428572e-06,
+      "loss": 1.4597,
+      "step": 29665
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.13256e-06,
+      "loss": 1.0816,
+      "step": 29666
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.132291428571428e-06,
+      "loss": 1.4036,
+      "step": 29667
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.132022857142857e-06,
+      "loss": 1.1917,
+      "step": 29668
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1317542857142855e-06,
+      "loss": 1.0878,
+      "step": 29669
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1314857142857145e-06,
+      "loss": 1.4603,
+      "step": 29670
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1312171428571435e-06,
+      "loss": 1.0453,
+      "step": 29671
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.130948571428572e-06,
+      "loss": 1.2245,
+      "step": 29672
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.130680000000001e-06,
+      "loss": 0.9861,
+      "step": 29673
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.130411428571429e-06,
+      "loss": 0.966,
+      "step": 29674
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.130142857142857e-06,
+      "loss": 1.0534,
+      "step": 29675
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.129874285714286e-06,
+      "loss": 0.9737,
+      "step": 29676
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.129605714285714e-06,
+      "loss": 1.0969,
+      "step": 29677
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.129337142857143e-06,
+      "loss": 1.0346,
+      "step": 29678
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1290685714285715e-06,
+      "loss": 1.092,
+      "step": 29679
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1288e-06,
+      "loss": 1.4347,
+      "step": 29680
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.128531428571429e-06,
+      "loss": 1.7464,
+      "step": 29681
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.128262857142857e-06,
+      "loss": 1.2126,
+      "step": 29682
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.127994285714286e-06,
+      "loss": 1.0261,
+      "step": 29683
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.127725714285715e-06,
+      "loss": 1.2116,
+      "step": 29684
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.127457142857143e-06,
+      "loss": 0.9144,
+      "step": 29685
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.127188571428572e-06,
+      "loss": 1.4435,
+      "step": 29686
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.12692e-06,
+      "loss": 1.0203,
+      "step": 29687
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.126651428571429e-06,
+      "loss": 0.8664,
+      "step": 29688
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1263828571428575e-06,
+      "loss": 1.2867,
+      "step": 29689
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.126114285714286e-06,
+      "loss": 1.2454,
+      "step": 29690
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.125845714285715e-06,
+      "loss": 1.0254,
+      "step": 29691
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.125577142857143e-06,
+      "loss": 1.1225,
+      "step": 29692
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.125308571428571e-06,
+      "loss": 0.9782,
+      "step": 29693
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.12504e-06,
+      "loss": 1.3899,
+      "step": 29694
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.124771428571428e-06,
+      "loss": 1.5575,
+      "step": 29695
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.124502857142857e-06,
+      "loss": 1.2584,
+      "step": 29696
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.124234285714286e-06,
+      "loss": 1.7013,
+      "step": 29697
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1239657142857145e-06,
+      "loss": 1.0146,
+      "step": 29698
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1236971428571435e-06,
+      "loss": 1.1693,
+      "step": 29699
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.123428571428572e-06,
+      "loss": 0.895,
+      "step": 29700
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.123160000000001e-06,
+      "loss": 1.0723,
+      "step": 29701
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.122891428571429e-06,
+      "loss": 1.4532,
+      "step": 29702
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.122622857142857e-06,
+      "loss": 1.0472,
+      "step": 29703
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.122354285714286e-06,
+      "loss": 1.165,
+      "step": 29704
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.122085714285714e-06,
+      "loss": 1.4184,
+      "step": 29705
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.121817142857143e-06,
+      "loss": 0.9348,
+      "step": 29706
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1215485714285714e-06,
+      "loss": 1.4284,
+      "step": 29707
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.12128e-06,
+      "loss": 1.1032,
+      "step": 29708
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.121011428571429e-06,
+      "loss": 0.6533,
+      "step": 29709
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.120742857142858e-06,
+      "loss": 1.0387,
+      "step": 29710
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.120474285714286e-06,
+      "loss": 1.3935,
+      "step": 29711
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.120205714285715e-06,
+      "loss": 1.1286,
+      "step": 29712
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.119937142857143e-06,
+      "loss": 1.5002,
+      "step": 29713
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.119668571428572e-06,
+      "loss": 1.0865,
+      "step": 29714
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1194e-06,
+      "loss": 1.1687,
+      "step": 29715
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.119131428571428e-06,
+      "loss": 1.135,
+      "step": 29716
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1188628571428574e-06,
+      "loss": 1.5756,
+      "step": 29717
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.118594285714286e-06,
+      "loss": 1.0828,
+      "step": 29718
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.118325714285715e-06,
+      "loss": 1.1159,
+      "step": 29719
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.118057142857143e-06,
+      "loss": 1.5829,
+      "step": 29720
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.117788571428571e-06,
+      "loss": 1.1664,
+      "step": 29721
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.11752e-06,
+      "loss": 1.0001,
+      "step": 29722
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.117251428571429e-06,
+      "loss": 1.6605,
+      "step": 29723
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.116982857142858e-06,
+      "loss": 0.8002,
+      "step": 29724
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.116714285714286e-06,
+      "loss": 1.027,
+      "step": 29725
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.116445714285714e-06,
+      "loss": 1.2724,
+      "step": 29726
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1161771428571434e-06,
+      "loss": 1.1991,
+      "step": 29727
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.115908571428572e-06,
+      "loss": 0.866,
+      "step": 29728
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.115640000000001e-06,
+      "loss": 1.5297,
+      "step": 29729
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.115371428571429e-06,
+      "loss": 1.4896,
+      "step": 29730
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.115102857142857e-06,
+      "loss": 1.2632,
+      "step": 29731
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.114834285714286e-06,
+      "loss": 0.8427,
+      "step": 29732
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.114565714285714e-06,
+      "loss": 1.4359,
+      "step": 29733
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.114297142857142e-06,
+      "loss": 1.3424,
+      "step": 29734
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.114028571428571e-06,
+      "loss": 1.0897,
+      "step": 29735
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.11376e-06,
+      "loss": 1.6026,
+      "step": 29736
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.113491428571429e-06,
+      "loss": 1.3274,
+      "step": 29737
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.113222857142858e-06,
+      "loss": 1.1366,
+      "step": 29738
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.112954285714286e-06,
+      "loss": 1.2998,
+      "step": 29739
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.112685714285715e-06,
+      "loss": 1.273,
+      "step": 29740
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.112417142857143e-06,
+      "loss": 1.0982,
+      "step": 29741
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.112148571428572e-06,
+      "loss": 1.4901,
+      "step": 29742
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.11188e-06,
+      "loss": 0.6677,
+      "step": 29743
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.111611428571428e-06,
+      "loss": 1.2656,
+      "step": 29744
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.111342857142857e-06,
+      "loss": 1.0222,
+      "step": 29745
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1110742857142855e-06,
+      "loss": 0.9049,
+      "step": 29746
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1108057142857146e-06,
+      "loss": 0.9007,
+      "step": 29747
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.110537142857143e-06,
+      "loss": 1.3295,
+      "step": 29748
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.110268571428572e-06,
+      "loss": 1.2079,
+      "step": 29749
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.110000000000001e-06,
+      "loss": 1.2461,
+      "step": 29750
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.109731428571429e-06,
+      "loss": 1.1146,
+      "step": 29751
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.109462857142858e-06,
+      "loss": 1.07,
+      "step": 29752
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.109194285714286e-06,
+      "loss": 1.2391,
+      "step": 29753
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.108925714285714e-06,
+      "loss": 0.9603,
+      "step": 29754
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.108657142857143e-06,
+      "loss": 0.8779,
+      "step": 29755
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1083885714285715e-06,
+      "loss": 1.1307,
+      "step": 29756
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.10812e-06,
+      "loss": 1.1518,
+      "step": 29757
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.107851428571429e-06,
+      "loss": 1.1554,
+      "step": 29758
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.107582857142857e-06,
+      "loss": 1.0395,
+      "step": 29759
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.107314285714286e-06,
+      "loss": 1.2485,
+      "step": 29760
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.107045714285714e-06,
+      "loss": 1.1667,
+      "step": 29761
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.106777142857143e-06,
+      "loss": 1.0979,
+      "step": 29762
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.106508571428572e-06,
+      "loss": 1.5529,
+      "step": 29763
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.10624e-06,
+      "loss": 1.3337,
+      "step": 29764
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.105971428571429e-06,
+      "loss": 0.9125,
+      "step": 29765
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1057028571428575e-06,
+      "loss": 1.2559,
+      "step": 29766
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.105434285714286e-06,
+      "loss": 0.9578,
+      "step": 29767
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.105165714285715e-06,
+      "loss": 1.1392,
+      "step": 29768
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.104897142857143e-06,
+      "loss": 1.4181,
+      "step": 29769
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.104628571428572e-06,
+      "loss": 1.3775,
+      "step": 29770
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.10436e-06,
+      "loss": 1.5066,
+      "step": 29771
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.104091428571428e-06,
+      "loss": 1.1763,
+      "step": 29772
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.103822857142857e-06,
+      "loss": 1.5419,
+      "step": 29773
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1035542857142855e-06,
+      "loss": 1.1739,
+      "step": 29774
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1032857142857145e-06,
+      "loss": 1.4969,
+      "step": 29775
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1030171428571435e-06,
+      "loss": 1.1669,
+      "step": 29776
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.102748571428572e-06,
+      "loss": 1.0086,
+      "step": 29777
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.102480000000001e-06,
+      "loss": 1.0445,
+      "step": 29778
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.102211428571429e-06,
+      "loss": 1.3996,
+      "step": 29779
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.101942857142857e-06,
+      "loss": 1.305,
+      "step": 29780
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.101674285714286e-06,
+      "loss": 0.8302,
+      "step": 29781
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.101405714285714e-06,
+      "loss": 1.1386,
+      "step": 29782
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.101137142857143e-06,
+      "loss": 1.0288,
+      "step": 29783
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1008685714285715e-06,
+      "loss": 1.4266,
+      "step": 29784
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.1006e-06,
+      "loss": 0.7278,
+      "step": 29785
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.100331428571429e-06,
+      "loss": 1.1608,
+      "step": 29786
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.100062857142857e-06,
+      "loss": 1.317,
+      "step": 29787
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.099794285714286e-06,
+      "loss": 1.5427,
+      "step": 29788
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.099525714285715e-06,
+      "loss": 1.4364,
+      "step": 29789
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.099257142857143e-06,
+      "loss": 1.0703,
+      "step": 29790
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.098988571428572e-06,
+      "loss": 1.0907,
+      "step": 29791
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.09872e-06,
+      "loss": 1.0243,
+      "step": 29792
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.098451428571429e-06,
+      "loss": 1.0566,
+      "step": 29793
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.0981828571428575e-06,
+      "loss": 1.7555,
+      "step": 29794
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.097914285714286e-06,
+      "loss": 1.3852,
+      "step": 29795
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.097645714285715e-06,
+      "loss": 1.1527,
+      "step": 29796
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.097377142857143e-06,
+      "loss": 1.1339,
+      "step": 29797
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.097108571428571e-06,
+      "loss": 1.3201,
+      "step": 29798
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.09684e-06,
+      "loss": 1.2186,
+      "step": 29799
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.096571428571428e-06,
+      "loss": 1.3484,
+      "step": 29800
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.096302857142857e-06,
+      "loss": 1.3997,
+      "step": 29801
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.096034285714286e-06,
+      "loss": 1.1421,
+      "step": 29802
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.0957657142857145e-06,
+      "loss": 1.0281,
+      "step": 29803
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.0954971428571435e-06,
+      "loss": 1.0604,
+      "step": 29804
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.095228571428572e-06,
+      "loss": 1.3714,
+      "step": 29805
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.094960000000001e-06,
+      "loss": 1.0378,
+      "step": 29806
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.094691428571429e-06,
+      "loss": 0.6768,
+      "step": 29807
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.094422857142857e-06,
+      "loss": 1.3495,
+      "step": 29808
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.094154285714286e-06,
+      "loss": 0.9574,
+      "step": 29809
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.093885714285714e-06,
+      "loss": 1.3349,
+      "step": 29810
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.093617142857143e-06,
+      "loss": 1.1917,
+      "step": 29811
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.0933485714285714e-06,
+      "loss": 0.7157,
+      "step": 29812
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.09308e-06,
+      "loss": 1.2633,
+      "step": 29813
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.092811428571429e-06,
+      "loss": 1.1226,
+      "step": 29814
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.092542857142858e-06,
+      "loss": 1.238,
+      "step": 29815
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.092274285714287e-06,
+      "loss": 0.8944,
+      "step": 29816
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.092005714285715e-06,
+      "loss": 0.9991,
+      "step": 29817
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.091737142857143e-06,
+      "loss": 1.2334,
+      "step": 29818
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.091468571428572e-06,
+      "loss": 1.5972,
+      "step": 29819
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.0912e-06,
+      "loss": 1.2566,
+      "step": 29820
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.090931428571428e-06,
+      "loss": 1.2172,
+      "step": 29821
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.0906628571428574e-06,
+      "loss": 1.496,
+      "step": 29822
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.090394285714286e-06,
+      "loss": 0.9463,
+      "step": 29823
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.090125714285715e-06,
+      "loss": 1.3824,
+      "step": 29824
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.089857142857143e-06,
+      "loss": 1.4321,
+      "step": 29825
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.089588571428571e-06,
+      "loss": 1.3482,
+      "step": 29826
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.08932e-06,
+      "loss": 1.4265,
+      "step": 29827
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.089051428571429e-06,
+      "loss": 1.1762,
+      "step": 29828
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.088782857142858e-06,
+      "loss": 1.2471,
+      "step": 29829
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.088514285714286e-06,
+      "loss": 1.542,
+      "step": 29830
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.088245714285714e-06,
+      "loss": 1.4171,
+      "step": 29831
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.087977142857143e-06,
+      "loss": 1.2945,
+      "step": 29832
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.087708571428572e-06,
+      "loss": 1.19,
+      "step": 29833
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.087440000000001e-06,
+      "loss": 1.2141,
+      "step": 29834
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.087171428571429e-06,
+      "loss": 1.0564,
+      "step": 29835
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.086902857142857e-06,
+      "loss": 0.884,
+      "step": 29836
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6.086634285714286e-06,
+      "loss": 1.0094,
+      "step": 29837
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.086365714285714e-06,
+      "loss": 1.0758,
+      "step": 29838
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.086097142857143e-06,
+      "loss": 1.2053,
+      "step": 29839
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.085828571428571e-06,
+      "loss": 1.3078,
+      "step": 29840
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.08556e-06,
+      "loss": 1.8678,
+      "step": 29841
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.085291428571429e-06,
+      "loss": 1.1844,
+      "step": 29842
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.085022857142858e-06,
+      "loss": 1.3896,
+      "step": 29843
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.084754285714286e-06,
+      "loss": 0.9141,
+      "step": 29844
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.084485714285715e-06,
+      "loss": 0.9208,
+      "step": 29845
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.084217142857143e-06,
+      "loss": 1.2518,
+      "step": 29846
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.083948571428572e-06,
+      "loss": 1.0792,
+      "step": 29847
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.08368e-06,
+      "loss": 1.1155,
+      "step": 29848
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.083411428571428e-06,
+      "loss": 1.417,
+      "step": 29849
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.083142857142857e-06,
+      "loss": 0.9338,
+      "step": 29850
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0828742857142855e-06,
+      "loss": 0.9741,
+      "step": 29851
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0826057142857146e-06,
+      "loss": 0.9787,
+      "step": 29852
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.082337142857143e-06,
+      "loss": 0.9763,
+      "step": 29853
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.082068571428572e-06,
+      "loss": 1.1877,
+      "step": 29854
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.081800000000001e-06,
+      "loss": 1.207,
+      "step": 29855
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.081531428571429e-06,
+      "loss": 1.0398,
+      "step": 29856
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.081262857142858e-06,
+      "loss": 1.1323,
+      "step": 29857
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.080994285714286e-06,
+      "loss": 1.1776,
+      "step": 29858
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.080725714285714e-06,
+      "loss": 0.8862,
+      "step": 29859
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.080457142857143e-06,
+      "loss": 1.3897,
+      "step": 29860
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0801885714285715e-06,
+      "loss": 0.7106,
+      "step": 29861
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0799200000000006e-06,
+      "loss": 1.3287,
+      "step": 29862
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.079651428571429e-06,
+      "loss": 1.0902,
+      "step": 29863
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.079382857142857e-06,
+      "loss": 0.8122,
+      "step": 29864
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.079114285714286e-06,
+      "loss": 0.9383,
+      "step": 29865
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.078845714285714e-06,
+      "loss": 1.2958,
+      "step": 29866
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.078577142857143e-06,
+      "loss": 1.2855,
+      "step": 29867
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.078308571428572e-06,
+      "loss": 1.381,
+      "step": 29868
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.07804e-06,
+      "loss": 1.0392,
+      "step": 29869
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.077771428571429e-06,
+      "loss": 0.9539,
+      "step": 29870
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0775028571428575e-06,
+      "loss": 1.1979,
+      "step": 29871
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.077234285714286e-06,
+      "loss": 1.2562,
+      "step": 29872
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.076965714285715e-06,
+      "loss": 1.4064,
+      "step": 29873
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.076697142857143e-06,
+      "loss": 1.1501,
+      "step": 29874
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.076428571428572e-06,
+      "loss": 1.0988,
+      "step": 29875
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.07616e-06,
+      "loss": 1.3775,
+      "step": 29876
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.075891428571428e-06,
+      "loss": 0.9862,
+      "step": 29877
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.075622857142857e-06,
+      "loss": 1.1984,
+      "step": 29878
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0753542857142855e-06,
+      "loss": 1.1372,
+      "step": 29879
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0750857142857145e-06,
+      "loss": 0.7153,
+      "step": 29880
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0748171428571435e-06,
+      "loss": 0.965,
+      "step": 29881
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.074548571428572e-06,
+      "loss": 1.208,
+      "step": 29882
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.074280000000001e-06,
+      "loss": 1.2558,
+      "step": 29883
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.074011428571429e-06,
+      "loss": 1.4516,
+      "step": 29884
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.073742857142857e-06,
+      "loss": 0.8435,
+      "step": 29885
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.073474285714286e-06,
+      "loss": 0.6861,
+      "step": 29886
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.073205714285714e-06,
+      "loss": 0.9539,
+      "step": 29887
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.072937142857143e-06,
+      "loss": 1.5791,
+      "step": 29888
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0726685714285715e-06,
+      "loss": 1.1929,
+      "step": 29889
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0724e-06,
+      "loss": 1.1101,
+      "step": 29890
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.072131428571429e-06,
+      "loss": 1.3432,
+      "step": 29891
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.071862857142857e-06,
+      "loss": 1.2002,
+      "step": 29892
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.071594285714286e-06,
+      "loss": 1.6191,
+      "step": 29893
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.071325714285715e-06,
+      "loss": 0.8912,
+      "step": 29894
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.071057142857143e-06,
+      "loss": 1.2567,
+      "step": 29895
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.070788571428572e-06,
+      "loss": 1.5749,
+      "step": 29896
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.07052e-06,
+      "loss": 0.8883,
+      "step": 29897
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.070251428571429e-06,
+      "loss": 1.1621,
+      "step": 29898
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0699828571428575e-06,
+      "loss": 1.3849,
+      "step": 29899
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.069714285714286e-06,
+      "loss": 0.9115,
+      "step": 29900
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.069445714285715e-06,
+      "loss": 0.9371,
+      "step": 29901
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.069177142857143e-06,
+      "loss": 1.1907,
+      "step": 29902
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.068908571428572e-06,
+      "loss": 0.8945,
+      "step": 29903
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.06864e-06,
+      "loss": 0.9922,
+      "step": 29904
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.068371428571428e-06,
+      "loss": 0.9797,
+      "step": 29905
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.068102857142857e-06,
+      "loss": 0.8103,
+      "step": 29906
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.067834285714286e-06,
+      "loss": 1.0622,
+      "step": 29907
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0675657142857145e-06,
+      "loss": 0.7998,
+      "step": 29908
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0672971428571435e-06,
+      "loss": 1.429,
+      "step": 29909
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.067028571428572e-06,
+      "loss": 1.0847,
+      "step": 29910
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.066760000000001e-06,
+      "loss": 0.8065,
+      "step": 29911
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.066491428571429e-06,
+      "loss": 0.5776,
+      "step": 29912
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.066222857142857e-06,
+      "loss": 0.9326,
+      "step": 29913
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.065954285714286e-06,
+      "loss": 1.4539,
+      "step": 29914
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.065685714285714e-06,
+      "loss": 1.4872,
+      "step": 29915
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.065417142857143e-06,
+      "loss": 1.0834,
+      "step": 29916
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0651485714285714e-06,
+      "loss": 1.0838,
+      "step": 29917
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.06488e-06,
+      "loss": 1.3984,
+      "step": 29918
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.064611428571429e-06,
+      "loss": 1.1024,
+      "step": 29919
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.064342857142858e-06,
+      "loss": 1.0273,
+      "step": 29920
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.064074285714287e-06,
+      "loss": 1.0508,
+      "step": 29921
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.063805714285715e-06,
+      "loss": 1.1131,
+      "step": 29922
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.063537142857143e-06,
+      "loss": 0.8203,
+      "step": 29923
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.063268571428572e-06,
+      "loss": 1.1849,
+      "step": 29924
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.063e-06,
+      "loss": 0.9611,
+      "step": 29925
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.062731428571429e-06,
+      "loss": 1.0677,
+      "step": 29926
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0624628571428574e-06,
+      "loss": 1.3713,
+      "step": 29927
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.062194285714286e-06,
+      "loss": 1.0304,
+      "step": 29928
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.061925714285715e-06,
+      "loss": 1.1739,
+      "step": 29929
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.061657142857143e-06,
+      "loss": 1.5121,
+      "step": 29930
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.061388571428571e-06,
+      "loss": 1.0011,
+      "step": 29931
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.06112e-06,
+      "loss": 1.1829,
+      "step": 29932
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.060851428571429e-06,
+      "loss": 1.0727,
+      "step": 29933
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.060582857142858e-06,
+      "loss": 0.9561,
+      "step": 29934
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.060314285714286e-06,
+      "loss": 0.9369,
+      "step": 29935
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.060045714285714e-06,
+      "loss": 1.2147,
+      "step": 29936
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.059777142857143e-06,
+      "loss": 1.2636,
+      "step": 29937
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.059508571428572e-06,
+      "loss": 1.1368,
+      "step": 29938
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.059240000000001e-06,
+      "loss": 0.6169,
+      "step": 29939
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.058971428571429e-06,
+      "loss": 0.849,
+      "step": 29940
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.058702857142857e-06,
+      "loss": 1.2723,
+      "step": 29941
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.058434285714286e-06,
+      "loss": 1.0429,
+      "step": 29942
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.058165714285714e-06,
+      "loss": 0.8913,
+      "step": 29943
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.057897142857143e-06,
+      "loss": 1.0112,
+      "step": 29944
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.057628571428571e-06,
+      "loss": 0.7011,
+      "step": 29945
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0573599999999995e-06,
+      "loss": 1.0753,
+      "step": 29946
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.057091428571429e-06,
+      "loss": 1.3024,
+      "step": 29947
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.056822857142858e-06,
+      "loss": 1.2147,
+      "step": 29948
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.056554285714286e-06,
+      "loss": 0.5387,
+      "step": 29949
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.056285714285715e-06,
+      "loss": 1.3535,
+      "step": 29950
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.056017142857143e-06,
+      "loss": 1.0589,
+      "step": 29951
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.055748571428572e-06,
+      "loss": 1.0081,
+      "step": 29952
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.05548e-06,
+      "loss": 1.0948,
+      "step": 29953
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.055211428571428e-06,
+      "loss": 0.639,
+      "step": 29954
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.054942857142857e-06,
+      "loss": 1.2542,
+      "step": 29955
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0546742857142855e-06,
+      "loss": 1.2175,
+      "step": 29956
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0544057142857146e-06,
+      "loss": 1.1715,
+      "step": 29957
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.054137142857143e-06,
+      "loss": 0.622,
+      "step": 29958
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.053868571428571e-06,
+      "loss": 1.2543,
+      "step": 29959
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.053600000000001e-06,
+      "loss": 1.1665,
+      "step": 29960
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.053331428571429e-06,
+      "loss": 1.3688,
+      "step": 29961
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.053062857142858e-06,
+      "loss": 1.2246,
+      "step": 29962
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.052794285714286e-06,
+      "loss": 1.0762,
+      "step": 29963
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.052525714285714e-06,
+      "loss": 1.0314,
+      "step": 29964
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.052257142857143e-06,
+      "loss": 1.2788,
+      "step": 29965
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0519885714285715e-06,
+      "loss": 1.4696,
+      "step": 29966
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0517200000000006e-06,
+      "loss": 0.9793,
+      "step": 29967
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.051451428571429e-06,
+      "loss": 1.4653,
+      "step": 29968
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.051182857142857e-06,
+      "loss": 1.6198,
+      "step": 29969
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.050914285714286e-06,
+      "loss": 1.3387,
+      "step": 29970
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.050645714285714e-06,
+      "loss": 1.5952,
+      "step": 29971
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.050377142857142e-06,
+      "loss": 1.0549,
+      "step": 29972
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.050108571428572e-06,
+      "loss": 0.7444,
+      "step": 29973
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.04984e-06,
+      "loss": 1.2214,
+      "step": 29974
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.049571428571429e-06,
+      "loss": 1.1543,
+      "step": 29975
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0493028571428575e-06,
+      "loss": 1.4405,
+      "step": 29976
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.049034285714286e-06,
+      "loss": 1.0932,
+      "step": 29977
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.048765714285715e-06,
+      "loss": 0.891,
+      "step": 29978
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.048497142857143e-06,
+      "loss": 1.2074,
+      "step": 29979
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.048228571428572e-06,
+      "loss": 1.2705,
+      "step": 29980
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.04796e-06,
+      "loss": 1.3403,
+      "step": 29981
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.047691428571428e-06,
+      "loss": 0.8757,
+      "step": 29982
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.047422857142857e-06,
+      "loss": 1.281,
+      "step": 29983
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0471542857142855e-06,
+      "loss": 0.7192,
+      "step": 29984
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0468857142857145e-06,
+      "loss": 1.2997,
+      "step": 29985
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0466171428571435e-06,
+      "loss": 1.1667,
+      "step": 29986
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.046348571428572e-06,
+      "loss": 1.258,
+      "step": 29987
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.046080000000001e-06,
+      "loss": 1.0254,
+      "step": 29988
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.045811428571429e-06,
+      "loss": 1.2123,
+      "step": 29989
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.045542857142858e-06,
+      "loss": 1.094,
+      "step": 29990
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.045274285714286e-06,
+      "loss": 1.1336,
+      "step": 29991
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.045005714285714e-06,
+      "loss": 1.1437,
+      "step": 29992
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.044737142857143e-06,
+      "loss": 1.0974,
+      "step": 29993
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0444685714285715e-06,
+      "loss": 1.1608,
+      "step": 29994
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0442e-06,
+      "loss": 1.3589,
+      "step": 29995
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.043931428571429e-06,
+      "loss": 1.0324,
+      "step": 29996
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.043662857142857e-06,
+      "loss": 0.9763,
+      "step": 29997
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.043394285714286e-06,
+      "loss": 1.3661,
+      "step": 29998
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.043125714285714e-06,
+      "loss": 0.8868,
+      "step": 29999
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.042857142857143e-06,
+      "loss": 0.9327,
+      "step": 30000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.042588571428572e-06,
+      "loss": 0.8559,
+      "step": 30001
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.04232e-06,
+      "loss": 0.9144,
+      "step": 30002
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.042051428571429e-06,
+      "loss": 0.882,
+      "step": 30003
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0417828571428575e-06,
+      "loss": 1.2956,
+      "step": 30004
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.041514285714286e-06,
+      "loss": 1.1169,
+      "step": 30005
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.041245714285715e-06,
+      "loss": 0.921,
+      "step": 30006
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.040977142857143e-06,
+      "loss": 1.3954,
+      "step": 30007
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.040708571428572e-06,
+      "loss": 1.0915,
+      "step": 30008
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.04044e-06,
+      "loss": 0.8341,
+      "step": 30009
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.040171428571428e-06,
+      "loss": 1.0427,
+      "step": 30010
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.039902857142857e-06,
+      "loss": 1.2506,
+      "step": 30011
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 6.0396342857142854e-06,
+      "loss": 1.1137,
+      "step": 30012
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.039365714285715e-06,
+      "loss": 1.0186,
+      "step": 30013
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0390971428571435e-06,
+      "loss": 1.0513,
+      "step": 30014
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.038828571428572e-06,
+      "loss": 1.0877,
+      "step": 30015
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.038560000000001e-06,
+      "loss": 0.6999,
+      "step": 30016
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.038291428571429e-06,
+      "loss": 0.9763,
+      "step": 30017
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.038022857142857e-06,
+      "loss": 1.4864,
+      "step": 30018
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.037754285714286e-06,
+      "loss": 1.3144,
+      "step": 30019
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.037485714285714e-06,
+      "loss": 1.4597,
+      "step": 30020
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.037217142857143e-06,
+      "loss": 1.2107,
+      "step": 30021
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0369485714285714e-06,
+      "loss": 1.179,
+      "step": 30022
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.03668e-06,
+      "loss": 1.2255,
+      "step": 30023
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.036411428571429e-06,
+      "loss": 1.178,
+      "step": 30024
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.036142857142857e-06,
+      "loss": 1.2129,
+      "step": 30025
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.035874285714287e-06,
+      "loss": 0.842,
+      "step": 30026
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.035605714285715e-06,
+      "loss": 0.7799,
+      "step": 30027
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.035337142857143e-06,
+      "loss": 0.953,
+      "step": 30028
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.035068571428572e-06,
+      "loss": 0.9091,
+      "step": 30029
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0348e-06,
+      "loss": 1.092,
+      "step": 30030
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.034531428571429e-06,
+      "loss": 0.862,
+      "step": 30031
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0342628571428574e-06,
+      "loss": 1.0997,
+      "step": 30032
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.033994285714286e-06,
+      "loss": 0.9941,
+      "step": 30033
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.033725714285715e-06,
+      "loss": 1.0379,
+      "step": 30034
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.033457142857143e-06,
+      "loss": 1.1464,
+      "step": 30035
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.033188571428571e-06,
+      "loss": 1.0009,
+      "step": 30036
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.03292e-06,
+      "loss": 1.199,
+      "step": 30037
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.032651428571428e-06,
+      "loss": 1.1922,
+      "step": 30038
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.032382857142858e-06,
+      "loss": 1.1557,
+      "step": 30039
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.032114285714286e-06,
+      "loss": 0.5754,
+      "step": 30040
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.031845714285714e-06,
+      "loss": 0.9589,
+      "step": 30041
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.031577142857143e-06,
+      "loss": 1.141,
+      "step": 30042
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.031308571428572e-06,
+      "loss": 1.2034,
+      "step": 30043
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.031040000000001e-06,
+      "loss": 1.1129,
+      "step": 30044
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.030771428571429e-06,
+      "loss": 1.0368,
+      "step": 30045
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.030502857142857e-06,
+      "loss": 1.2504,
+      "step": 30046
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.030234285714286e-06,
+      "loss": 1.1486,
+      "step": 30047
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.029965714285714e-06,
+      "loss": 1.2393,
+      "step": 30048
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.029697142857143e-06,
+      "loss": 0.9076,
+      "step": 30049
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.029428571428571e-06,
+      "loss": 0.9516,
+      "step": 30050
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0291599999999995e-06,
+      "loss": 0.9595,
+      "step": 30051
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0288914285714286e-06,
+      "loss": 1.2877,
+      "step": 30052
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.028622857142858e-06,
+      "loss": 0.6819,
+      "step": 30053
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.028354285714287e-06,
+      "loss": 1.1314,
+      "step": 30054
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.028085714285715e-06,
+      "loss": 1.4814,
+      "step": 30055
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.027817142857143e-06,
+      "loss": 1.4699,
+      "step": 30056
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.027548571428572e-06,
+      "loss": 1.5211,
+      "step": 30057
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.02728e-06,
+      "loss": 1.1574,
+      "step": 30058
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.027011428571428e-06,
+      "loss": 1.5229,
+      "step": 30059
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.026742857142857e-06,
+      "loss": 1.4151,
+      "step": 30060
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0264742857142855e-06,
+      "loss": 0.9185,
+      "step": 30061
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0262057142857146e-06,
+      "loss": 1.6007,
+      "step": 30062
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.025937142857143e-06,
+      "loss": 0.7778,
+      "step": 30063
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.025668571428571e-06,
+      "loss": 1.0011,
+      "step": 30064
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0254e-06,
+      "loss": 1.4463,
+      "step": 30065
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.025131428571429e-06,
+      "loss": 1.3633,
+      "step": 30066
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.024862857142858e-06,
+      "loss": 1.1327,
+      "step": 30067
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.024594285714286e-06,
+      "loss": 1.1301,
+      "step": 30068
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.024325714285714e-06,
+      "loss": 1.186,
+      "step": 30069
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.024057142857143e-06,
+      "loss": 1.183,
+      "step": 30070
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0237885714285715e-06,
+      "loss": 1.0452,
+      "step": 30071
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0235200000000006e-06,
+      "loss": 1.0923,
+      "step": 30072
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.023251428571429e-06,
+      "loss": 1.4649,
+      "step": 30073
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.022982857142857e-06,
+      "loss": 1.0102,
+      "step": 30074
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.022714285714286e-06,
+      "loss": 1.3799,
+      "step": 30075
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.022445714285714e-06,
+      "loss": 1.2232,
+      "step": 30076
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.022177142857143e-06,
+      "loss": 1.007,
+      "step": 30077
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.021908571428571e-06,
+      "loss": 0.8477,
+      "step": 30078
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.02164e-06,
+      "loss": 0.9119,
+      "step": 30079
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.021371428571429e-06,
+      "loss": 0.9267,
+      "step": 30080
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0211028571428575e-06,
+      "loss": 1.2883,
+      "step": 30081
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.020834285714286e-06,
+      "loss": 0.7761,
+      "step": 30082
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.020565714285715e-06,
+      "loss": 0.8765,
+      "step": 30083
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.020297142857143e-06,
+      "loss": 1.0894,
+      "step": 30084
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.020028571428572e-06,
+      "loss": 1.2774,
+      "step": 30085
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.01976e-06,
+      "loss": 1.3494,
+      "step": 30086
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.019491428571428e-06,
+      "loss": 0.8304,
+      "step": 30087
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.019222857142857e-06,
+      "loss": 1.4359,
+      "step": 30088
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0189542857142855e-06,
+      "loss": 1.0217,
+      "step": 30089
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0186857142857145e-06,
+      "loss": 1.3405,
+      "step": 30090
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.018417142857143e-06,
+      "loss": 1.3164,
+      "step": 30091
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.018148571428572e-06,
+      "loss": 0.8234,
+      "step": 30092
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.017880000000001e-06,
+      "loss": 1.0336,
+      "step": 30093
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.017611428571429e-06,
+      "loss": 1.0938,
+      "step": 30094
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.017342857142858e-06,
+      "loss": 1.1451,
+      "step": 30095
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.017074285714286e-06,
+      "loss": 1.4746,
+      "step": 30096
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.016805714285714e-06,
+      "loss": 1.0515,
+      "step": 30097
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.016537142857143e-06,
+      "loss": 1.3755,
+      "step": 30098
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0162685714285715e-06,
+      "loss": 1.2248,
+      "step": 30099
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0160000000000005e-06,
+      "loss": 0.986,
+      "step": 30100
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.015731428571429e-06,
+      "loss": 1.5135,
+      "step": 30101
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.015462857142857e-06,
+      "loss": 1.4195,
+      "step": 30102
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.015194285714286e-06,
+      "loss": 1.1245,
+      "step": 30103
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.014925714285714e-06,
+      "loss": 1.2916,
+      "step": 30104
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.014657142857143e-06,
+      "loss": 1.0531,
+      "step": 30105
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.014388571428572e-06,
+      "loss": 1.5224,
+      "step": 30106
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.01412e-06,
+      "loss": 0.9992,
+      "step": 30107
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.013851428571429e-06,
+      "loss": 0.9717,
+      "step": 30108
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0135828571428575e-06,
+      "loss": 1.4717,
+      "step": 30109
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.013314285714286e-06,
+      "loss": 1.293,
+      "step": 30110
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.013045714285715e-06,
+      "loss": 1.3874,
+      "step": 30111
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.012777142857143e-06,
+      "loss": 0.8535,
+      "step": 30112
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.012508571428572e-06,
+      "loss": 1.4887,
+      "step": 30113
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.01224e-06,
+      "loss": 0.7096,
+      "step": 30114
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.011971428571428e-06,
+      "loss": 1.3837,
+      "step": 30115
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.011702857142857e-06,
+      "loss": 1.536,
+      "step": 30116
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0114342857142854e-06,
+      "loss": 1.2341,
+      "step": 30117
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0111657142857145e-06,
+      "loss": 1.5108,
+      "step": 30118
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0108971428571435e-06,
+      "loss": 0.5974,
+      "step": 30119
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.010628571428572e-06,
+      "loss": 0.664,
+      "step": 30120
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.010360000000001e-06,
+      "loss": 1.1325,
+      "step": 30121
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.010091428571429e-06,
+      "loss": 0.9996,
+      "step": 30122
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.009822857142857e-06,
+      "loss": 0.9652,
+      "step": 30123
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.009554285714286e-06,
+      "loss": 1.1692,
+      "step": 30124
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.009285714285714e-06,
+      "loss": 1.1129,
+      "step": 30125
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.009017142857143e-06,
+      "loss": 1.6833,
+      "step": 30126
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0087485714285714e-06,
+      "loss": 1.48,
+      "step": 30127
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.00848e-06,
+      "loss": 1.3355,
+      "step": 30128
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.008211428571429e-06,
+      "loss": 1.3016,
+      "step": 30129
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.007942857142857e-06,
+      "loss": 1.4972,
+      "step": 30130
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.007674285714286e-06,
+      "loss": 1.235,
+      "step": 30131
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.007405714285715e-06,
+      "loss": 0.7829,
+      "step": 30132
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.007137142857143e-06,
+      "loss": 0.9288,
+      "step": 30133
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.006868571428572e-06,
+      "loss": 1.4941,
+      "step": 30134
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0066e-06,
+      "loss": 1.4044,
+      "step": 30135
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.006331428571429e-06,
+      "loss": 1.4614,
+      "step": 30136
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0060628571428574e-06,
+      "loss": 1.4045,
+      "step": 30137
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.005794285714286e-06,
+      "loss": 1.2187,
+      "step": 30138
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.005525714285715e-06,
+      "loss": 1.2379,
+      "step": 30139
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.005257142857143e-06,
+      "loss": 0.9248,
+      "step": 30140
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.004988571428572e-06,
+      "loss": 0.9947,
+      "step": 30141
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.00472e-06,
+      "loss": 1.1841,
+      "step": 30142
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.004451428571428e-06,
+      "loss": 1.4086,
+      "step": 30143
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.004182857142857e-06,
+      "loss": 0.8418,
+      "step": 30144
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.003914285714286e-06,
+      "loss": 1.4916,
+      "step": 30145
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.003645714285714e-06,
+      "loss": 0.8676,
+      "step": 30146
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.003377142857143e-06,
+      "loss": 1.3775,
+      "step": 30147
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.003108571428572e-06,
+      "loss": 1.0963,
+      "step": 30148
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.002840000000001e-06,
+      "loss": 1.1388,
+      "step": 30149
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.002571428571429e-06,
+      "loss": 1.504,
+      "step": 30150
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.002302857142857e-06,
+      "loss": 1.6786,
+      "step": 30151
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.002034285714286e-06,
+      "loss": 0.8947,
+      "step": 30152
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.001765714285714e-06,
+      "loss": 1.0836,
+      "step": 30153
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.001497142857143e-06,
+      "loss": 1.267,
+      "step": 30154
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.001228571428571e-06,
+      "loss": 0.8381,
+      "step": 30155
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0009599999999995e-06,
+      "loss": 1.0684,
+      "step": 30156
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.0006914285714286e-06,
+      "loss": 1.4214,
+      "step": 30157
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.000422857142858e-06,
+      "loss": 0.9083,
+      "step": 30158
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 6.000154285714287e-06,
+      "loss": 0.7293,
+      "step": 30159
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.999885714285715e-06,
+      "loss": 1.4336,
+      "step": 30160
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.999617142857143e-06,
+      "loss": 0.9709,
+      "step": 30161
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.999348571428572e-06,
+      "loss": 0.808,
+      "step": 30162
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.99908e-06,
+      "loss": 0.8787,
+      "step": 30163
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.998811428571429e-06,
+      "loss": 0.9781,
+      "step": 30164
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.998542857142857e-06,
+      "loss": 1.3225,
+      "step": 30165
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.9982742857142855e-06,
+      "loss": 1.5069,
+      "step": 30166
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.9980057142857146e-06,
+      "loss": 1.6531,
+      "step": 30167
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.997737142857143e-06,
+      "loss": 1.1767,
+      "step": 30168
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.997468571428571e-06,
+      "loss": 0.791,
+      "step": 30169
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.9972e-06,
+      "loss": 1.7845,
+      "step": 30170
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.996931428571429e-06,
+      "loss": 0.9624,
+      "step": 30171
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.996662857142858e-06,
+      "loss": 1.0009,
+      "step": 30172
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.996394285714286e-06,
+      "loss": 1.1737,
+      "step": 30173
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.996125714285714e-06,
+      "loss": 1.387,
+      "step": 30174
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.995857142857143e-06,
+      "loss": 1.3071,
+      "step": 30175
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.9955885714285715e-06,
+      "loss": 1.3964,
+      "step": 30176
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.9953200000000006e-06,
+      "loss": 1.0749,
+      "step": 30177
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.995051428571429e-06,
+      "loss": 1.1667,
+      "step": 30178
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.994782857142857e-06,
+      "loss": 0.9696,
+      "step": 30179
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.994514285714286e-06,
+      "loss": 1.7072,
+      "step": 30180
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.994245714285714e-06,
+      "loss": 1.0824,
+      "step": 30181
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.993977142857143e-06,
+      "loss": 0.8365,
+      "step": 30182
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.993708571428571e-06,
+      "loss": 1.6784,
+      "step": 30183
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.99344e-06,
+      "loss": 1.484,
+      "step": 30184
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.993171428571429e-06,
+      "loss": 1.4371,
+      "step": 30185
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.9929028571428575e-06,
+      "loss": 1.3834,
+      "step": 30186
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 5.9926342857142866e-06,
+      "loss": 1.1641,
+      "step": 30187
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.992365714285715e-06,
+      "loss": 1.1817,
+      "step": 30188
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.992097142857143e-06,
+      "loss": 1.1858,
+      "step": 30189
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.991828571428572e-06,
+      "loss": 1.0763,
+      "step": 30190
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.99156e-06,
+      "loss": 1.0605,
+      "step": 30191
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.991291428571428e-06,
+      "loss": 0.7892,
+      "step": 30192
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.991022857142857e-06,
+      "loss": 1.4566,
+      "step": 30193
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9907542857142855e-06,
+      "loss": 0.8241,
+      "step": 30194
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9904857142857145e-06,
+      "loss": 1.1424,
+      "step": 30195
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.990217142857143e-06,
+      "loss": 1.0916,
+      "step": 30196
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.989948571428572e-06,
+      "loss": 1.1137,
+      "step": 30197
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.989680000000001e-06,
+      "loss": 1.0682,
+      "step": 30198
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.989411428571429e-06,
+      "loss": 1.0344,
+      "step": 30199
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.989142857142858e-06,
+      "loss": 0.8148,
+      "step": 30200
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.988874285714286e-06,
+      "loss": 1.1251,
+      "step": 30201
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.988605714285714e-06,
+      "loss": 1.3569,
+      "step": 30202
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.988337142857143e-06,
+      "loss": 1.1769,
+      "step": 30203
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9880685714285715e-06,
+      "loss": 1.1172,
+      "step": 30204
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9878000000000005e-06,
+      "loss": 0.8501,
+      "step": 30205
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.987531428571429e-06,
+      "loss": 1.3781,
+      "step": 30206
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.987262857142857e-06,
+      "loss": 1.3279,
+      "step": 30207
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.986994285714286e-06,
+      "loss": 1.4939,
+      "step": 30208
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.986725714285714e-06,
+      "loss": 1.1869,
+      "step": 30209
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.986457142857143e-06,
+      "loss": 1.5138,
+      "step": 30210
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.986188571428572e-06,
+      "loss": 1.3826,
+      "step": 30211
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.98592e-06,
+      "loss": 1.0517,
+      "step": 30212
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.985651428571429e-06,
+      "loss": 1.428,
+      "step": 30213
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9853828571428575e-06,
+      "loss": 1.1715,
+      "step": 30214
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.985114285714286e-06,
+      "loss": 1.2303,
+      "step": 30215
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.984845714285715e-06,
+      "loss": 1.3429,
+      "step": 30216
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.984577142857143e-06,
+      "loss": 1.0501,
+      "step": 30217
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.984308571428572e-06,
+      "loss": 1.1005,
+      "step": 30218
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.98404e-06,
+      "loss": 1.2402,
+      "step": 30219
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.983771428571428e-06,
+      "loss": 1.7072,
+      "step": 30220
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.983502857142857e-06,
+      "loss": 0.9447,
+      "step": 30221
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9832342857142854e-06,
+      "loss": 1.3911,
+      "step": 30222
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9829657142857145e-06,
+      "loss": 0.8263,
+      "step": 30223
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9826971428571435e-06,
+      "loss": 1.0312,
+      "step": 30224
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.982428571428572e-06,
+      "loss": 1.2059,
+      "step": 30225
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.982160000000001e-06,
+      "loss": 1.0865,
+      "step": 30226
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.981891428571429e-06,
+      "loss": 1.4411,
+      "step": 30227
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.981622857142858e-06,
+      "loss": 1.3104,
+      "step": 30228
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.981354285714286e-06,
+      "loss": 1.1378,
+      "step": 30229
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.981085714285714e-06,
+      "loss": 1.261,
+      "step": 30230
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.980817142857143e-06,
+      "loss": 1.0878,
+      "step": 30231
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9805485714285714e-06,
+      "loss": 1.2229,
+      "step": 30232
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.98028e-06,
+      "loss": 1.4171,
+      "step": 30233
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.980011428571429e-06,
+      "loss": 1.2533,
+      "step": 30234
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.979742857142857e-06,
+      "loss": 1.1031,
+      "step": 30235
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.979474285714286e-06,
+      "loss": 1.1767,
+      "step": 30236
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.979205714285715e-06,
+      "loss": 1.0982,
+      "step": 30237
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.978937142857143e-06,
+      "loss": 0.825,
+      "step": 30238
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.978668571428572e-06,
+      "loss": 1.4403,
+      "step": 30239
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9784e-06,
+      "loss": 1.1021,
+      "step": 30240
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.978131428571429e-06,
+      "loss": 1.3528,
+      "step": 30241
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.977862857142857e-06,
+      "loss": 1.2986,
+      "step": 30242
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.977594285714286e-06,
+      "loss": 1.3902,
+      "step": 30243
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.977325714285715e-06,
+      "loss": 1.2193,
+      "step": 30244
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.977057142857143e-06,
+      "loss": 1.1882,
+      "step": 30245
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.976788571428572e-06,
+      "loss": 1.0467,
+      "step": 30246
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.97652e-06,
+      "loss": 0.9087,
+      "step": 30247
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.976251428571428e-06,
+      "loss": 1.2576,
+      "step": 30248
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.975982857142857e-06,
+      "loss": 0.9413,
+      "step": 30249
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.975714285714286e-06,
+      "loss": 0.9921,
+      "step": 30250
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.975445714285715e-06,
+      "loss": 1.3166,
+      "step": 30251
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.975177142857143e-06,
+      "loss": 0.852,
+      "step": 30252
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.974908571428572e-06,
+      "loss": 0.6898,
+      "step": 30253
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.974640000000001e-06,
+      "loss": 1.3388,
+      "step": 30254
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.974371428571429e-06,
+      "loss": 1.5521,
+      "step": 30255
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.974102857142857e-06,
+      "loss": 0.9788,
+      "step": 30256
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.973834285714286e-06,
+      "loss": 1.461,
+      "step": 30257
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.973565714285714e-06,
+      "loss": 1.0549,
+      "step": 30258
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.973297142857143e-06,
+      "loss": 1.1441,
+      "step": 30259
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.973028571428571e-06,
+      "loss": 1.4354,
+      "step": 30260
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9727599999999995e-06,
+      "loss": 1.3331,
+      "step": 30261
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9724914285714286e-06,
+      "loss": 1.0751,
+      "step": 30262
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.972222857142858e-06,
+      "loss": 1.1897,
+      "step": 30263
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.971954285714287e-06,
+      "loss": 1.2834,
+      "step": 30264
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.971685714285715e-06,
+      "loss": 1.0434,
+      "step": 30265
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.971417142857143e-06,
+      "loss": 1.2653,
+      "step": 30266
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.971148571428572e-06,
+      "loss": 1.4624,
+      "step": 30267
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.97088e-06,
+      "loss": 1.2831,
+      "step": 30268
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.970611428571429e-06,
+      "loss": 1.3101,
+      "step": 30269
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.970342857142857e-06,
+      "loss": 1.2056,
+      "step": 30270
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9700742857142855e-06,
+      "loss": 1.0954,
+      "step": 30271
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9698057142857146e-06,
+      "loss": 0.8852,
+      "step": 30272
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.969537142857143e-06,
+      "loss": 0.7289,
+      "step": 30273
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.969268571428572e-06,
+      "loss": 0.9045,
+      "step": 30274
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.969e-06,
+      "loss": 1.1565,
+      "step": 30275
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.968731428571429e-06,
+      "loss": 1.0936,
+      "step": 30276
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.968462857142858e-06,
+      "loss": 1.3165,
+      "step": 30277
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.968194285714286e-06,
+      "loss": 1.4092,
+      "step": 30278
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.967925714285714e-06,
+      "loss": 1.079,
+      "step": 30279
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.967657142857143e-06,
+      "loss": 0.9283,
+      "step": 30280
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9673885714285715e-06,
+      "loss": 1.0809,
+      "step": 30281
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9671200000000006e-06,
+      "loss": 0.6889,
+      "step": 30282
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.966851428571429e-06,
+      "loss": 1.4436,
+      "step": 30283
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.966582857142857e-06,
+      "loss": 1.2229,
+      "step": 30284
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.966314285714286e-06,
+      "loss": 1.42,
+      "step": 30285
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.966045714285714e-06,
+      "loss": 1.102,
+      "step": 30286
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.965777142857143e-06,
+      "loss": 1.0817,
+      "step": 30287
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.965508571428571e-06,
+      "loss": 1.2844,
+      "step": 30288
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.96524e-06,
+      "loss": 1.2211,
+      "step": 30289
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.964971428571429e-06,
+      "loss": 1.3568,
+      "step": 30290
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9647028571428575e-06,
+      "loss": 1.3377,
+      "step": 30291
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9644342857142866e-06,
+      "loss": 0.869,
+      "step": 30292
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.964165714285715e-06,
+      "loss": 0.8948,
+      "step": 30293
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.963897142857143e-06,
+      "loss": 0.9282,
+      "step": 30294
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.963628571428572e-06,
+      "loss": 1.3255,
+      "step": 30295
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.96336e-06,
+      "loss": 1.2591,
+      "step": 30296
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.963091428571428e-06,
+      "loss": 1.1487,
+      "step": 30297
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.962822857142857e-06,
+      "loss": 1.1606,
+      "step": 30298
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9625542857142855e-06,
+      "loss": 1.1453,
+      "step": 30299
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9622857142857145e-06,
+      "loss": 1.0427,
+      "step": 30300
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.962017142857143e-06,
+      "loss": 1.3068,
+      "step": 30301
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.961748571428572e-06,
+      "loss": 1.2124,
+      "step": 30302
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.961480000000001e-06,
+      "loss": 1.3544,
+      "step": 30303
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.961211428571429e-06,
+      "loss": 1.5097,
+      "step": 30304
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.960942857142858e-06,
+      "loss": 1.0991,
+      "step": 30305
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.960674285714286e-06,
+      "loss": 1.0215,
+      "step": 30306
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.960405714285714e-06,
+      "loss": 1.0206,
+      "step": 30307
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.960137142857143e-06,
+      "loss": 1.1281,
+      "step": 30308
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9598685714285715e-06,
+      "loss": 1.3842,
+      "step": 30309
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9596000000000005e-06,
+      "loss": 0.7631,
+      "step": 30310
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.959331428571429e-06,
+      "loss": 1.194,
+      "step": 30311
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.959062857142857e-06,
+      "loss": 1.0371,
+      "step": 30312
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.958794285714286e-06,
+      "loss": 1.6873,
+      "step": 30313
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.958525714285714e-06,
+      "loss": 1.0688,
+      "step": 30314
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.958257142857143e-06,
+      "loss": 1.0281,
+      "step": 30315
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.957988571428572e-06,
+      "loss": 1.1987,
+      "step": 30316
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.95772e-06,
+      "loss": 1.3495,
+      "step": 30317
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.957451428571429e-06,
+      "loss": 1.2479,
+      "step": 30318
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9571828571428575e-06,
+      "loss": 0.8589,
+      "step": 30319
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.956914285714286e-06,
+      "loss": 1.0466,
+      "step": 30320
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.956645714285715e-06,
+      "loss": 0.9221,
+      "step": 30321
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.956377142857143e-06,
+      "loss": 0.9928,
+      "step": 30322
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.956108571428572e-06,
+      "loss": 1.1226,
+      "step": 30323
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.95584e-06,
+      "loss": 0.8833,
+      "step": 30324
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.955571428571428e-06,
+      "loss": 1.2581,
+      "step": 30325
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.955302857142857e-06,
+      "loss": 1.1786,
+      "step": 30326
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9550342857142854e-06,
+      "loss": 1.1486,
+      "step": 30327
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9547657142857145e-06,
+      "loss": 1.2024,
+      "step": 30328
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9544971428571435e-06,
+      "loss": 0.7544,
+      "step": 30329
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.954228571428572e-06,
+      "loss": 1.4612,
+      "step": 30330
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.953960000000001e-06,
+      "loss": 1.124,
+      "step": 30331
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.953691428571429e-06,
+      "loss": 1.5993,
+      "step": 30332
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.953422857142858e-06,
+      "loss": 1.5094,
+      "step": 30333
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.953154285714286e-06,
+      "loss": 1.9657,
+      "step": 30334
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.952885714285714e-06,
+      "loss": 0.9869,
+      "step": 30335
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.952617142857143e-06,
+      "loss": 1.0787,
+      "step": 30336
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9523485714285714e-06,
+      "loss": 1.1966,
+      "step": 30337
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9520800000000004e-06,
+      "loss": 1.0793,
+      "step": 30338
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.951811428571429e-06,
+      "loss": 1.1749,
+      "step": 30339
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.951542857142857e-06,
+      "loss": 1.0563,
+      "step": 30340
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.951274285714286e-06,
+      "loss": 1.0656,
+      "step": 30341
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.951005714285715e-06,
+      "loss": 1.4523,
+      "step": 30342
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.950737142857143e-06,
+      "loss": 1.2123,
+      "step": 30343
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.950468571428572e-06,
+      "loss": 1.1824,
+      "step": 30344
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.9502e-06,
+      "loss": 1.1131,
+      "step": 30345
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.949931428571429e-06,
+      "loss": 0.9431,
+      "step": 30346
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.949662857142857e-06,
+      "loss": 1.3391,
+      "step": 30347
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.949394285714286e-06,
+      "loss": 0.8476,
+      "step": 30348
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.949125714285715e-06,
+      "loss": 0.8449,
+      "step": 30349
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.948857142857143e-06,
+      "loss": 1.0953,
+      "step": 30350
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.948588571428572e-06,
+      "loss": 1.3202,
+      "step": 30351
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.94832e-06,
+      "loss": 1.246,
+      "step": 30352
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.948051428571428e-06,
+      "loss": 1.4676,
+      "step": 30353
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.947782857142857e-06,
+      "loss": 0.9462,
+      "step": 30354
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.947514285714286e-06,
+      "loss": 0.8245,
+      "step": 30355
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.947245714285715e-06,
+      "loss": 1.2186,
+      "step": 30356
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.946977142857143e-06,
+      "loss": 1.1441,
+      "step": 30357
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.946708571428572e-06,
+      "loss": 1.069,
+      "step": 30358
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.946440000000001e-06,
+      "loss": 1.2782,
+      "step": 30359
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.946171428571429e-06,
+      "loss": 0.7729,
+      "step": 30360
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.945902857142858e-06,
+      "loss": 1.4778,
+      "step": 30361
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 5.945634285714286e-06,
+      "loss": 1.1484,
+      "step": 30362
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.945365714285714e-06,
+      "loss": 1.262,
+      "step": 30363
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.945097142857143e-06,
+      "loss": 1.4161,
+      "step": 30364
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.944828571428571e-06,
+      "loss": 0.9534,
+      "step": 30365
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9445599999999995e-06,
+      "loss": 0.871,
+      "step": 30366
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9442914285714286e-06,
+      "loss": 1.3479,
+      "step": 30367
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.944022857142858e-06,
+      "loss": 0.696,
+      "step": 30368
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.943754285714287e-06,
+      "loss": 0.943,
+      "step": 30369
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.943485714285715e-06,
+      "loss": 0.7289,
+      "step": 30370
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.943217142857143e-06,
+      "loss": 1.0319,
+      "step": 30371
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.942948571428572e-06,
+      "loss": 1.0279,
+      "step": 30372
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.94268e-06,
+      "loss": 0.9997,
+      "step": 30373
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.942411428571429e-06,
+      "loss": 0.6385,
+      "step": 30374
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.942142857142857e-06,
+      "loss": 1.0071,
+      "step": 30375
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9418742857142855e-06,
+      "loss": 1.1311,
+      "step": 30376
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9416057142857146e-06,
+      "loss": 0.9435,
+      "step": 30377
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.941337142857143e-06,
+      "loss": 1.0958,
+      "step": 30378
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.941068571428572e-06,
+      "loss": 1.4374,
+      "step": 30379
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9408e-06,
+      "loss": 0.9153,
+      "step": 30380
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.940531428571429e-06,
+      "loss": 1.0236,
+      "step": 30381
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.940262857142858e-06,
+      "loss": 0.6808,
+      "step": 30382
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.939994285714286e-06,
+      "loss": 0.6544,
+      "step": 30383
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.939725714285714e-06,
+      "loss": 1.1347,
+      "step": 30384
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.939457142857143e-06,
+      "loss": 0.6995,
+      "step": 30385
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9391885714285715e-06,
+      "loss": 0.9001,
+      "step": 30386
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9389200000000006e-06,
+      "loss": 1.0407,
+      "step": 30387
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.938651428571429e-06,
+      "loss": 1.3092,
+      "step": 30388
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.938382857142857e-06,
+      "loss": 1.1976,
+      "step": 30389
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.938114285714286e-06,
+      "loss": 0.8069,
+      "step": 30390
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.937845714285714e-06,
+      "loss": 1.0155,
+      "step": 30391
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.937577142857143e-06,
+      "loss": 1.0324,
+      "step": 30392
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.937308571428571e-06,
+      "loss": 0.8602,
+      "step": 30393
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.93704e-06,
+      "loss": 1.2941,
+      "step": 30394
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.936771428571429e-06,
+      "loss": 1.1817,
+      "step": 30395
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9365028571428575e-06,
+      "loss": 1.3144,
+      "step": 30396
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9362342857142866e-06,
+      "loss": 0.9135,
+      "step": 30397
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.935965714285715e-06,
+      "loss": 1.2287,
+      "step": 30398
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.935697142857143e-06,
+      "loss": 0.8587,
+      "step": 30399
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.935428571428572e-06,
+      "loss": 1.5308,
+      "step": 30400
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.93516e-06,
+      "loss": 1.6064,
+      "step": 30401
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.934891428571429e-06,
+      "loss": 0.9621,
+      "step": 30402
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.934622857142857e-06,
+      "loss": 1.2764,
+      "step": 30403
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9343542857142855e-06,
+      "loss": 1.2518,
+      "step": 30404
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9340857142857145e-06,
+      "loss": 0.936,
+      "step": 30405
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.933817142857143e-06,
+      "loss": 1.0752,
+      "step": 30406
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.933548571428571e-06,
+      "loss": 0.9599,
+      "step": 30407
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.933280000000001e-06,
+      "loss": 1.5655,
+      "step": 30408
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.933011428571429e-06,
+      "loss": 1.4616,
+      "step": 30409
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.932742857142858e-06,
+      "loss": 1.0908,
+      "step": 30410
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.932474285714286e-06,
+      "loss": 1.215,
+      "step": 30411
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.932205714285714e-06,
+      "loss": 1.0001,
+      "step": 30412
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.931937142857143e-06,
+      "loss": 1.5545,
+      "step": 30413
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9316685714285715e-06,
+      "loss": 1.0798,
+      "step": 30414
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9314000000000005e-06,
+      "loss": 1.0722,
+      "step": 30415
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.931131428571429e-06,
+      "loss": 1.1573,
+      "step": 30416
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.930862857142857e-06,
+      "loss": 1.1004,
+      "step": 30417
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.930594285714286e-06,
+      "loss": 1.302,
+      "step": 30418
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.930325714285714e-06,
+      "loss": 1.087,
+      "step": 30419
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.930057142857143e-06,
+      "loss": 1.3793,
+      "step": 30420
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.929788571428572e-06,
+      "loss": 1.1315,
+      "step": 30421
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.92952e-06,
+      "loss": 1.0701,
+      "step": 30422
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.929251428571429e-06,
+      "loss": 1.0718,
+      "step": 30423
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9289828571428575e-06,
+      "loss": 1.2948,
+      "step": 30424
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9287142857142865e-06,
+      "loss": 1.2589,
+      "step": 30425
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.928445714285715e-06,
+      "loss": 1.3097,
+      "step": 30426
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.928177142857143e-06,
+      "loss": 0.8692,
+      "step": 30427
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.927908571428572e-06,
+      "loss": 0.8267,
+      "step": 30428
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.92764e-06,
+      "loss": 0.9203,
+      "step": 30429
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.927371428571428e-06,
+      "loss": 0.9468,
+      "step": 30430
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.927102857142857e-06,
+      "loss": 1.0238,
+      "step": 30431
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9268342857142854e-06,
+      "loss": 1.409,
+      "step": 30432
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9265657142857145e-06,
+      "loss": 1.1286,
+      "step": 30433
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9262971428571435e-06,
+      "loss": 0.7808,
+      "step": 30434
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.926028571428572e-06,
+      "loss": 1.5167,
+      "step": 30435
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.925760000000001e-06,
+      "loss": 1.4729,
+      "step": 30436
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.925491428571429e-06,
+      "loss": 1.265,
+      "step": 30437
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.925222857142858e-06,
+      "loss": 0.7647,
+      "step": 30438
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.924954285714286e-06,
+      "loss": 0.8735,
+      "step": 30439
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.924685714285714e-06,
+      "loss": 1.0169,
+      "step": 30440
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.924417142857143e-06,
+      "loss": 0.9467,
+      "step": 30441
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9241485714285714e-06,
+      "loss": 1.3862,
+      "step": 30442
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9238800000000004e-06,
+      "loss": 1.8163,
+      "step": 30443
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.923611428571429e-06,
+      "loss": 1.3295,
+      "step": 30444
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.923342857142857e-06,
+      "loss": 1.7065,
+      "step": 30445
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.923074285714286e-06,
+      "loss": 0.8731,
+      "step": 30446
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.922805714285715e-06,
+      "loss": 0.8894,
+      "step": 30447
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.922537142857143e-06,
+      "loss": 0.9425,
+      "step": 30448
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.922268571428572e-06,
+      "loss": 0.9231,
+      "step": 30449
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.922e-06,
+      "loss": 1.1184,
+      "step": 30450
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.921731428571429e-06,
+      "loss": 1.0805,
+      "step": 30451
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.921462857142857e-06,
+      "loss": 1.2707,
+      "step": 30452
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.921194285714286e-06,
+      "loss": 1.2829,
+      "step": 30453
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.920925714285715e-06,
+      "loss": 0.9843,
+      "step": 30454
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.920657142857143e-06,
+      "loss": 0.8269,
+      "step": 30455
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.920388571428572e-06,
+      "loss": 1.3215,
+      "step": 30456
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.92012e-06,
+      "loss": 1.1475,
+      "step": 30457
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.919851428571428e-06,
+      "loss": 0.9944,
+      "step": 30458
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.919582857142857e-06,
+      "loss": 1.142,
+      "step": 30459
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.919314285714285e-06,
+      "loss": 1.1207,
+      "step": 30460
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.919045714285715e-06,
+      "loss": 1.3583,
+      "step": 30461
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.918777142857143e-06,
+      "loss": 0.9523,
+      "step": 30462
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.918508571428572e-06,
+      "loss": 1.547,
+      "step": 30463
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.918240000000001e-06,
+      "loss": 0.8966,
+      "step": 30464
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.917971428571429e-06,
+      "loss": 1.334,
+      "step": 30465
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.917702857142858e-06,
+      "loss": 1.225,
+      "step": 30466
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.917434285714286e-06,
+      "loss": 1.0896,
+      "step": 30467
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.917165714285714e-06,
+      "loss": 1.4579,
+      "step": 30468
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.916897142857143e-06,
+      "loss": 1.3086,
+      "step": 30469
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.916628571428571e-06,
+      "loss": 1.086,
+      "step": 30470
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9163599999999995e-06,
+      "loss": 0.9247,
+      "step": 30471
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9160914285714286e-06,
+      "loss": 1.0598,
+      "step": 30472
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.915822857142857e-06,
+      "loss": 1.1323,
+      "step": 30473
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.915554285714287e-06,
+      "loss": 1.2492,
+      "step": 30474
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.915285714285715e-06,
+      "loss": 1.3448,
+      "step": 30475
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.915017142857143e-06,
+      "loss": 1.3949,
+      "step": 30476
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.914748571428572e-06,
+      "loss": 0.8192,
+      "step": 30477
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.91448e-06,
+      "loss": 1.1559,
+      "step": 30478
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.914211428571429e-06,
+      "loss": 1.3691,
+      "step": 30479
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.913942857142857e-06,
+      "loss": 0.875,
+      "step": 30480
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9136742857142855e-06,
+      "loss": 0.691,
+      "step": 30481
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9134057142857146e-06,
+      "loss": 1.0064,
+      "step": 30482
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.913137142857143e-06,
+      "loss": 0.9623,
+      "step": 30483
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.912868571428572e-06,
+      "loss": 1.5072,
+      "step": 30484
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9126e-06,
+      "loss": 1.175,
+      "step": 30485
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.912331428571428e-06,
+      "loss": 0.8077,
+      "step": 30486
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.912062857142858e-06,
+      "loss": 1.034,
+      "step": 30487
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.911794285714286e-06,
+      "loss": 0.8883,
+      "step": 30488
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.911525714285715e-06,
+      "loss": 0.9773,
+      "step": 30489
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.911257142857143e-06,
+      "loss": 1.0983,
+      "step": 30490
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9109885714285715e-06,
+      "loss": 0.7683,
+      "step": 30491
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9107200000000006e-06,
+      "loss": 1.1649,
+      "step": 30492
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.910451428571429e-06,
+      "loss": 1.3373,
+      "step": 30493
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.910182857142857e-06,
+      "loss": 1.0302,
+      "step": 30494
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.909914285714286e-06,
+      "loss": 1.0164,
+      "step": 30495
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.909645714285714e-06,
+      "loss": 1.3621,
+      "step": 30496
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.909377142857143e-06,
+      "loss": 0.9519,
+      "step": 30497
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.909108571428571e-06,
+      "loss": 1.4949,
+      "step": 30498
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9088399999999995e-06,
+      "loss": 1.2009,
+      "step": 30499
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.908571428571429e-06,
+      "loss": 1.2201,
+      "step": 30500
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9083028571428575e-06,
+      "loss": 1.283,
+      "step": 30501
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9080342857142866e-06,
+      "loss": 1.1996,
+      "step": 30502
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.907765714285715e-06,
+      "loss": 1.2218,
+      "step": 30503
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.907497142857143e-06,
+      "loss": 0.8137,
+      "step": 30504
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.907228571428572e-06,
+      "loss": 1.1878,
+      "step": 30505
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.90696e-06,
+      "loss": 1.1913,
+      "step": 30506
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.906691428571429e-06,
+      "loss": 1.343,
+      "step": 30507
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.906422857142857e-06,
+      "loss": 0.7649,
+      "step": 30508
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9061542857142855e-06,
+      "loss": 0.7934,
+      "step": 30509
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9058857142857145e-06,
+      "loss": 1.2423,
+      "step": 30510
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.905617142857143e-06,
+      "loss": 1.2227,
+      "step": 30511
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.905348571428572e-06,
+      "loss": 0.7231,
+      "step": 30512
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.90508e-06,
+      "loss": 1.4803,
+      "step": 30513
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.904811428571429e-06,
+      "loss": 0.7206,
+      "step": 30514
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.904542857142858e-06,
+      "loss": 1.136,
+      "step": 30515
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.904274285714286e-06,
+      "loss": 1.483,
+      "step": 30516
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.904005714285714e-06,
+      "loss": 1.3681,
+      "step": 30517
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.903737142857143e-06,
+      "loss": 1.1339,
+      "step": 30518
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9034685714285715e-06,
+      "loss": 1.2997,
+      "step": 30519
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9032000000000005e-06,
+      "loss": 1.0121,
+      "step": 30520
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.902931428571429e-06,
+      "loss": 1.0862,
+      "step": 30521
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.902662857142857e-06,
+      "loss": 0.9758,
+      "step": 30522
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.902394285714286e-06,
+      "loss": 0.6212,
+      "step": 30523
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.902125714285714e-06,
+      "loss": 1.3827,
+      "step": 30524
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.901857142857143e-06,
+      "loss": 0.7502,
+      "step": 30525
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.901588571428571e-06,
+      "loss": 1.1555,
+      "step": 30526
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.90132e-06,
+      "loss": 0.8456,
+      "step": 30527
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.901051428571429e-06,
+      "loss": 0.8107,
+      "step": 30528
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9007828571428575e-06,
+      "loss": 1.0888,
+      "step": 30529
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.9005142857142865e-06,
+      "loss": 1.3221,
+      "step": 30530
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.900245714285715e-06,
+      "loss": 1.592,
+      "step": 30531
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.899977142857143e-06,
+      "loss": 0.975,
+      "step": 30532
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.899708571428572e-06,
+      "loss": 1.2434,
+      "step": 30533
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.89944e-06,
+      "loss": 1.4735,
+      "step": 30534
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.899171428571428e-06,
+      "loss": 1.4074,
+      "step": 30535
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.898902857142857e-06,
+      "loss": 0.8683,
+      "step": 30536
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 5.8986342857142854e-06,
+      "loss": 1.2042,
+      "step": 30537
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8983657142857144e-06,
+      "loss": 1.0632,
+      "step": 30538
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.898097142857143e-06,
+      "loss": 1.8324,
+      "step": 30539
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.897828571428572e-06,
+      "loss": 1.3628,
+      "step": 30540
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.897560000000001e-06,
+      "loss": 1.1634,
+      "step": 30541
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.897291428571429e-06,
+      "loss": 0.8558,
+      "step": 30542
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.897022857142858e-06,
+      "loss": 0.7218,
+      "step": 30543
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.896754285714286e-06,
+      "loss": 1.0858,
+      "step": 30544
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.896485714285714e-06,
+      "loss": 1.3583,
+      "step": 30545
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.896217142857143e-06,
+      "loss": 1.2507,
+      "step": 30546
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.895948571428571e-06,
+      "loss": 0.9526,
+      "step": 30547
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8956800000000004e-06,
+      "loss": 1.0979,
+      "step": 30548
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.895411428571429e-06,
+      "loss": 1.1201,
+      "step": 30549
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.895142857142857e-06,
+      "loss": 0.8866,
+      "step": 30550
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.894874285714286e-06,
+      "loss": 0.621,
+      "step": 30551
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.894605714285714e-06,
+      "loss": 1.6197,
+      "step": 30552
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.894337142857144e-06,
+      "loss": 1.0344,
+      "step": 30553
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.894068571428572e-06,
+      "loss": 1.1716,
+      "step": 30554
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8938e-06,
+      "loss": 0.9134,
+      "step": 30555
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.893531428571429e-06,
+      "loss": 1.2283,
+      "step": 30556
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.893262857142857e-06,
+      "loss": 1.0351,
+      "step": 30557
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.892994285714286e-06,
+      "loss": 0.758,
+      "step": 30558
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.892725714285715e-06,
+      "loss": 1.2724,
+      "step": 30559
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.892457142857143e-06,
+      "loss": 0.905,
+      "step": 30560
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.892188571428572e-06,
+      "loss": 1.4447,
+      "step": 30561
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.89192e-06,
+      "loss": 1.3223,
+      "step": 30562
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.891651428571428e-06,
+      "loss": 1.0251,
+      "step": 30563
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.891382857142857e-06,
+      "loss": 1.1677,
+      "step": 30564
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.891114285714285e-06,
+      "loss": 1.4023,
+      "step": 30565
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.890845714285714e-06,
+      "loss": 0.9129,
+      "step": 30566
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.890577142857143e-06,
+      "loss": 0.8886,
+      "step": 30567
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.890308571428572e-06,
+      "loss": 1.0972,
+      "step": 30568
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.890040000000001e-06,
+      "loss": 1.2455,
+      "step": 30569
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.889771428571429e-06,
+      "loss": 1.3005,
+      "step": 30570
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.889502857142858e-06,
+      "loss": 1.0404,
+      "step": 30571
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.889234285714286e-06,
+      "loss": 1.3669,
+      "step": 30572
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.888965714285714e-06,
+      "loss": 1.0028,
+      "step": 30573
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.888697142857143e-06,
+      "loss": 0.8674,
+      "step": 30574
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.888428571428571e-06,
+      "loss": 1.0264,
+      "step": 30575
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.88816e-06,
+      "loss": 1.4277,
+      "step": 30576
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8878914285714286e-06,
+      "loss": 0.6293,
+      "step": 30577
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.887622857142857e-06,
+      "loss": 1.0549,
+      "step": 30578
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.887354285714286e-06,
+      "loss": 1.0525,
+      "step": 30579
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.887085714285715e-06,
+      "loss": 1.2977,
+      "step": 30580
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.886817142857143e-06,
+      "loss": 1.2783,
+      "step": 30581
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.886548571428572e-06,
+      "loss": 1.0598,
+      "step": 30582
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.88628e-06,
+      "loss": 1.3804,
+      "step": 30583
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.886011428571429e-06,
+      "loss": 1.2757,
+      "step": 30584
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.885742857142857e-06,
+      "loss": 1.3005,
+      "step": 30585
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8854742857142855e-06,
+      "loss": 1.0132,
+      "step": 30586
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8852057142857146e-06,
+      "loss": 1.215,
+      "step": 30587
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.884937142857143e-06,
+      "loss": 1.0835,
+      "step": 30588
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.884668571428572e-06,
+      "loss": 1.0252,
+      "step": 30589
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8844e-06,
+      "loss": 1.0444,
+      "step": 30590
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.884131428571428e-06,
+      "loss": 1.0587,
+      "step": 30591
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.883862857142857e-06,
+      "loss": 1.5911,
+      "step": 30592
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.883594285714286e-06,
+      "loss": 0.9708,
+      "step": 30593
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.883325714285715e-06,
+      "loss": 1.659,
+      "step": 30594
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.883057142857143e-06,
+      "loss": 1.0451,
+      "step": 30595
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8827885714285715e-06,
+      "loss": 1.3329,
+      "step": 30596
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8825200000000006e-06,
+      "loss": 1.0687,
+      "step": 30597
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.882251428571429e-06,
+      "loss": 1.3456,
+      "step": 30598
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.881982857142858e-06,
+      "loss": 1.3954,
+      "step": 30599
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.881714285714286e-06,
+      "loss": 0.9403,
+      "step": 30600
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.881445714285714e-06,
+      "loss": 0.8777,
+      "step": 30601
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.881177142857143e-06,
+      "loss": 0.9302,
+      "step": 30602
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.880908571428571e-06,
+      "loss": 0.8128,
+      "step": 30603
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8806399999999995e-06,
+      "loss": 1.201,
+      "step": 30604
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8803714285714285e-06,
+      "loss": 0.9491,
+      "step": 30605
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8801028571428575e-06,
+      "loss": 1.0166,
+      "step": 30606
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8798342857142866e-06,
+      "loss": 1.2467,
+      "step": 30607
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.879565714285715e-06,
+      "loss": 1.1726,
+      "step": 30608
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.879297142857143e-06,
+      "loss": 0.8145,
+      "step": 30609
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.879028571428572e-06,
+      "loss": 1.1186,
+      "step": 30610
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.87876e-06,
+      "loss": 1.0745,
+      "step": 30611
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.878491428571429e-06,
+      "loss": 0.7807,
+      "step": 30612
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.878222857142857e-06,
+      "loss": 1.5508,
+      "step": 30613
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8779542857142855e-06,
+      "loss": 1.6071,
+      "step": 30614
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8776857142857145e-06,
+      "loss": 0.8271,
+      "step": 30615
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.877417142857143e-06,
+      "loss": 1.0861,
+      "step": 30616
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.877148571428572e-06,
+      "loss": 0.9412,
+      "step": 30617
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.87688e-06,
+      "loss": 1.0209,
+      "step": 30618
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.876611428571429e-06,
+      "loss": 1.0454,
+      "step": 30619
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.876342857142858e-06,
+      "loss": 1.0724,
+      "step": 30620
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.876074285714286e-06,
+      "loss": 1.1136,
+      "step": 30621
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.875805714285714e-06,
+      "loss": 1.4667,
+      "step": 30622
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.875537142857143e-06,
+      "loss": 1.5127,
+      "step": 30623
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8752685714285715e-06,
+      "loss": 1.273,
+      "step": 30624
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8750000000000005e-06,
+      "loss": 1.3068,
+      "step": 30625
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.874731428571429e-06,
+      "loss": 1.1316,
+      "step": 30626
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.874462857142857e-06,
+      "loss": 1.3114,
+      "step": 30627
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.874194285714286e-06,
+      "loss": 0.6806,
+      "step": 30628
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.873925714285714e-06,
+      "loss": 1.4449,
+      "step": 30629
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.873657142857143e-06,
+      "loss": 1.2791,
+      "step": 30630
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.873388571428571e-06,
+      "loss": 0.6097,
+      "step": 30631
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.87312e-06,
+      "loss": 0.9084,
+      "step": 30632
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.872851428571429e-06,
+      "loss": 1.3293,
+      "step": 30633
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8725828571428575e-06,
+      "loss": 1.2651,
+      "step": 30634
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8723142857142865e-06,
+      "loss": 1.2662,
+      "step": 30635
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.872045714285715e-06,
+      "loss": 1.1388,
+      "step": 30636
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.871777142857143e-06,
+      "loss": 1.0728,
+      "step": 30637
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.871508571428572e-06,
+      "loss": 1.1431,
+      "step": 30638
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.87124e-06,
+      "loss": 1.2872,
+      "step": 30639
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.870971428571429e-06,
+      "loss": 0.7659,
+      "step": 30640
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.870702857142857e-06,
+      "loss": 1.1111,
+      "step": 30641
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8704342857142854e-06,
+      "loss": 0.8264,
+      "step": 30642
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8701657142857144e-06,
+      "loss": 1.0495,
+      "step": 30643
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.869897142857143e-06,
+      "loss": 1.4604,
+      "step": 30644
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.869628571428572e-06,
+      "loss": 0.9657,
+      "step": 30645
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.869360000000001e-06,
+      "loss": 0.8033,
+      "step": 30646
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.869091428571429e-06,
+      "loss": 1.0521,
+      "step": 30647
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.868822857142858e-06,
+      "loss": 0.8925,
+      "step": 30648
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.868554285714286e-06,
+      "loss": 0.9409,
+      "step": 30649
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.868285714285714e-06,
+      "loss": 1.183,
+      "step": 30650
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.868017142857143e-06,
+      "loss": 1.4067,
+      "step": 30651
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.867748571428571e-06,
+      "loss": 1.4875,
+      "step": 30652
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8674800000000004e-06,
+      "loss": 1.0043,
+      "step": 30653
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.867211428571429e-06,
+      "loss": 1.0548,
+      "step": 30654
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.866942857142857e-06,
+      "loss": 0.7508,
+      "step": 30655
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.866674285714286e-06,
+      "loss": 1.5537,
+      "step": 30656
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.866405714285714e-06,
+      "loss": 1.4658,
+      "step": 30657
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.866137142857143e-06,
+      "loss": 1.043,
+      "step": 30658
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.865868571428572e-06,
+      "loss": 1.1714,
+      "step": 30659
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8656e-06,
+      "loss": 1.2769,
+      "step": 30660
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.865331428571429e-06,
+      "loss": 0.9452,
+      "step": 30661
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.865062857142857e-06,
+      "loss": 1.3416,
+      "step": 30662
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8647942857142864e-06,
+      "loss": 1.3881,
+      "step": 30663
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.864525714285715e-06,
+      "loss": 1.033,
+      "step": 30664
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.864257142857143e-06,
+      "loss": 1.1118,
+      "step": 30665
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.863988571428572e-06,
+      "loss": 1.0839,
+      "step": 30666
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.86372e-06,
+      "loss": 1.2065,
+      "step": 30667
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.863451428571428e-06,
+      "loss": 0.8427,
+      "step": 30668
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.863182857142857e-06,
+      "loss": 1.0306,
+      "step": 30669
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.862914285714285e-06,
+      "loss": 1.0433,
+      "step": 30670
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.862645714285714e-06,
+      "loss": 1.1181,
+      "step": 30671
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.862377142857143e-06,
+      "loss": 0.9287,
+      "step": 30672
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.862108571428572e-06,
+      "loss": 1.2299,
+      "step": 30673
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.861840000000001e-06,
+      "loss": 1.2659,
+      "step": 30674
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.861571428571429e-06,
+      "loss": 1.0655,
+      "step": 30675
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.861302857142858e-06,
+      "loss": 1.0346,
+      "step": 30676
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.861034285714286e-06,
+      "loss": 1.2347,
+      "step": 30677
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.860765714285714e-06,
+      "loss": 1.1045,
+      "step": 30678
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.860497142857143e-06,
+      "loss": 1.0441,
+      "step": 30679
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.860228571428571e-06,
+      "loss": 1.6486,
+      "step": 30680
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.85996e-06,
+      "loss": 1.3497,
+      "step": 30681
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8596914285714286e-06,
+      "loss": 1.3415,
+      "step": 30682
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.859422857142857e-06,
+      "loss": 1.09,
+      "step": 30683
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.859154285714286e-06,
+      "loss": 1.6396,
+      "step": 30684
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.858885714285715e-06,
+      "loss": 0.7954,
+      "step": 30685
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.858617142857144e-06,
+      "loss": 1.2784,
+      "step": 30686
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.858348571428572e-06,
+      "loss": 1.1012,
+      "step": 30687
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.85808e-06,
+      "loss": 1.3917,
+      "step": 30688
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.857811428571429e-06,
+      "loss": 1.2271,
+      "step": 30689
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.857542857142857e-06,
+      "loss": 1.0117,
+      "step": 30690
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8572742857142855e-06,
+      "loss": 1.1161,
+      "step": 30691
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8570057142857146e-06,
+      "loss": 1.2034,
+      "step": 30692
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.856737142857143e-06,
+      "loss": 1.0677,
+      "step": 30693
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.856468571428572e-06,
+      "loss": 0.6929,
+      "step": 30694
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8562e-06,
+      "loss": 0.9623,
+      "step": 30695
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.855931428571428e-06,
+      "loss": 1.3382,
+      "step": 30696
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.855662857142857e-06,
+      "loss": 1.0342,
+      "step": 30697
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.855394285714286e-06,
+      "loss": 0.8386,
+      "step": 30698
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.855125714285715e-06,
+      "loss": 1.235,
+      "step": 30699
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.854857142857143e-06,
+      "loss": 1.0858,
+      "step": 30700
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8545885714285715e-06,
+      "loss": 1.0324,
+      "step": 30701
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8543200000000006e-06,
+      "loss": 1.2916,
+      "step": 30702
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.854051428571429e-06,
+      "loss": 1.0759,
+      "step": 30703
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.853782857142858e-06,
+      "loss": 1.087,
+      "step": 30704
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.853514285714286e-06,
+      "loss": 1.1965,
+      "step": 30705
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.853245714285714e-06,
+      "loss": 1.3948,
+      "step": 30706
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.852977142857143e-06,
+      "loss": 1.2193,
+      "step": 30707
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.852708571428571e-06,
+      "loss": 1.0222,
+      "step": 30708
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8524399999999995e-06,
+      "loss": 1.5222,
+      "step": 30709
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8521714285714285e-06,
+      "loss": 1.4547,
+      "step": 30710
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8519028571428575e-06,
+      "loss": 1.0134,
+      "step": 30711
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 5.8516342857142866e-06,
+      "loss": 1.1379,
+      "step": 30712
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.851365714285715e-06,
+      "loss": 1.2329,
+      "step": 30713
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.851097142857143e-06,
+      "loss": 1.113,
+      "step": 30714
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.850828571428572e-06,
+      "loss": 0.9771,
+      "step": 30715
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.85056e-06,
+      "loss": 1.2994,
+      "step": 30716
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.850291428571429e-06,
+      "loss": 1.1663,
+      "step": 30717
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.850022857142857e-06,
+      "loss": 1.0668,
+      "step": 30718
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8497542857142855e-06,
+      "loss": 1.2549,
+      "step": 30719
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8494857142857145e-06,
+      "loss": 1.5462,
+      "step": 30720
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.849217142857143e-06,
+      "loss": 0.9514,
+      "step": 30721
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.848948571428572e-06,
+      "loss": 1.2894,
+      "step": 30722
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.84868e-06,
+      "loss": 0.9647,
+      "step": 30723
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.848411428571429e-06,
+      "loss": 0.9695,
+      "step": 30724
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.848142857142858e-06,
+      "loss": 1.1732,
+      "step": 30725
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.847874285714286e-06,
+      "loss": 1.1545,
+      "step": 30726
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.847605714285715e-06,
+      "loss": 1.6766,
+      "step": 30727
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.847337142857143e-06,
+      "loss": 0.9974,
+      "step": 30728
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8470685714285715e-06,
+      "loss": 1.2929,
+      "step": 30729
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8468000000000005e-06,
+      "loss": 1.1371,
+      "step": 30730
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.846531428571429e-06,
+      "loss": 1.0119,
+      "step": 30731
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.846262857142857e-06,
+      "loss": 1.4041,
+      "step": 30732
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.845994285714286e-06,
+      "loss": 0.6848,
+      "step": 30733
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.845725714285714e-06,
+      "loss": 1.1357,
+      "step": 30734
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.845457142857143e-06,
+      "loss": 1.2152,
+      "step": 30735
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.845188571428571e-06,
+      "loss": 1.289,
+      "step": 30736
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.84492e-06,
+      "loss": 1.3025,
+      "step": 30737
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.844651428571429e-06,
+      "loss": 1.6382,
+      "step": 30738
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8443828571428575e-06,
+      "loss": 1.2857,
+      "step": 30739
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8441142857142865e-06,
+      "loss": 0.894,
+      "step": 30740
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.843845714285715e-06,
+      "loss": 1.3717,
+      "step": 30741
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.843577142857143e-06,
+      "loss": 1.3248,
+      "step": 30742
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.843308571428572e-06,
+      "loss": 1.0544,
+      "step": 30743
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.84304e-06,
+      "loss": 1.2664,
+      "step": 30744
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.842771428571429e-06,
+      "loss": 1.4166,
+      "step": 30745
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.842502857142857e-06,
+      "loss": 1.2133,
+      "step": 30746
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8422342857142854e-06,
+      "loss": 1.1981,
+      "step": 30747
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8419657142857144e-06,
+      "loss": 0.9618,
+      "step": 30748
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.841697142857143e-06,
+      "loss": 1.3803,
+      "step": 30749
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.841428571428572e-06,
+      "loss": 0.9116,
+      "step": 30750
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.841160000000001e-06,
+      "loss": 1.3815,
+      "step": 30751
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.840891428571429e-06,
+      "loss": 1.2791,
+      "step": 30752
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.840622857142858e-06,
+      "loss": 1.0926,
+      "step": 30753
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.840354285714286e-06,
+      "loss": 0.6797,
+      "step": 30754
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.840085714285714e-06,
+      "loss": 1.0013,
+      "step": 30755
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.839817142857143e-06,
+      "loss": 0.9621,
+      "step": 30756
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.839548571428571e-06,
+      "loss": 1.1137,
+      "step": 30757
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8392800000000004e-06,
+      "loss": 0.9659,
+      "step": 30758
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.839011428571429e-06,
+      "loss": 1.1179,
+      "step": 30759
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.838742857142857e-06,
+      "loss": 1.5386,
+      "step": 30760
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.838474285714286e-06,
+      "loss": 1.1105,
+      "step": 30761
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.838205714285714e-06,
+      "loss": 1.3743,
+      "step": 30762
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.837937142857143e-06,
+      "loss": 1.1961,
+      "step": 30763
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.837668571428572e-06,
+      "loss": 1.4039,
+      "step": 30764
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8374e-06,
+      "loss": 1.6502,
+      "step": 30765
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.837131428571429e-06,
+      "loss": 0.8882,
+      "step": 30766
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.836862857142857e-06,
+      "loss": 1.4033,
+      "step": 30767
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8365942857142864e-06,
+      "loss": 1.3773,
+      "step": 30768
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.836325714285715e-06,
+      "loss": 1.1152,
+      "step": 30769
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.836057142857143e-06,
+      "loss": 0.9981,
+      "step": 30770
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.835788571428572e-06,
+      "loss": 0.8993,
+      "step": 30771
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.83552e-06,
+      "loss": 1.0622,
+      "step": 30772
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.835251428571429e-06,
+      "loss": 1.3155,
+      "step": 30773
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.834982857142857e-06,
+      "loss": 1.1054,
+      "step": 30774
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.834714285714285e-06,
+      "loss": 1.3888,
+      "step": 30775
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.834445714285714e-06,
+      "loss": 1.0907,
+      "step": 30776
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.834177142857143e-06,
+      "loss": 1.1237,
+      "step": 30777
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.833908571428572e-06,
+      "loss": 1.1263,
+      "step": 30778
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.833640000000001e-06,
+      "loss": 1.1487,
+      "step": 30779
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.833371428571429e-06,
+      "loss": 1.0761,
+      "step": 30780
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.833102857142858e-06,
+      "loss": 1.2476,
+      "step": 30781
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.832834285714286e-06,
+      "loss": 0.8938,
+      "step": 30782
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.832565714285714e-06,
+      "loss": 1.1275,
+      "step": 30783
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.832297142857143e-06,
+      "loss": 0.8476,
+      "step": 30784
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.832028571428571e-06,
+      "loss": 1.2268,
+      "step": 30785
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.83176e-06,
+      "loss": 1.23,
+      "step": 30786
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8314914285714286e-06,
+      "loss": 0.8608,
+      "step": 30787
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.831222857142857e-06,
+      "loss": 1.2154,
+      "step": 30788
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.830954285714286e-06,
+      "loss": 1.1137,
+      "step": 30789
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.830685714285715e-06,
+      "loss": 1.1008,
+      "step": 30790
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.830417142857144e-06,
+      "loss": 1.2277,
+      "step": 30791
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.830148571428572e-06,
+      "loss": 1.4302,
+      "step": 30792
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.82988e-06,
+      "loss": 1.2077,
+      "step": 30793
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.829611428571429e-06,
+      "loss": 1.1108,
+      "step": 30794
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.829342857142857e-06,
+      "loss": 1.4811,
+      "step": 30795
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8290742857142855e-06,
+      "loss": 0.6615,
+      "step": 30796
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8288057142857146e-06,
+      "loss": 0.8941,
+      "step": 30797
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.828537142857143e-06,
+      "loss": 1.2952,
+      "step": 30798
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.828268571428572e-06,
+      "loss": 1.218,
+      "step": 30799
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.828e-06,
+      "loss": 1.214,
+      "step": 30800
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.827731428571428e-06,
+      "loss": 0.9668,
+      "step": 30801
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.827462857142857e-06,
+      "loss": 1.2312,
+      "step": 30802
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.827194285714286e-06,
+      "loss": 0.9234,
+      "step": 30803
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.826925714285715e-06,
+      "loss": 1.1066,
+      "step": 30804
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.826657142857143e-06,
+      "loss": 1.2089,
+      "step": 30805
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8263885714285715e-06,
+      "loss": 1.0435,
+      "step": 30806
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8261200000000006e-06,
+      "loss": 1.4668,
+      "step": 30807
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.825851428571429e-06,
+      "loss": 1.1167,
+      "step": 30808
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.825582857142858e-06,
+      "loss": 0.9072,
+      "step": 30809
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.825314285714286e-06,
+      "loss": 1.499,
+      "step": 30810
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.825045714285714e-06,
+      "loss": 1.3279,
+      "step": 30811
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.824777142857143e-06,
+      "loss": 1.1858,
+      "step": 30812
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.824508571428571e-06,
+      "loss": 1.4061,
+      "step": 30813
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.82424e-06,
+      "loss": 1.2504,
+      "step": 30814
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8239714285714285e-06,
+      "loss": 1.3134,
+      "step": 30815
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8237028571428575e-06,
+      "loss": 0.959,
+      "step": 30816
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8234342857142866e-06,
+      "loss": 1.119,
+      "step": 30817
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.823165714285715e-06,
+      "loss": 1.2797,
+      "step": 30818
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.822897142857143e-06,
+      "loss": 1.2754,
+      "step": 30819
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.822628571428572e-06,
+      "loss": 1.4471,
+      "step": 30820
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.82236e-06,
+      "loss": 0.6779,
+      "step": 30821
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.822091428571429e-06,
+      "loss": 1.3542,
+      "step": 30822
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.821822857142857e-06,
+      "loss": 1.1278,
+      "step": 30823
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8215542857142855e-06,
+      "loss": 0.9303,
+      "step": 30824
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8212857142857145e-06,
+      "loss": 1.1391,
+      "step": 30825
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.821017142857143e-06,
+      "loss": 1.0259,
+      "step": 30826
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.820748571428572e-06,
+      "loss": 1.2621,
+      "step": 30827
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.82048e-06,
+      "loss": 1.4148,
+      "step": 30828
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.820211428571429e-06,
+      "loss": 1.2441,
+      "step": 30829
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.819942857142858e-06,
+      "loss": 0.8409,
+      "step": 30830
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.819674285714286e-06,
+      "loss": 1.0742,
+      "step": 30831
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.819405714285715e-06,
+      "loss": 1.6058,
+      "step": 30832
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.819137142857143e-06,
+      "loss": 1.7441,
+      "step": 30833
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8188685714285715e-06,
+      "loss": 1.3455,
+      "step": 30834
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8186000000000005e-06,
+      "loss": 1.2236,
+      "step": 30835
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.818331428571429e-06,
+      "loss": 1.3177,
+      "step": 30836
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.818062857142858e-06,
+      "loss": 1.1969,
+      "step": 30837
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.817794285714286e-06,
+      "loss": 1.072,
+      "step": 30838
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.817525714285714e-06,
+      "loss": 0.8459,
+      "step": 30839
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.817257142857143e-06,
+      "loss": 1.287,
+      "step": 30840
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.816988571428571e-06,
+      "loss": 1.3843,
+      "step": 30841
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.81672e-06,
+      "loss": 1.2468,
+      "step": 30842
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.816451428571429e-06,
+      "loss": 1.2154,
+      "step": 30843
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8161828571428575e-06,
+      "loss": 1.1692,
+      "step": 30844
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8159142857142865e-06,
+      "loss": 1.0729,
+      "step": 30845
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.815645714285715e-06,
+      "loss": 1.2134,
+      "step": 30846
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.815377142857143e-06,
+      "loss": 1.3171,
+      "step": 30847
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.815108571428572e-06,
+      "loss": 1.0832,
+      "step": 30848
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.81484e-06,
+      "loss": 1.3078,
+      "step": 30849
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.814571428571429e-06,
+      "loss": 0.937,
+      "step": 30850
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.814302857142857e-06,
+      "loss": 1.4389,
+      "step": 30851
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.814034285714285e-06,
+      "loss": 0.7326,
+      "step": 30852
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8137657142857144e-06,
+      "loss": 1.1853,
+      "step": 30853
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.813497142857143e-06,
+      "loss": 1.4132,
+      "step": 30854
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.813228571428572e-06,
+      "loss": 1.3127,
+      "step": 30855
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.812960000000001e-06,
+      "loss": 0.9529,
+      "step": 30856
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.812691428571429e-06,
+      "loss": 0.9201,
+      "step": 30857
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.812422857142858e-06,
+      "loss": 0.8619,
+      "step": 30858
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.812154285714286e-06,
+      "loss": 1.199,
+      "step": 30859
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.811885714285715e-06,
+      "loss": 1.2156,
+      "step": 30860
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.811617142857143e-06,
+      "loss": 0.9404,
+      "step": 30861
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.811348571428571e-06,
+      "loss": 1.2183,
+      "step": 30862
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8110800000000004e-06,
+      "loss": 0.9813,
+      "step": 30863
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.810811428571429e-06,
+      "loss": 1.1235,
+      "step": 30864
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.810542857142857e-06,
+      "loss": 1.5449,
+      "step": 30865
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.810274285714286e-06,
+      "loss": 1.1159,
+      "step": 30866
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.810005714285714e-06,
+      "loss": 0.9896,
+      "step": 30867
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.809737142857143e-06,
+      "loss": 0.9872,
+      "step": 30868
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.809468571428572e-06,
+      "loss": 1.1685,
+      "step": 30869
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8092e-06,
+      "loss": 1.0811,
+      "step": 30870
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.808931428571429e-06,
+      "loss": 0.7475,
+      "step": 30871
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.808662857142857e-06,
+      "loss": 1.1787,
+      "step": 30872
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.8083942857142864e-06,
+      "loss": 0.9599,
+      "step": 30873
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.808125714285715e-06,
+      "loss": 1.2687,
+      "step": 30874
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.807857142857143e-06,
+      "loss": 1.1921,
+      "step": 30875
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.807588571428572e-06,
+      "loss": 1.2724,
+      "step": 30876
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.80732e-06,
+      "loss": 1.2446,
+      "step": 30877
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.807051428571429e-06,
+      "loss": 1.4127,
+      "step": 30878
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.806782857142857e-06,
+      "loss": 1.4164,
+      "step": 30879
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.806514285714285e-06,
+      "loss": 1.5178,
+      "step": 30880
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.806245714285714e-06,
+      "loss": 0.853,
+      "step": 30881
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.805977142857143e-06,
+      "loss": 1.3005,
+      "step": 30882
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.805708571428572e-06,
+      "loss": 1.2005,
+      "step": 30883
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.805440000000001e-06,
+      "loss": 1.2933,
+      "step": 30884
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.805171428571429e-06,
+      "loss": 1.2789,
+      "step": 30885
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.804902857142858e-06,
+      "loss": 0.9842,
+      "step": 30886
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 5.804634285714286e-06,
+      "loss": 1.123,
+      "step": 30887
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.804365714285714e-06,
+      "loss": 1.1533,
+      "step": 30888
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.804097142857143e-06,
+      "loss": 1.1843,
+      "step": 30889
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.803828571428571e-06,
+      "loss": 1.0819,
+      "step": 30890
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.80356e-06,
+      "loss": 1.4513,
+      "step": 30891
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.8032914285714286e-06,
+      "loss": 0.9498,
+      "step": 30892
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.803022857142857e-06,
+      "loss": 1.0297,
+      "step": 30893
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.802754285714286e-06,
+      "loss": 1.4323,
+      "step": 30894
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.802485714285715e-06,
+      "loss": 1.5641,
+      "step": 30895
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.802217142857144e-06,
+      "loss": 1.1066,
+      "step": 30896
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.801948571428572e-06,
+      "loss": 0.9263,
+      "step": 30897
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.80168e-06,
+      "loss": 1.1977,
+      "step": 30898
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.801411428571429e-06,
+      "loss": 1.2582,
+      "step": 30899
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.801142857142857e-06,
+      "loss": 1.3845,
+      "step": 30900
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.800874285714286e-06,
+      "loss": 0.5636,
+      "step": 30901
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.8006057142857146e-06,
+      "loss": 1.2396,
+      "step": 30902
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.800337142857143e-06,
+      "loss": 1.0549,
+      "step": 30903
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.800068571428572e-06,
+      "loss": 0.6655,
+      "step": 30904
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7998e-06,
+      "loss": 1.5222,
+      "step": 30905
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.799531428571428e-06,
+      "loss": 0.8312,
+      "step": 30906
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.799262857142857e-06,
+      "loss": 1.0105,
+      "step": 30907
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.798994285714285e-06,
+      "loss": 1.3765,
+      "step": 30908
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.798725714285715e-06,
+      "loss": 1.0812,
+      "step": 30909
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.798457142857143e-06,
+      "loss": 0.8232,
+      "step": 30910
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7981885714285715e-06,
+      "loss": 1.5354,
+      "step": 30911
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7979200000000006e-06,
+      "loss": 1.3977,
+      "step": 30912
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.797651428571429e-06,
+      "loss": 1.55,
+      "step": 30913
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.797382857142858e-06,
+      "loss": 1.6826,
+      "step": 30914
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.797114285714286e-06,
+      "loss": 1.1999,
+      "step": 30915
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.796845714285714e-06,
+      "loss": 0.9737,
+      "step": 30916
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.796577142857143e-06,
+      "loss": 1.5011,
+      "step": 30917
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.796308571428571e-06,
+      "loss": 1.1604,
+      "step": 30918
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.79604e-06,
+      "loss": 1.0467,
+      "step": 30919
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7957714285714285e-06,
+      "loss": 1.0171,
+      "step": 30920
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.795502857142857e-06,
+      "loss": 1.1547,
+      "step": 30921
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7952342857142866e-06,
+      "loss": 1.5963,
+      "step": 30922
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.794965714285715e-06,
+      "loss": 1.2198,
+      "step": 30923
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.794697142857144e-06,
+      "loss": 1.1235,
+      "step": 30924
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.794428571428572e-06,
+      "loss": 1.2179,
+      "step": 30925
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.79416e-06,
+      "loss": 0.9015,
+      "step": 30926
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.793891428571429e-06,
+      "loss": 1.1554,
+      "step": 30927
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.793622857142857e-06,
+      "loss": 0.7823,
+      "step": 30928
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7933542857142855e-06,
+      "loss": 0.8963,
+      "step": 30929
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7930857142857145e-06,
+      "loss": 0.9584,
+      "step": 30930
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.792817142857143e-06,
+      "loss": 1.164,
+      "step": 30931
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.792548571428572e-06,
+      "loss": 1.2922,
+      "step": 30932
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.79228e-06,
+      "loss": 0.9404,
+      "step": 30933
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.792011428571428e-06,
+      "loss": 1.4302,
+      "step": 30934
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.791742857142858e-06,
+      "loss": 1.3605,
+      "step": 30935
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.791474285714286e-06,
+      "loss": 0.7193,
+      "step": 30936
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.791205714285715e-06,
+      "loss": 0.9269,
+      "step": 30937
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.790937142857143e-06,
+      "loss": 1.1148,
+      "step": 30938
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7906685714285715e-06,
+      "loss": 1.1515,
+      "step": 30939
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7904000000000005e-06,
+      "loss": 1.0732,
+      "step": 30940
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.790131428571429e-06,
+      "loss": 1.0958,
+      "step": 30941
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.789862857142858e-06,
+      "loss": 1.453,
+      "step": 30942
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.789594285714286e-06,
+      "loss": 1.4043,
+      "step": 30943
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.789325714285714e-06,
+      "loss": 1.3332,
+      "step": 30944
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.789057142857143e-06,
+      "loss": 0.8577,
+      "step": 30945
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.788788571428571e-06,
+      "loss": 1.0865,
+      "step": 30946
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7885199999999994e-06,
+      "loss": 1.3649,
+      "step": 30947
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.788251428571429e-06,
+      "loss": 1.0947,
+      "step": 30948
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7879828571428575e-06,
+      "loss": 1.5039,
+      "step": 30949
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7877142857142865e-06,
+      "loss": 1.1094,
+      "step": 30950
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.787445714285715e-06,
+      "loss": 1.3427,
+      "step": 30951
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.787177142857143e-06,
+      "loss": 1.3031,
+      "step": 30952
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.786908571428572e-06,
+      "loss": 1.1847,
+      "step": 30953
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.78664e-06,
+      "loss": 0.8667,
+      "step": 30954
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.786371428571429e-06,
+      "loss": 1.4192,
+      "step": 30955
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.786102857142857e-06,
+      "loss": 1.2348,
+      "step": 30956
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.785834285714285e-06,
+      "loss": 1.2892,
+      "step": 30957
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7855657142857144e-06,
+      "loss": 0.8932,
+      "step": 30958
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.785297142857143e-06,
+      "loss": 1.1363,
+      "step": 30959
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.785028571428572e-06,
+      "loss": 0.7888,
+      "step": 30960
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.78476e-06,
+      "loss": 1.3793,
+      "step": 30961
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.784491428571429e-06,
+      "loss": 1.4266,
+      "step": 30962
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.784222857142858e-06,
+      "loss": 0.9807,
+      "step": 30963
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.783954285714286e-06,
+      "loss": 1.0893,
+      "step": 30964
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.783685714285715e-06,
+      "loss": 1.412,
+      "step": 30965
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.783417142857143e-06,
+      "loss": 1.2615,
+      "step": 30966
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.783148571428571e-06,
+      "loss": 1.0635,
+      "step": 30967
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7828800000000004e-06,
+      "loss": 0.8599,
+      "step": 30968
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.782611428571429e-06,
+      "loss": 1.2782,
+      "step": 30969
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.782342857142857e-06,
+      "loss": 0.9366,
+      "step": 30970
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.782074285714286e-06,
+      "loss": 1.399,
+      "step": 30971
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.781805714285714e-06,
+      "loss": 1.2385,
+      "step": 30972
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.781537142857143e-06,
+      "loss": 0.8992,
+      "step": 30973
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.781268571428571e-06,
+      "loss": 0.9317,
+      "step": 30974
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.781e-06,
+      "loss": 1.504,
+      "step": 30975
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.780731428571429e-06,
+      "loss": 1.1028,
+      "step": 30976
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.780462857142857e-06,
+      "loss": 0.9808,
+      "step": 30977
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7801942857142864e-06,
+      "loss": 1.0999,
+      "step": 30978
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.779925714285715e-06,
+      "loss": 1.0263,
+      "step": 30979
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.779657142857143e-06,
+      "loss": 1.5733,
+      "step": 30980
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.779388571428572e-06,
+      "loss": 1.559,
+      "step": 30981
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.77912e-06,
+      "loss": 1.0091,
+      "step": 30982
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.778851428571429e-06,
+      "loss": 1.0012,
+      "step": 30983
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.778582857142857e-06,
+      "loss": 1.1449,
+      "step": 30984
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.778314285714285e-06,
+      "loss": 0.9909,
+      "step": 30985
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.778045714285714e-06,
+      "loss": 1.371,
+      "step": 30986
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7777771428571426e-06,
+      "loss": 1.1617,
+      "step": 30987
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7775085714285724e-06,
+      "loss": 1.2724,
+      "step": 30988
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.777240000000001e-06,
+      "loss": 1.0147,
+      "step": 30989
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.776971428571429e-06,
+      "loss": 1.2195,
+      "step": 30990
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.776702857142858e-06,
+      "loss": 1.3646,
+      "step": 30991
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.776434285714286e-06,
+      "loss": 1.0119,
+      "step": 30992
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.776165714285714e-06,
+      "loss": 1.093,
+      "step": 30993
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.775897142857143e-06,
+      "loss": 1.2775,
+      "step": 30994
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.775628571428571e-06,
+      "loss": 1.0564,
+      "step": 30995
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.77536e-06,
+      "loss": 1.0056,
+      "step": 30996
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7750914285714286e-06,
+      "loss": 1.2519,
+      "step": 30997
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.774822857142857e-06,
+      "loss": 1.0853,
+      "step": 30998
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.774554285714286e-06,
+      "loss": 1.2552,
+      "step": 30999
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.774285714285714e-06,
+      "loss": 1.2329,
+      "step": 31000
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.774017142857144e-06,
+      "loss": 1.1782,
+      "step": 31001
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.773748571428572e-06,
+      "loss": 1.2324,
+      "step": 31002
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.77348e-06,
+      "loss": 0.8095,
+      "step": 31003
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.773211428571429e-06,
+      "loss": 1.132,
+      "step": 31004
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.772942857142857e-06,
+      "loss": 1.1261,
+      "step": 31005
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.772674285714286e-06,
+      "loss": 0.8918,
+      "step": 31006
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7724057142857146e-06,
+      "loss": 1.5634,
+      "step": 31007
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.772137142857143e-06,
+      "loss": 1.4821,
+      "step": 31008
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.771868571428572e-06,
+      "loss": 1.0084,
+      "step": 31009
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7716e-06,
+      "loss": 0.7801,
+      "step": 31010
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.771331428571429e-06,
+      "loss": 1.2716,
+      "step": 31011
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.771062857142857e-06,
+      "loss": 0.8569,
+      "step": 31012
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.770794285714285e-06,
+      "loss": 1.7316,
+      "step": 31013
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.770525714285714e-06,
+      "loss": 1.0365,
+      "step": 31014
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.770257142857143e-06,
+      "loss": 1.0445,
+      "step": 31015
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7699885714285715e-06,
+      "loss": 1.4446,
+      "step": 31016
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7697200000000006e-06,
+      "loss": 1.049,
+      "step": 31017
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.769451428571429e-06,
+      "loss": 1.3749,
+      "step": 31018
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.769182857142858e-06,
+      "loss": 0.9272,
+      "step": 31019
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.768914285714286e-06,
+      "loss": 0.9444,
+      "step": 31020
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.768645714285714e-06,
+      "loss": 1.0101,
+      "step": 31021
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.768377142857143e-06,
+      "loss": 0.9576,
+      "step": 31022
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.768108571428571e-06,
+      "loss": 0.8523,
+      "step": 31023
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.76784e-06,
+      "loss": 0.9988,
+      "step": 31024
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7675714285714285e-06,
+      "loss": 1.1759,
+      "step": 31025
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.767302857142857e-06,
+      "loss": 1.4158,
+      "step": 31026
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.767034285714286e-06,
+      "loss": 0.8354,
+      "step": 31027
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.766765714285715e-06,
+      "loss": 1.0596,
+      "step": 31028
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.766497142857144e-06,
+      "loss": 1.2194,
+      "step": 31029
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.766228571428572e-06,
+      "loss": 0.9117,
+      "step": 31030
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.76596e-06,
+      "loss": 0.71,
+      "step": 31031
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.765691428571429e-06,
+      "loss": 0.8023,
+      "step": 31032
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.765422857142857e-06,
+      "loss": 1.4852,
+      "step": 31033
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7651542857142855e-06,
+      "loss": 1.162,
+      "step": 31034
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7648857142857145e-06,
+      "loss": 1.4361,
+      "step": 31035
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.764617142857143e-06,
+      "loss": 1.4338,
+      "step": 31036
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.764348571428572e-06,
+      "loss": 0.9316,
+      "step": 31037
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.76408e-06,
+      "loss": 0.9684,
+      "step": 31038
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.763811428571428e-06,
+      "loss": 1.4789,
+      "step": 31039
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.763542857142857e-06,
+      "loss": 0.9898,
+      "step": 31040
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.763274285714286e-06,
+      "loss": 0.8991,
+      "step": 31041
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.763005714285715e-06,
+      "loss": 1.0156,
+      "step": 31042
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.762737142857143e-06,
+      "loss": 1.2237,
+      "step": 31043
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7624685714285715e-06,
+      "loss": 1.6222,
+      "step": 31044
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7622000000000005e-06,
+      "loss": 1.081,
+      "step": 31045
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.761931428571429e-06,
+      "loss": 1.3051,
+      "step": 31046
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.761662857142858e-06,
+      "loss": 0.8534,
+      "step": 31047
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.761394285714286e-06,
+      "loss": 0.9694,
+      "step": 31048
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.761125714285714e-06,
+      "loss": 1.2657,
+      "step": 31049
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.760857142857143e-06,
+      "loss": 0.7286,
+      "step": 31050
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.760588571428571e-06,
+      "loss": 0.9455,
+      "step": 31051
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.76032e-06,
+      "loss": 0.9137,
+      "step": 31052
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7600514285714284e-06,
+      "loss": 1.4088,
+      "step": 31053
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7597828571428575e-06,
+      "loss": 0.8925,
+      "step": 31054
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.7595142857142865e-06,
+      "loss": 0.6853,
+      "step": 31055
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.759245714285715e-06,
+      "loss": 0.9586,
+      "step": 31056
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.758977142857143e-06,
+      "loss": 1.4781,
+      "step": 31057
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.758708571428572e-06,
+      "loss": 0.8447,
+      "step": 31058
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.75844e-06,
+      "loss": 0.7527,
+      "step": 31059
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.758171428571429e-06,
+      "loss": 0.8465,
+      "step": 31060
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.757902857142857e-06,
+      "loss": 0.9493,
+      "step": 31061
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 5.757634285714285e-06,
+      "loss": 1.2316,
+      "step": 31062
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7573657142857144e-06,
+      "loss": 1.3509,
+      "step": 31063
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.757097142857143e-06,
+      "loss": 0.9313,
+      "step": 31064
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.756828571428572e-06,
+      "loss": 1.2305,
+      "step": 31065
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.75656e-06,
+      "loss": 1.0965,
+      "step": 31066
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.756291428571429e-06,
+      "loss": 1.1451,
+      "step": 31067
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.756022857142858e-06,
+      "loss": 1.1025,
+      "step": 31068
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.755754285714286e-06,
+      "loss": 1.4298,
+      "step": 31069
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.755485714285715e-06,
+      "loss": 1.4792,
+      "step": 31070
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.755217142857143e-06,
+      "loss": 0.8432,
+      "step": 31071
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.754948571428571e-06,
+      "loss": 1.1023,
+      "step": 31072
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7546800000000004e-06,
+      "loss": 1.4615,
+      "step": 31073
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.754411428571429e-06,
+      "loss": 1.1406,
+      "step": 31074
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.754142857142858e-06,
+      "loss": 0.9168,
+      "step": 31075
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.753874285714286e-06,
+      "loss": 1.0565,
+      "step": 31076
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.753605714285714e-06,
+      "loss": 1.6424,
+      "step": 31077
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.753337142857143e-06,
+      "loss": 1.1789,
+      "step": 31078
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.753068571428571e-06,
+      "loss": 1.124,
+      "step": 31079
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7528e-06,
+      "loss": 1.4243,
+      "step": 31080
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.752531428571429e-06,
+      "loss": 1.4531,
+      "step": 31081
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.752262857142857e-06,
+      "loss": 1.1245,
+      "step": 31082
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7519942857142864e-06,
+      "loss": 1.0689,
+      "step": 31083
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.751725714285715e-06,
+      "loss": 1.3028,
+      "step": 31084
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.751457142857143e-06,
+      "loss": 1.346,
+      "step": 31085
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.751188571428572e-06,
+      "loss": 1.2228,
+      "step": 31086
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.75092e-06,
+      "loss": 1.3259,
+      "step": 31087
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.750651428571429e-06,
+      "loss": 1.2491,
+      "step": 31088
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.750382857142857e-06,
+      "loss": 0.8644,
+      "step": 31089
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.750114285714285e-06,
+      "loss": 1.1154,
+      "step": 31090
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.749845714285714e-06,
+      "loss": 0.8871,
+      "step": 31091
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7495771428571426e-06,
+      "loss": 0.8616,
+      "step": 31092
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.749308571428572e-06,
+      "loss": 1.2293,
+      "step": 31093
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.749040000000001e-06,
+      "loss": 1.1094,
+      "step": 31094
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.748771428571429e-06,
+      "loss": 1.0854,
+      "step": 31095
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.748502857142858e-06,
+      "loss": 1.7455,
+      "step": 31096
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.748234285714286e-06,
+      "loss": 1.2116,
+      "step": 31097
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.747965714285715e-06,
+      "loss": 0.8982,
+      "step": 31098
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.747697142857143e-06,
+      "loss": 1.075,
+      "step": 31099
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.747428571428571e-06,
+      "loss": 0.9012,
+      "step": 31100
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.74716e-06,
+      "loss": 1.1898,
+      "step": 31101
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7468914285714286e-06,
+      "loss": 1.328,
+      "step": 31102
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.746622857142857e-06,
+      "loss": 1.1714,
+      "step": 31103
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.746354285714286e-06,
+      "loss": 0.9163,
+      "step": 31104
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.746085714285714e-06,
+      "loss": 1.1736,
+      "step": 31105
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.745817142857143e-06,
+      "loss": 1.3275,
+      "step": 31106
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.745548571428572e-06,
+      "loss": 0.901,
+      "step": 31107
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.74528e-06,
+      "loss": 1.4061,
+      "step": 31108
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.745011428571429e-06,
+      "loss": 1.4765,
+      "step": 31109
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.744742857142857e-06,
+      "loss": 0.8801,
+      "step": 31110
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.744474285714286e-06,
+      "loss": 1.1315,
+      "step": 31111
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7442057142857146e-06,
+      "loss": 0.8686,
+      "step": 31112
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.743937142857143e-06,
+      "loss": 1.0949,
+      "step": 31113
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.743668571428572e-06,
+      "loss": 0.9962,
+      "step": 31114
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7434e-06,
+      "loss": 0.8725,
+      "step": 31115
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.743131428571429e-06,
+      "loss": 1.2547,
+      "step": 31116
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.742862857142857e-06,
+      "loss": 1.0469,
+      "step": 31117
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.742594285714285e-06,
+      "loss": 1.524,
+      "step": 31118
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.742325714285714e-06,
+      "loss": 1.008,
+      "step": 31119
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.742057142857143e-06,
+      "loss": 1.1302,
+      "step": 31120
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7417885714285715e-06,
+      "loss": 0.9093,
+      "step": 31121
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7415200000000006e-06,
+      "loss": 1.1399,
+      "step": 31122
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.741251428571429e-06,
+      "loss": 1.0504,
+      "step": 31123
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.740982857142858e-06,
+      "loss": 0.8984,
+      "step": 31124
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.740714285714286e-06,
+      "loss": 1.1698,
+      "step": 31125
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.740445714285714e-06,
+      "loss": 0.7683,
+      "step": 31126
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.740177142857143e-06,
+      "loss": 1.3579,
+      "step": 31127
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.739908571428571e-06,
+      "loss": 1.3162,
+      "step": 31128
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.73964e-06,
+      "loss": 0.9943,
+      "step": 31129
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7393714285714285e-06,
+      "loss": 1.3786,
+      "step": 31130
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.739102857142857e-06,
+      "loss": 1.2795,
+      "step": 31131
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.738834285714286e-06,
+      "loss": 1.083,
+      "step": 31132
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.738565714285715e-06,
+      "loss": 1.2313,
+      "step": 31133
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.738297142857144e-06,
+      "loss": 0.9216,
+      "step": 31134
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.738028571428572e-06,
+      "loss": 1.1873,
+      "step": 31135
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.73776e-06,
+      "loss": 1.0864,
+      "step": 31136
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.737491428571429e-06,
+      "loss": 0.9655,
+      "step": 31137
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.737222857142857e-06,
+      "loss": 1.0803,
+      "step": 31138
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.736954285714286e-06,
+      "loss": 1.3241,
+      "step": 31139
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7366857142857145e-06,
+      "loss": 1.1476,
+      "step": 31140
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.736417142857143e-06,
+      "loss": 1.2383,
+      "step": 31141
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.736148571428572e-06,
+      "loss": 1.5021,
+      "step": 31142
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.73588e-06,
+      "loss": 1.7057,
+      "step": 31143
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.735611428571428e-06,
+      "loss": 1.0245,
+      "step": 31144
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.735342857142857e-06,
+      "loss": 0.7482,
+      "step": 31145
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.735074285714286e-06,
+      "loss": 1.1816,
+      "step": 31146
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.734805714285715e-06,
+      "loss": 1.0096,
+      "step": 31147
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.734537142857143e-06,
+      "loss": 1.2588,
+      "step": 31148
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7342685714285715e-06,
+      "loss": 1.2843,
+      "step": 31149
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7340000000000005e-06,
+      "loss": 1.4887,
+      "step": 31150
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.733731428571429e-06,
+      "loss": 1.0511,
+      "step": 31151
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.733462857142858e-06,
+      "loss": 1.3038,
+      "step": 31152
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.733194285714286e-06,
+      "loss": 1.0509,
+      "step": 31153
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.732925714285714e-06,
+      "loss": 1.4429,
+      "step": 31154
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.732657142857143e-06,
+      "loss": 1.5286,
+      "step": 31155
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.732388571428571e-06,
+      "loss": 1.2255,
+      "step": 31156
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.73212e-06,
+      "loss": 1.1939,
+      "step": 31157
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7318514285714284e-06,
+      "loss": 0.9924,
+      "step": 31158
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7315828571428575e-06,
+      "loss": 1.433,
+      "step": 31159
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7313142857142865e-06,
+      "loss": 1.0121,
+      "step": 31160
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.731045714285715e-06,
+      "loss": 1.6123,
+      "step": 31161
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.730777142857144e-06,
+      "loss": 0.7495,
+      "step": 31162
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.730508571428572e-06,
+      "loss": 1.2327,
+      "step": 31163
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.73024e-06,
+      "loss": 1.4455,
+      "step": 31164
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.729971428571429e-06,
+      "loss": 0.9223,
+      "step": 31165
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.729702857142857e-06,
+      "loss": 1.406,
+      "step": 31166
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.729434285714285e-06,
+      "loss": 0.9605,
+      "step": 31167
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7291657142857144e-06,
+      "loss": 1.3545,
+      "step": 31168
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.728897142857143e-06,
+      "loss": 0.7627,
+      "step": 31169
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.728628571428572e-06,
+      "loss": 1.4458,
+      "step": 31170
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.72836e-06,
+      "loss": 1.3326,
+      "step": 31171
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.728091428571429e-06,
+      "loss": 1.1864,
+      "step": 31172
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.727822857142858e-06,
+      "loss": 1.438,
+      "step": 31173
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.727554285714286e-06,
+      "loss": 1.3543,
+      "step": 31174
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.727285714285715e-06,
+      "loss": 1.3983,
+      "step": 31175
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.727017142857143e-06,
+      "loss": 1.2418,
+      "step": 31176
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.726748571428571e-06,
+      "loss": 0.7394,
+      "step": 31177
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7264800000000004e-06,
+      "loss": 0.9848,
+      "step": 31178
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.726211428571429e-06,
+      "loss": 1.432,
+      "step": 31179
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.725942857142858e-06,
+      "loss": 0.7735,
+      "step": 31180
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.725674285714286e-06,
+      "loss": 1.0782,
+      "step": 31181
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.725405714285714e-06,
+      "loss": 1.1813,
+      "step": 31182
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.725137142857143e-06,
+      "loss": 1.3737,
+      "step": 31183
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.724868571428571e-06,
+      "loss": 1.0864,
+      "step": 31184
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7246e-06,
+      "loss": 1.2182,
+      "step": 31185
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.724331428571429e-06,
+      "loss": 0.8668,
+      "step": 31186
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.724062857142857e-06,
+      "loss": 0.7598,
+      "step": 31187
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7237942857142864e-06,
+      "loss": 1.0815,
+      "step": 31188
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.723525714285715e-06,
+      "loss": 1.1486,
+      "step": 31189
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.723257142857143e-06,
+      "loss": 0.9123,
+      "step": 31190
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.722988571428572e-06,
+      "loss": 0.9237,
+      "step": 31191
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.72272e-06,
+      "loss": 0.7779,
+      "step": 31192
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.722451428571429e-06,
+      "loss": 1.1239,
+      "step": 31193
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.722182857142857e-06,
+      "loss": 1.7166,
+      "step": 31194
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.721914285714285e-06,
+      "loss": 1.3978,
+      "step": 31195
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.721645714285714e-06,
+      "loss": 1.2259,
+      "step": 31196
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7213771428571426e-06,
+      "loss": 0.8224,
+      "step": 31197
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.721108571428572e-06,
+      "loss": 0.7942,
+      "step": 31198
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.720840000000001e-06,
+      "loss": 0.8636,
+      "step": 31199
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.720571428571429e-06,
+      "loss": 0.9449,
+      "step": 31200
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.720302857142858e-06,
+      "loss": 0.8211,
+      "step": 31201
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.720034285714286e-06,
+      "loss": 1.2486,
+      "step": 31202
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.719765714285715e-06,
+      "loss": 1.4567,
+      "step": 31203
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.719497142857143e-06,
+      "loss": 1.2148,
+      "step": 31204
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.719228571428571e-06,
+      "loss": 1.1779,
+      "step": 31205
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.71896e-06,
+      "loss": 1.198,
+      "step": 31206
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7186914285714286e-06,
+      "loss": 0.9631,
+      "step": 31207
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.718422857142857e-06,
+      "loss": 1.3771,
+      "step": 31208
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.718154285714286e-06,
+      "loss": 1.2496,
+      "step": 31209
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.717885714285714e-06,
+      "loss": 1.2388,
+      "step": 31210
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.717617142857143e-06,
+      "loss": 1.0659,
+      "step": 31211
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.717348571428572e-06,
+      "loss": 0.8067,
+      "step": 31212
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.71708e-06,
+      "loss": 1.0615,
+      "step": 31213
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.716811428571429e-06,
+      "loss": 0.9758,
+      "step": 31214
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.716542857142857e-06,
+      "loss": 0.7828,
+      "step": 31215
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.716274285714286e-06,
+      "loss": 1.0057,
+      "step": 31216
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7160057142857146e-06,
+      "loss": 1.021,
+      "step": 31217
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.715737142857143e-06,
+      "loss": 0.9256,
+      "step": 31218
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.715468571428572e-06,
+      "loss": 0.9009,
+      "step": 31219
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7152e-06,
+      "loss": 1.1208,
+      "step": 31220
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.714931428571429e-06,
+      "loss": 1.0208,
+      "step": 31221
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.714662857142857e-06,
+      "loss": 0.8423,
+      "step": 31222
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.714394285714285e-06,
+      "loss": 1.2748,
+      "step": 31223
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.714125714285714e-06,
+      "loss": 1.2645,
+      "step": 31224
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.713857142857143e-06,
+      "loss": 1.14,
+      "step": 31225
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.713588571428572e-06,
+      "loss": 0.7298,
+      "step": 31226
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7133200000000006e-06,
+      "loss": 1.3537,
+      "step": 31227
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.713051428571429e-06,
+      "loss": 1.2076,
+      "step": 31228
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.712782857142858e-06,
+      "loss": 1.129,
+      "step": 31229
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.712514285714286e-06,
+      "loss": 1.0324,
+      "step": 31230
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.712245714285714e-06,
+      "loss": 1.0408,
+      "step": 31231
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.711977142857143e-06,
+      "loss": 1.0246,
+      "step": 31232
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.711708571428571e-06,
+      "loss": 1.3418,
+      "step": 31233
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.71144e-06,
+      "loss": 0.8424,
+      "step": 31234
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.7111714285714285e-06,
+      "loss": 1.4964,
+      "step": 31235
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.710902857142857e-06,
+      "loss": 1.2814,
+      "step": 31236
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 5.710634285714286e-06,
+      "loss": 1.0271,
+      "step": 31237
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.710365714285715e-06,
+      "loss": 1.4736,
+      "step": 31238
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.710097142857144e-06,
+      "loss": 1.1601,
+      "step": 31239
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.709828571428572e-06,
+      "loss": 1.041,
+      "step": 31240
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.70956e-06,
+      "loss": 1.1531,
+      "step": 31241
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.709291428571429e-06,
+      "loss": 1.428,
+      "step": 31242
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.709022857142857e-06,
+      "loss": 1.4301,
+      "step": 31243
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.708754285714286e-06,
+      "loss": 0.8301,
+      "step": 31244
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7084857142857145e-06,
+      "loss": 1.0003,
+      "step": 31245
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.708217142857143e-06,
+      "loss": 1.0148,
+      "step": 31246
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.707948571428572e-06,
+      "loss": 1.2436,
+      "step": 31247
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.70768e-06,
+      "loss": 0.9903,
+      "step": 31248
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.707411428571429e-06,
+      "loss": 1.1129,
+      "step": 31249
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.707142857142857e-06,
+      "loss": 1.3058,
+      "step": 31250
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.706874285714286e-06,
+      "loss": 0.8219,
+      "step": 31251
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.706605714285715e-06,
+      "loss": 1.0047,
+      "step": 31252
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.706337142857143e-06,
+      "loss": 1.2715,
+      "step": 31253
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7060685714285715e-06,
+      "loss": 0.9411,
+      "step": 31254
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7058000000000005e-06,
+      "loss": 1.1023,
+      "step": 31255
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.705531428571429e-06,
+      "loss": 1.6751,
+      "step": 31256
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.705262857142858e-06,
+      "loss": 0.9036,
+      "step": 31257
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.704994285714286e-06,
+      "loss": 1.1918,
+      "step": 31258
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.704725714285714e-06,
+      "loss": 1.06,
+      "step": 31259
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.704457142857143e-06,
+      "loss": 1.5875,
+      "step": 31260
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.704188571428571e-06,
+      "loss": 1.027,
+      "step": 31261
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.70392e-06,
+      "loss": 1.2267,
+      "step": 31262
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7036514285714284e-06,
+      "loss": 0.9938,
+      "step": 31263
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7033828571428575e-06,
+      "loss": 1.6173,
+      "step": 31264
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7031142857142865e-06,
+      "loss": 1.3441,
+      "step": 31265
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.702845714285715e-06,
+      "loss": 1.1056,
+      "step": 31266
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.702577142857144e-06,
+      "loss": 1.3617,
+      "step": 31267
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.702308571428572e-06,
+      "loss": 1.4034,
+      "step": 31268
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.70204e-06,
+      "loss": 1.1054,
+      "step": 31269
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.701771428571429e-06,
+      "loss": 1.2241,
+      "step": 31270
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.701502857142857e-06,
+      "loss": 1.5119,
+      "step": 31271
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.701234285714286e-06,
+      "loss": 0.8293,
+      "step": 31272
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.7009657142857144e-06,
+      "loss": 1.2585,
+      "step": 31273
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.700697142857143e-06,
+      "loss": 1.4435,
+      "step": 31274
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.700428571428572e-06,
+      "loss": 1.0123,
+      "step": 31275
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.70016e-06,
+      "loss": 0.9505,
+      "step": 31276
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.699891428571429e-06,
+      "loss": 1.1885,
+      "step": 31277
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.699622857142858e-06,
+      "loss": 1.4879,
+      "step": 31278
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.699354285714286e-06,
+      "loss": 0.8475,
+      "step": 31279
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.699085714285715e-06,
+      "loss": 1.1049,
+      "step": 31280
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.698817142857143e-06,
+      "loss": 1.254,
+      "step": 31281
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.698548571428571e-06,
+      "loss": 0.9757,
+      "step": 31282
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6982800000000004e-06,
+      "loss": 1.524,
+      "step": 31283
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.698011428571429e-06,
+      "loss": 1.0037,
+      "step": 31284
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.697742857142858e-06,
+      "loss": 1.1253,
+      "step": 31285
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.697474285714286e-06,
+      "loss": 1.0404,
+      "step": 31286
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.697205714285714e-06,
+      "loss": 1.0649,
+      "step": 31287
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.696937142857143e-06,
+      "loss": 1.4967,
+      "step": 31288
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.696668571428571e-06,
+      "loss": 1.2889,
+      "step": 31289
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6964e-06,
+      "loss": 1.2416,
+      "step": 31290
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.696131428571429e-06,
+      "loss": 0.9849,
+      "step": 31291
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.695862857142857e-06,
+      "loss": 1.44,
+      "step": 31292
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6955942857142864e-06,
+      "loss": 1.4708,
+      "step": 31293
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.695325714285715e-06,
+      "loss": 1.158,
+      "step": 31294
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.695057142857143e-06,
+      "loss": 1.142,
+      "step": 31295
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.694788571428572e-06,
+      "loss": 0.9115,
+      "step": 31296
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.69452e-06,
+      "loss": 0.8303,
+      "step": 31297
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.694251428571429e-06,
+      "loss": 1.2764,
+      "step": 31298
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.693982857142857e-06,
+      "loss": 1.3049,
+      "step": 31299
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.693714285714285e-06,
+      "loss": 1.0794,
+      "step": 31300
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.693445714285714e-06,
+      "loss": 1.0073,
+      "step": 31301
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6931771428571426e-06,
+      "loss": 0.7925,
+      "step": 31302
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.692908571428572e-06,
+      "loss": 1.3843,
+      "step": 31303
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.692640000000001e-06,
+      "loss": 1.0655,
+      "step": 31304
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.692371428571429e-06,
+      "loss": 0.6274,
+      "step": 31305
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.692102857142858e-06,
+      "loss": 1.1817,
+      "step": 31306
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.691834285714286e-06,
+      "loss": 1.433,
+      "step": 31307
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.691565714285715e-06,
+      "loss": 0.9933,
+      "step": 31308
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.691297142857143e-06,
+      "loss": 1.5347,
+      "step": 31309
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.691028571428571e-06,
+      "loss": 1.3081,
+      "step": 31310
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.69076e-06,
+      "loss": 0.8612,
+      "step": 31311
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6904914285714286e-06,
+      "loss": 1.0308,
+      "step": 31312
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.690222857142858e-06,
+      "loss": 0.8642,
+      "step": 31313
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.689954285714286e-06,
+      "loss": 1.1897,
+      "step": 31314
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.689685714285714e-06,
+      "loss": 0.9211,
+      "step": 31315
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.689417142857143e-06,
+      "loss": 1.4212,
+      "step": 31316
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.689148571428572e-06,
+      "loss": 1.4323,
+      "step": 31317
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.68888e-06,
+      "loss": 1.0375,
+      "step": 31318
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.688611428571429e-06,
+      "loss": 1.4448,
+      "step": 31319
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.688342857142857e-06,
+      "loss": 1.2385,
+      "step": 31320
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.688074285714286e-06,
+      "loss": 1.4569,
+      "step": 31321
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6878057142857146e-06,
+      "loss": 1.4676,
+      "step": 31322
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.687537142857143e-06,
+      "loss": 0.7665,
+      "step": 31323
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.687268571428572e-06,
+      "loss": 1.1604,
+      "step": 31324
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.687e-06,
+      "loss": 0.9822,
+      "step": 31325
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.686731428571429e-06,
+      "loss": 1.0398,
+      "step": 31326
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.686462857142857e-06,
+      "loss": 1.0539,
+      "step": 31327
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.686194285714285e-06,
+      "loss": 1.0913,
+      "step": 31328
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.685925714285714e-06,
+      "loss": 0.9258,
+      "step": 31329
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.685657142857143e-06,
+      "loss": 1.1014,
+      "step": 31330
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.685388571428572e-06,
+      "loss": 1.1275,
+      "step": 31331
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6851200000000006e-06,
+      "loss": 1.3108,
+      "step": 31332
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.684851428571429e-06,
+      "loss": 1.0289,
+      "step": 31333
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.684582857142858e-06,
+      "loss": 1.7909,
+      "step": 31334
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.684314285714286e-06,
+      "loss": 1.0824,
+      "step": 31335
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.684045714285715e-06,
+      "loss": 0.9615,
+      "step": 31336
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.683777142857143e-06,
+      "loss": 0.9261,
+      "step": 31337
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.683508571428571e-06,
+      "loss": 0.9492,
+      "step": 31338
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.68324e-06,
+      "loss": 1.0718,
+      "step": 31339
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6829714285714285e-06,
+      "loss": 1.024,
+      "step": 31340
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.682702857142857e-06,
+      "loss": 1.1777,
+      "step": 31341
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.682434285714286e-06,
+      "loss": 1.0235,
+      "step": 31342
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.682165714285715e-06,
+      "loss": 1.1355,
+      "step": 31343
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.681897142857144e-06,
+      "loss": 1.3029,
+      "step": 31344
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.681628571428572e-06,
+      "loss": 1.4008,
+      "step": 31345
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.68136e-06,
+      "loss": 0.9608,
+      "step": 31346
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.681091428571429e-06,
+      "loss": 1.0619,
+      "step": 31347
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.680822857142857e-06,
+      "loss": 1.2045,
+      "step": 31348
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.680554285714286e-06,
+      "loss": 1.3253,
+      "step": 31349
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6802857142857145e-06,
+      "loss": 1.2367,
+      "step": 31350
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.680017142857143e-06,
+      "loss": 1.2656,
+      "step": 31351
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.679748571428572e-06,
+      "loss": 0.7948,
+      "step": 31352
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.67948e-06,
+      "loss": 1.1471,
+      "step": 31353
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.679211428571429e-06,
+      "loss": 1.3696,
+      "step": 31354
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.678942857142857e-06,
+      "loss": 0.8606,
+      "step": 31355
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.678674285714285e-06,
+      "loss": 1.1463,
+      "step": 31356
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.678405714285715e-06,
+      "loss": 1.2486,
+      "step": 31357
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.678137142857143e-06,
+      "loss": 1.3593,
+      "step": 31358
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6778685714285715e-06,
+      "loss": 1.0299,
+      "step": 31359
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6776000000000005e-06,
+      "loss": 1.0745,
+      "step": 31360
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.677331428571429e-06,
+      "loss": 1.1773,
+      "step": 31361
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.677062857142858e-06,
+      "loss": 1.1868,
+      "step": 31362
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.676794285714286e-06,
+      "loss": 1.1674,
+      "step": 31363
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.676525714285714e-06,
+      "loss": 1.0695,
+      "step": 31364
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.676257142857143e-06,
+      "loss": 0.9576,
+      "step": 31365
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.675988571428571e-06,
+      "loss": 0.6576,
+      "step": 31366
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.67572e-06,
+      "loss": 1.4095,
+      "step": 31367
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6754514285714284e-06,
+      "loss": 1.2853,
+      "step": 31368
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.675182857142857e-06,
+      "loss": 1.4911,
+      "step": 31369
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6749142857142865e-06,
+      "loss": 1.3351,
+      "step": 31370
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.674645714285715e-06,
+      "loss": 1.1503,
+      "step": 31371
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.674377142857144e-06,
+      "loss": 1.4272,
+      "step": 31372
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.674108571428572e-06,
+      "loss": 1.1323,
+      "step": 31373
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.67384e-06,
+      "loss": 1.3552,
+      "step": 31374
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.673571428571429e-06,
+      "loss": 1.2334,
+      "step": 31375
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.673302857142857e-06,
+      "loss": 1.1144,
+      "step": 31376
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.673034285714286e-06,
+      "loss": 1.0783,
+      "step": 31377
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6727657142857144e-06,
+      "loss": 0.9016,
+      "step": 31378
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.672497142857143e-06,
+      "loss": 1.1527,
+      "step": 31379
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.672228571428572e-06,
+      "loss": 1.1956,
+      "step": 31380
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.67196e-06,
+      "loss": 1.1271,
+      "step": 31381
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.671691428571428e-06,
+      "loss": 1.3018,
+      "step": 31382
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.671422857142858e-06,
+      "loss": 1.2745,
+      "step": 31383
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.671154285714286e-06,
+      "loss": 1.4263,
+      "step": 31384
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.670885714285715e-06,
+      "loss": 1.0986,
+      "step": 31385
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.670617142857143e-06,
+      "loss": 1.3905,
+      "step": 31386
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.670348571428571e-06,
+      "loss": 1.0966,
+      "step": 31387
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6700800000000004e-06,
+      "loss": 1.2904,
+      "step": 31388
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.669811428571429e-06,
+      "loss": 0.973,
+      "step": 31389
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.669542857142858e-06,
+      "loss": 0.8799,
+      "step": 31390
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.669274285714286e-06,
+      "loss": 1.1099,
+      "step": 31391
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.669005714285714e-06,
+      "loss": 0.6141,
+      "step": 31392
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.668737142857143e-06,
+      "loss": 1.312,
+      "step": 31393
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.668468571428571e-06,
+      "loss": 1.3168,
+      "step": 31394
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6682e-06,
+      "loss": 1.0697,
+      "step": 31395
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.667931428571429e-06,
+      "loss": 0.9626,
+      "step": 31396
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.667662857142857e-06,
+      "loss": 1.1235,
+      "step": 31397
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6673942857142864e-06,
+      "loss": 1.5118,
+      "step": 31398
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.667125714285715e-06,
+      "loss": 1.0563,
+      "step": 31399
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.666857142857144e-06,
+      "loss": 1.331,
+      "step": 31400
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.666588571428572e-06,
+      "loss": 1.2396,
+      "step": 31401
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.66632e-06,
+      "loss": 0.9037,
+      "step": 31402
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.666051428571429e-06,
+      "loss": 1.7412,
+      "step": 31403
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.665782857142857e-06,
+      "loss": 0.9613,
+      "step": 31404
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.665514285714285e-06,
+      "loss": 1.5371,
+      "step": 31405
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.665245714285714e-06,
+      "loss": 1.5001,
+      "step": 31406
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.6649771428571426e-06,
+      "loss": 0.7501,
+      "step": 31407
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.664708571428572e-06,
+      "loss": 1.2168,
+      "step": 31408
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.66444e-06,
+      "loss": 1.3424,
+      "step": 31409
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.664171428571429e-06,
+      "loss": 1.329,
+      "step": 31410
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.663902857142858e-06,
+      "loss": 1.4901,
+      "step": 31411
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 5.663634285714286e-06,
+      "loss": 1.6337,
+      "step": 31412
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.663365714285715e-06,
+      "loss": 0.985,
+      "step": 31413
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.663097142857143e-06,
+      "loss": 1.0668,
+      "step": 31414
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.662828571428571e-06,
+      "loss": 1.2189,
+      "step": 31415
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.66256e-06,
+      "loss": 1.462,
+      "step": 31416
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6622914285714286e-06,
+      "loss": 1.6234,
+      "step": 31417
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.662022857142858e-06,
+      "loss": 0.9594,
+      "step": 31418
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.661754285714286e-06,
+      "loss": 1.6207,
+      "step": 31419
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.661485714285714e-06,
+      "loss": 0.9732,
+      "step": 31420
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.661217142857143e-06,
+      "loss": 1.2776,
+      "step": 31421
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.660948571428571e-06,
+      "loss": 0.9575,
+      "step": 31422
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.660680000000001e-06,
+      "loss": 0.856,
+      "step": 31423
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.660411428571429e-06,
+      "loss": 0.7348,
+      "step": 31424
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.660142857142857e-06,
+      "loss": 1.5092,
+      "step": 31425
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.659874285714286e-06,
+      "loss": 1.3313,
+      "step": 31426
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6596057142857146e-06,
+      "loss": 0.92,
+      "step": 31427
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.659337142857143e-06,
+      "loss": 0.7903,
+      "step": 31428
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.659068571428572e-06,
+      "loss": 1.0648,
+      "step": 31429
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6588e-06,
+      "loss": 0.8302,
+      "step": 31430
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.658531428571429e-06,
+      "loss": 1.0499,
+      "step": 31431
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.658262857142857e-06,
+      "loss": 1.0345,
+      "step": 31432
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.657994285714285e-06,
+      "loss": 1.1891,
+      "step": 31433
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.657725714285714e-06,
+      "loss": 1.0321,
+      "step": 31434
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6574571428571425e-06,
+      "loss": 1.3015,
+      "step": 31435
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.657188571428572e-06,
+      "loss": 1.269,
+      "step": 31436
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6569200000000006e-06,
+      "loss": 0.9504,
+      "step": 31437
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.656651428571429e-06,
+      "loss": 1.5563,
+      "step": 31438
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.656382857142858e-06,
+      "loss": 1.1794,
+      "step": 31439
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.656114285714286e-06,
+      "loss": 0.8975,
+      "step": 31440
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.655845714285715e-06,
+      "loss": 1.5734,
+      "step": 31441
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.655577142857143e-06,
+      "loss": 1.0008,
+      "step": 31442
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.655308571428571e-06,
+      "loss": 0.8146,
+      "step": 31443
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.65504e-06,
+      "loss": 1.2692,
+      "step": 31444
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6547714285714285e-06,
+      "loss": 1.0508,
+      "step": 31445
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.654502857142857e-06,
+      "loss": 0.8583,
+      "step": 31446
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.654234285714286e-06,
+      "loss": 0.9254,
+      "step": 31447
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.653965714285714e-06,
+      "loss": 0.9506,
+      "step": 31448
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.653697142857144e-06,
+      "loss": 1.3567,
+      "step": 31449
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.653428571428572e-06,
+      "loss": 1.2414,
+      "step": 31450
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.65316e-06,
+      "loss": 1.1416,
+      "step": 31451
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.652891428571429e-06,
+      "loss": 1.1465,
+      "step": 31452
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.652622857142857e-06,
+      "loss": 1.2854,
+      "step": 31453
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.652354285714286e-06,
+      "loss": 1.0354,
+      "step": 31454
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6520857142857145e-06,
+      "loss": 1.2912,
+      "step": 31455
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.651817142857143e-06,
+      "loss": 1.0321,
+      "step": 31456
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.651548571428572e-06,
+      "loss": 1.0085,
+      "step": 31457
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.65128e-06,
+      "loss": 1.3275,
+      "step": 31458
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.651011428571429e-06,
+      "loss": 1.3053,
+      "step": 31459
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.650742857142857e-06,
+      "loss": 1.1269,
+      "step": 31460
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.650474285714285e-06,
+      "loss": 1.0713,
+      "step": 31461
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.650205714285714e-06,
+      "loss": 1.1223,
+      "step": 31462
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.649937142857143e-06,
+      "loss": 0.9717,
+      "step": 31463
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.649668571428572e-06,
+      "loss": 1.2349,
+      "step": 31464
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6494000000000005e-06,
+      "loss": 1.5343,
+      "step": 31465
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.649131428571429e-06,
+      "loss": 1.1471,
+      "step": 31466
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.648862857142858e-06,
+      "loss": 0.9663,
+      "step": 31467
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.648594285714286e-06,
+      "loss": 1.2513,
+      "step": 31468
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.648325714285714e-06,
+      "loss": 1.1884,
+      "step": 31469
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.648057142857143e-06,
+      "loss": 0.9434,
+      "step": 31470
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.647788571428571e-06,
+      "loss": 1.3797,
+      "step": 31471
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.64752e-06,
+      "loss": 1.0658,
+      "step": 31472
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6472514285714284e-06,
+      "loss": 0.9275,
+      "step": 31473
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.646982857142857e-06,
+      "loss": 0.8241,
+      "step": 31474
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.646714285714286e-06,
+      "loss": 1.3157,
+      "step": 31475
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.646445714285715e-06,
+      "loss": 1.3776,
+      "step": 31476
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.646177142857144e-06,
+      "loss": 0.7702,
+      "step": 31477
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.645908571428572e-06,
+      "loss": 1.0087,
+      "step": 31478
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.64564e-06,
+      "loss": 1.1913,
+      "step": 31479
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.645371428571429e-06,
+      "loss": 0.9174,
+      "step": 31480
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.645102857142857e-06,
+      "loss": 1.3761,
+      "step": 31481
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.644834285714286e-06,
+      "loss": 1.4188,
+      "step": 31482
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6445657142857144e-06,
+      "loss": 0.9125,
+      "step": 31483
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.644297142857143e-06,
+      "loss": 1.4673,
+      "step": 31484
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.644028571428572e-06,
+      "loss": 1.0416,
+      "step": 31485
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.64376e-06,
+      "loss": 1.4028,
+      "step": 31486
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.643491428571429e-06,
+      "loss": 1.0813,
+      "step": 31487
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.643222857142857e-06,
+      "loss": 1.1396,
+      "step": 31488
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.642954285714286e-06,
+      "loss": 1.0838,
+      "step": 31489
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.642685714285715e-06,
+      "loss": 1.1679,
+      "step": 31490
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.642417142857143e-06,
+      "loss": 0.8449,
+      "step": 31491
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.642148571428571e-06,
+      "loss": 0.8488,
+      "step": 31492
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6418800000000004e-06,
+      "loss": 1.1316,
+      "step": 31493
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.641611428571429e-06,
+      "loss": 1.4158,
+      "step": 31494
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.641342857142858e-06,
+      "loss": 0.9074,
+      "step": 31495
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.641074285714286e-06,
+      "loss": 0.7563,
+      "step": 31496
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.640805714285714e-06,
+      "loss": 1.0571,
+      "step": 31497
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.640537142857143e-06,
+      "loss": 1.0267,
+      "step": 31498
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.640268571428571e-06,
+      "loss": 0.7769,
+      "step": 31499
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.64e-06,
+      "loss": 1.1705,
+      "step": 31500
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.639731428571428e-06,
+      "loss": 1.3741,
+      "step": 31501
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.639462857142857e-06,
+      "loss": 1.5193,
+      "step": 31502
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6391942857142864e-06,
+      "loss": 1.3336,
+      "step": 31503
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.638925714285715e-06,
+      "loss": 0.8943,
+      "step": 31504
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.638657142857144e-06,
+      "loss": 0.9356,
+      "step": 31505
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.638388571428572e-06,
+      "loss": 0.912,
+      "step": 31506
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.63812e-06,
+      "loss": 1.2998,
+      "step": 31507
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.637851428571429e-06,
+      "loss": 1.1692,
+      "step": 31508
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.637582857142857e-06,
+      "loss": 1.5188,
+      "step": 31509
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.637314285714286e-06,
+      "loss": 1.0153,
+      "step": 31510
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.637045714285714e-06,
+      "loss": 1.0911,
+      "step": 31511
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6367771428571426e-06,
+      "loss": 1.2517,
+      "step": 31512
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.636508571428572e-06,
+      "loss": 1.169,
+      "step": 31513
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.63624e-06,
+      "loss": 1.4102,
+      "step": 31514
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.635971428571429e-06,
+      "loss": 1.0664,
+      "step": 31515
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.635702857142858e-06,
+      "loss": 1.4815,
+      "step": 31516
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.635434285714286e-06,
+      "loss": 1.318,
+      "step": 31517
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.635165714285715e-06,
+      "loss": 0.915,
+      "step": 31518
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.634897142857143e-06,
+      "loss": 1.5686,
+      "step": 31519
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.634628571428571e-06,
+      "loss": 1.1167,
+      "step": 31520
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.63436e-06,
+      "loss": 1.0448,
+      "step": 31521
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6340914285714286e-06,
+      "loss": 1.101,
+      "step": 31522
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.633822857142858e-06,
+      "loss": 1.3104,
+      "step": 31523
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.633554285714286e-06,
+      "loss": 1.2469,
+      "step": 31524
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.633285714285714e-06,
+      "loss": 1.1866,
+      "step": 31525
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.633017142857143e-06,
+      "loss": 1.1259,
+      "step": 31526
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.632748571428571e-06,
+      "loss": 1.1806,
+      "step": 31527
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.63248e-06,
+      "loss": 1.1332,
+      "step": 31528
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.632211428571429e-06,
+      "loss": 1.3067,
+      "step": 31529
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.631942857142857e-06,
+      "loss": 1.244,
+      "step": 31530
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.631674285714286e-06,
+      "loss": 1.289,
+      "step": 31531
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6314057142857146e-06,
+      "loss": 1.0691,
+      "step": 31532
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.631137142857143e-06,
+      "loss": 1.3025,
+      "step": 31533
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.630868571428572e-06,
+      "loss": 1.1109,
+      "step": 31534
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6306e-06,
+      "loss": 1.1143,
+      "step": 31535
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.630331428571429e-06,
+      "loss": 1.1522,
+      "step": 31536
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.630062857142857e-06,
+      "loss": 0.6963,
+      "step": 31537
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.629794285714285e-06,
+      "loss": 1.1271,
+      "step": 31538
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.629525714285714e-06,
+      "loss": 1.3353,
+      "step": 31539
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6292571428571425e-06,
+      "loss": 1.0446,
+      "step": 31540
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6289885714285715e-06,
+      "loss": 0.7765,
+      "step": 31541
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6287200000000005e-06,
+      "loss": 1.0364,
+      "step": 31542
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.628451428571429e-06,
+      "loss": 1.3395,
+      "step": 31543
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.628182857142858e-06,
+      "loss": 0.9603,
+      "step": 31544
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.627914285714286e-06,
+      "loss": 1.5476,
+      "step": 31545
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.627645714285715e-06,
+      "loss": 1.3499,
+      "step": 31546
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.627377142857143e-06,
+      "loss": 1.1718,
+      "step": 31547
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.627108571428571e-06,
+      "loss": 1.1298,
+      "step": 31548
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.62684e-06,
+      "loss": 1.0813,
+      "step": 31549
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6265714285714285e-06,
+      "loss": 1.4257,
+      "step": 31550
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6263028571428575e-06,
+      "loss": 1.2216,
+      "step": 31551
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.626034285714286e-06,
+      "loss": 0.9982,
+      "step": 31552
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.625765714285714e-06,
+      "loss": 0.7602,
+      "step": 31553
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.625497142857143e-06,
+      "loss": 1.0749,
+      "step": 31554
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.625228571428572e-06,
+      "loss": 1.057,
+      "step": 31555
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.62496e-06,
+      "loss": 1.0811,
+      "step": 31556
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.624691428571429e-06,
+      "loss": 0.9781,
+      "step": 31557
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.624422857142857e-06,
+      "loss": 0.8843,
+      "step": 31558
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.624154285714286e-06,
+      "loss": 1.1699,
+      "step": 31559
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6238857142857145e-06,
+      "loss": 1.0502,
+      "step": 31560
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.623617142857143e-06,
+      "loss": 1.2314,
+      "step": 31561
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.623348571428572e-06,
+      "loss": 1.1465,
+      "step": 31562
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.62308e-06,
+      "loss": 1.404,
+      "step": 31563
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.622811428571429e-06,
+      "loss": 0.9282,
+      "step": 31564
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.622542857142857e-06,
+      "loss": 1.1324,
+      "step": 31565
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.622274285714285e-06,
+      "loss": 0.9826,
+      "step": 31566
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.622005714285714e-06,
+      "loss": 1.1662,
+      "step": 31567
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.621737142857143e-06,
+      "loss": 1.4155,
+      "step": 31568
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.621468571428572e-06,
+      "loss": 1.1112,
+      "step": 31569
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6212000000000005e-06,
+      "loss": 1.3091,
+      "step": 31570
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.620931428571429e-06,
+      "loss": 1.4239,
+      "step": 31571
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.620662857142858e-06,
+      "loss": 1.1827,
+      "step": 31572
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.620394285714286e-06,
+      "loss": 1.1459,
+      "step": 31573
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.620125714285715e-06,
+      "loss": 1.217,
+      "step": 31574
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.619857142857143e-06,
+      "loss": 0.6647,
+      "step": 31575
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.619588571428571e-06,
+      "loss": 1.2421,
+      "step": 31576
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.61932e-06,
+      "loss": 1.0681,
+      "step": 31577
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.6190514285714284e-06,
+      "loss": 0.7004,
+      "step": 31578
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.618782857142857e-06,
+      "loss": 1.4035,
+      "step": 31579
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.618514285714286e-06,
+      "loss": 1.0654,
+      "step": 31580
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.618245714285715e-06,
+      "loss": 1.1295,
+      "step": 31581
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.617977142857144e-06,
+      "loss": 1.0924,
+      "step": 31582
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.617708571428572e-06,
+      "loss": 0.7865,
+      "step": 31583
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.61744e-06,
+      "loss": 1.769,
+      "step": 31584
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.617171428571429e-06,
+      "loss": 1.4621,
+      "step": 31585
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.616902857142857e-06,
+      "loss": 1.1463,
+      "step": 31586
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 5.616634285714286e-06,
+      "loss": 1.2713,
+      "step": 31587
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6163657142857144e-06,
+      "loss": 1.2749,
+      "step": 31588
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.616097142857143e-06,
+      "loss": 1.5991,
+      "step": 31589
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.615828571428572e-06,
+      "loss": 0.6974,
+      "step": 31590
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.61556e-06,
+      "loss": 1.2287,
+      "step": 31591
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.615291428571429e-06,
+      "loss": 0.8836,
+      "step": 31592
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.615022857142857e-06,
+      "loss": 1.2434,
+      "step": 31593
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.614754285714286e-06,
+      "loss": 0.9528,
+      "step": 31594
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.614485714285715e-06,
+      "loss": 0.836,
+      "step": 31595
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.614217142857143e-06,
+      "loss": 1.0342,
+      "step": 31596
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.613948571428572e-06,
+      "loss": 1.1497,
+      "step": 31597
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6136800000000004e-06,
+      "loss": 1.0122,
+      "step": 31598
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.613411428571429e-06,
+      "loss": 0.9585,
+      "step": 31599
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.613142857142858e-06,
+      "loss": 1.1244,
+      "step": 31600
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.612874285714286e-06,
+      "loss": 1.2192,
+      "step": 31601
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.612605714285714e-06,
+      "loss": 0.8666,
+      "step": 31602
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.612337142857143e-06,
+      "loss": 1.0373,
+      "step": 31603
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.612068571428571e-06,
+      "loss": 0.7213,
+      "step": 31604
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6118e-06,
+      "loss": 1.0082,
+      "step": 31605
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.611531428571428e-06,
+      "loss": 1.2418,
+      "step": 31606
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.611262857142857e-06,
+      "loss": 1.4795,
+      "step": 31607
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6109942857142864e-06,
+      "loss": 1.3155,
+      "step": 31608
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.610725714285715e-06,
+      "loss": 1.4837,
+      "step": 31609
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.610457142857144e-06,
+      "loss": 1.3919,
+      "step": 31610
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.610188571428572e-06,
+      "loss": 1.0606,
+      "step": 31611
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.60992e-06,
+      "loss": 1.2854,
+      "step": 31612
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.609651428571429e-06,
+      "loss": 1.2238,
+      "step": 31613
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.609382857142857e-06,
+      "loss": 0.9573,
+      "step": 31614
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.609114285714286e-06,
+      "loss": 1.0136,
+      "step": 31615
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.608845714285714e-06,
+      "loss": 1.5174,
+      "step": 31616
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6085771428571426e-06,
+      "loss": 1.1986,
+      "step": 31617
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.608308571428572e-06,
+      "loss": 1.2222,
+      "step": 31618
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.60804e-06,
+      "loss": 1.4582,
+      "step": 31619
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.607771428571429e-06,
+      "loss": 1.4659,
+      "step": 31620
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.607502857142858e-06,
+      "loss": 1.0039,
+      "step": 31621
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.607234285714286e-06,
+      "loss": 0.9007,
+      "step": 31622
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.606965714285715e-06,
+      "loss": 1.3921,
+      "step": 31623
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.606697142857143e-06,
+      "loss": 1.2557,
+      "step": 31624
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.606428571428571e-06,
+      "loss": 1.2301,
+      "step": 31625
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.60616e-06,
+      "loss": 1.3052,
+      "step": 31626
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6058914285714286e-06,
+      "loss": 0.8637,
+      "step": 31627
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.605622857142858e-06,
+      "loss": 1.1497,
+      "step": 31628
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.605354285714286e-06,
+      "loss": 0.8598,
+      "step": 31629
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.605085714285714e-06,
+      "loss": 0.8855,
+      "step": 31630
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.604817142857143e-06,
+      "loss": 1.4084,
+      "step": 31631
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.604548571428571e-06,
+      "loss": 1.3469,
+      "step": 31632
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.60428e-06,
+      "loss": 0.653,
+      "step": 31633
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.604011428571429e-06,
+      "loss": 1.5179,
+      "step": 31634
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.603742857142857e-06,
+      "loss": 1.2929,
+      "step": 31635
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.603474285714286e-06,
+      "loss": 0.9034,
+      "step": 31636
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6032057142857146e-06,
+      "loss": 0.9002,
+      "step": 31637
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6029371428571436e-06,
+      "loss": 1.0701,
+      "step": 31638
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.602668571428572e-06,
+      "loss": 0.9279,
+      "step": 31639
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6024e-06,
+      "loss": 1.3206,
+      "step": 31640
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.602131428571429e-06,
+      "loss": 1.0164,
+      "step": 31641
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.601862857142857e-06,
+      "loss": 1.2056,
+      "step": 31642
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.601594285714285e-06,
+      "loss": 1.1769,
+      "step": 31643
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.601325714285714e-06,
+      "loss": 1.5952,
+      "step": 31644
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6010571428571425e-06,
+      "loss": 1.1019,
+      "step": 31645
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6007885714285715e-06,
+      "loss": 0.9419,
+      "step": 31646
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.6005200000000005e-06,
+      "loss": 1.559,
+      "step": 31647
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.600251428571429e-06,
+      "loss": 0.6684,
+      "step": 31648
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.599982857142858e-06,
+      "loss": 1.3045,
+      "step": 31649
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.599714285714286e-06,
+      "loss": 1.458,
+      "step": 31650
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.599445714285715e-06,
+      "loss": 0.9486,
+      "step": 31651
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.599177142857143e-06,
+      "loss": 1.5598,
+      "step": 31652
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.598908571428571e-06,
+      "loss": 0.9085,
+      "step": 31653
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.59864e-06,
+      "loss": 1.0873,
+      "step": 31654
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5983714285714285e-06,
+      "loss": 1.1915,
+      "step": 31655
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5981028571428575e-06,
+      "loss": 0.7852,
+      "step": 31656
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.597834285714286e-06,
+      "loss": 1.5413,
+      "step": 31657
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.597565714285714e-06,
+      "loss": 1.1261,
+      "step": 31658
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.597297142857143e-06,
+      "loss": 0.9894,
+      "step": 31659
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.597028571428572e-06,
+      "loss": 1.0122,
+      "step": 31660
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.596760000000001e-06,
+      "loss": 0.8743,
+      "step": 31661
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.596491428571429e-06,
+      "loss": 0.9419,
+      "step": 31662
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.596222857142857e-06,
+      "loss": 1.1917,
+      "step": 31663
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.595954285714286e-06,
+      "loss": 1.3551,
+      "step": 31664
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5956857142857145e-06,
+      "loss": 1.2593,
+      "step": 31665
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.595417142857143e-06,
+      "loss": 1.0275,
+      "step": 31666
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.595148571428572e-06,
+      "loss": 1.366,
+      "step": 31667
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.59488e-06,
+      "loss": 1.4625,
+      "step": 31668
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.594611428571429e-06,
+      "loss": 0.9286,
+      "step": 31669
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.594342857142857e-06,
+      "loss": 1.2249,
+      "step": 31670
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.594074285714285e-06,
+      "loss": 1.1175,
+      "step": 31671
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.593805714285714e-06,
+      "loss": 0.9657,
+      "step": 31672
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.593537142857143e-06,
+      "loss": 0.9989,
+      "step": 31673
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.593268571428572e-06,
+      "loss": 1.4184,
+      "step": 31674
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5930000000000005e-06,
+      "loss": 0.9409,
+      "step": 31675
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.592731428571429e-06,
+      "loss": 0.9821,
+      "step": 31676
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.592462857142858e-06,
+      "loss": 1.0126,
+      "step": 31677
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.592194285714286e-06,
+      "loss": 1.4066,
+      "step": 31678
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.591925714285715e-06,
+      "loss": 1.0836,
+      "step": 31679
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.591657142857143e-06,
+      "loss": 1.1255,
+      "step": 31680
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.591388571428571e-06,
+      "loss": 1.1242,
+      "step": 31681
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.59112e-06,
+      "loss": 0.8983,
+      "step": 31682
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5908514285714284e-06,
+      "loss": 1.2577,
+      "step": 31683
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5905828571428575e-06,
+      "loss": 1.6965,
+      "step": 31684
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.590314285714286e-06,
+      "loss": 1.1287,
+      "step": 31685
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.590045714285715e-06,
+      "loss": 0.7569,
+      "step": 31686
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.589777142857144e-06,
+      "loss": 1.1497,
+      "step": 31687
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.589508571428572e-06,
+      "loss": 1.3244,
+      "step": 31688
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.58924e-06,
+      "loss": 1.0199,
+      "step": 31689
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.588971428571429e-06,
+      "loss": 1.2113,
+      "step": 31690
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.588702857142857e-06,
+      "loss": 1.0838,
+      "step": 31691
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.588434285714286e-06,
+      "loss": 0.911,
+      "step": 31692
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5881657142857144e-06,
+      "loss": 1.3377,
+      "step": 31693
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.587897142857143e-06,
+      "loss": 1.1676,
+      "step": 31694
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.587628571428572e-06,
+      "loss": 0.8933,
+      "step": 31695
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.58736e-06,
+      "loss": 1.2035,
+      "step": 31696
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.587091428571429e-06,
+      "loss": 1.1443,
+      "step": 31697
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.586822857142857e-06,
+      "loss": 1.5282,
+      "step": 31698
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.586554285714286e-06,
+      "loss": 1.0593,
+      "step": 31699
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.586285714285715e-06,
+      "loss": 1.1631,
+      "step": 31700
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.586017142857143e-06,
+      "loss": 1.0268,
+      "step": 31701
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.585748571428572e-06,
+      "loss": 1.5658,
+      "step": 31702
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5854800000000004e-06,
+      "loss": 1.3256,
+      "step": 31703
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.585211428571429e-06,
+      "loss": 1.1392,
+      "step": 31704
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.584942857142858e-06,
+      "loss": 1.1766,
+      "step": 31705
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.584674285714286e-06,
+      "loss": 1.0799,
+      "step": 31706
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.584405714285714e-06,
+      "loss": 1.1221,
+      "step": 31707
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.584137142857143e-06,
+      "loss": 1.1794,
+      "step": 31708
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.583868571428571e-06,
+      "loss": 0.9995,
+      "step": 31709
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5836e-06,
+      "loss": 0.8331,
+      "step": 31710
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.583331428571428e-06,
+      "loss": 1.1613,
+      "step": 31711
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.583062857142857e-06,
+      "loss": 1.0996,
+      "step": 31712
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5827942857142864e-06,
+      "loss": 1.1921,
+      "step": 31713
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.582525714285715e-06,
+      "loss": 0.9627,
+      "step": 31714
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.582257142857144e-06,
+      "loss": 0.9824,
+      "step": 31715
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.581988571428572e-06,
+      "loss": 1.1186,
+      "step": 31716
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.58172e-06,
+      "loss": 1.4761,
+      "step": 31717
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.581451428571429e-06,
+      "loss": 1.2419,
+      "step": 31718
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.581182857142857e-06,
+      "loss": 0.9477,
+      "step": 31719
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.580914285714286e-06,
+      "loss": 0.7119,
+      "step": 31720
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.580645714285714e-06,
+      "loss": 0.9888,
+      "step": 31721
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5803771428571426e-06,
+      "loss": 1.1773,
+      "step": 31722
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.580108571428572e-06,
+      "loss": 0.8719,
+      "step": 31723
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.57984e-06,
+      "loss": 0.9357,
+      "step": 31724
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.579571428571429e-06,
+      "loss": 1.0252,
+      "step": 31725
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.579302857142858e-06,
+      "loss": 0.6783,
+      "step": 31726
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.579034285714286e-06,
+      "loss": 0.9728,
+      "step": 31727
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.578765714285715e-06,
+      "loss": 1.0486,
+      "step": 31728
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.578497142857143e-06,
+      "loss": 1.1172,
+      "step": 31729
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.578228571428571e-06,
+      "loss": 1.4219,
+      "step": 31730
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.57796e-06,
+      "loss": 1.4179,
+      "step": 31731
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5776914285714286e-06,
+      "loss": 1.417,
+      "step": 31732
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.577422857142858e-06,
+      "loss": 1.1345,
+      "step": 31733
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.577154285714286e-06,
+      "loss": 0.8555,
+      "step": 31734
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.576885714285714e-06,
+      "loss": 1.3914,
+      "step": 31735
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.576617142857143e-06,
+      "loss": 0.9629,
+      "step": 31736
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.576348571428571e-06,
+      "loss": 1.4401,
+      "step": 31737
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.57608e-06,
+      "loss": 1.3075,
+      "step": 31738
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.575811428571429e-06,
+      "loss": 0.8853,
+      "step": 31739
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.575542857142857e-06,
+      "loss": 0.7995,
+      "step": 31740
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.575274285714286e-06,
+      "loss": 1.0917,
+      "step": 31741
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5750057142857146e-06,
+      "loss": 1.0104,
+      "step": 31742
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5747371428571436e-06,
+      "loss": 0.8131,
+      "step": 31743
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.574468571428572e-06,
+      "loss": 1.3811,
+      "step": 31744
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5742e-06,
+      "loss": 0.8691,
+      "step": 31745
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.573931428571429e-06,
+      "loss": 0.9297,
+      "step": 31746
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.573662857142857e-06,
+      "loss": 1.0459,
+      "step": 31747
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.573394285714286e-06,
+      "loss": 1.1263,
+      "step": 31748
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.573125714285714e-06,
+      "loss": 0.9704,
+      "step": 31749
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5728571428571425e-06,
+      "loss": 0.76,
+      "step": 31750
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5725885714285715e-06,
+      "loss": 1.3306,
+      "step": 31751
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5723200000000005e-06,
+      "loss": 0.9969,
+      "step": 31752
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.572051428571429e-06,
+      "loss": 1.1141,
+      "step": 31753
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.571782857142858e-06,
+      "loss": 1.4059,
+      "step": 31754
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.571514285714286e-06,
+      "loss": 0.8965,
+      "step": 31755
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.571245714285715e-06,
+      "loss": 1.2252,
+      "step": 31756
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.570977142857143e-06,
+      "loss": 1.6095,
+      "step": 31757
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.570708571428571e-06,
+      "loss": 0.8863,
+      "step": 31758
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.57044e-06,
+      "loss": 1.2779,
+      "step": 31759
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5701714285714285e-06,
+      "loss": 1.5343,
+      "step": 31760
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.5699028571428575e-06,
+      "loss": 0.864,
+      "step": 31761
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 5.569634285714286e-06,
+      "loss": 1.2148,
+      "step": 31762
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.569365714285714e-06,
+      "loss": 1.0491,
+      "step": 31763
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.569097142857143e-06,
+      "loss": 1.4056,
+      "step": 31764
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.568828571428572e-06,
+      "loss": 0.957,
+      "step": 31765
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.568560000000001e-06,
+      "loss": 1.1555,
+      "step": 31766
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.568291428571429e-06,
+      "loss": 1.551,
+      "step": 31767
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.568022857142857e-06,
+      "loss": 1.0265,
+      "step": 31768
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.567754285714286e-06,
+      "loss": 0.9647,
+      "step": 31769
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5674857142857145e-06,
+      "loss": 1.4278,
+      "step": 31770
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5672171428571435e-06,
+      "loss": 1.1097,
+      "step": 31771
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.566948571428572e-06,
+      "loss": 0.9274,
+      "step": 31772
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.56668e-06,
+      "loss": 0.6511,
+      "step": 31773
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.566411428571429e-06,
+      "loss": 1.3245,
+      "step": 31774
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.566142857142857e-06,
+      "loss": 0.9851,
+      "step": 31775
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.565874285714285e-06,
+      "loss": 1.3346,
+      "step": 31776
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.565605714285714e-06,
+      "loss": 0.9541,
+      "step": 31777
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.565337142857143e-06,
+      "loss": 1.148,
+      "step": 31778
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.565068571428572e-06,
+      "loss": 1.1374,
+      "step": 31779
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5648000000000005e-06,
+      "loss": 0.8889,
+      "step": 31780
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.564531428571429e-06,
+      "loss": 1.1579,
+      "step": 31781
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.564262857142858e-06,
+      "loss": 0.9044,
+      "step": 31782
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.563994285714286e-06,
+      "loss": 1.0108,
+      "step": 31783
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.563725714285715e-06,
+      "loss": 1.6079,
+      "step": 31784
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.563457142857143e-06,
+      "loss": 1.3952,
+      "step": 31785
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.563188571428571e-06,
+      "loss": 0.9419,
+      "step": 31786
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.56292e-06,
+      "loss": 1.1872,
+      "step": 31787
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5626514285714284e-06,
+      "loss": 1.1665,
+      "step": 31788
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5623828571428575e-06,
+      "loss": 0.9969,
+      "step": 31789
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.562114285714286e-06,
+      "loss": 1.1121,
+      "step": 31790
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.561845714285715e-06,
+      "loss": 1.0719,
+      "step": 31791
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.561577142857144e-06,
+      "loss": 1.4308,
+      "step": 31792
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.561308571428572e-06,
+      "loss": 1.3672,
+      "step": 31793
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.56104e-06,
+      "loss": 1.5399,
+      "step": 31794
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.560771428571429e-06,
+      "loss": 1.2548,
+      "step": 31795
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.560502857142857e-06,
+      "loss": 1.2711,
+      "step": 31796
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.560234285714286e-06,
+      "loss": 0.9439,
+      "step": 31797
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5599657142857144e-06,
+      "loss": 1.4221,
+      "step": 31798
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.559697142857143e-06,
+      "loss": 1.5282,
+      "step": 31799
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.559428571428572e-06,
+      "loss": 0.8713,
+      "step": 31800
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.55916e-06,
+      "loss": 1.1679,
+      "step": 31801
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.558891428571429e-06,
+      "loss": 1.3288,
+      "step": 31802
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.558622857142857e-06,
+      "loss": 1.1142,
+      "step": 31803
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.558354285714285e-06,
+      "loss": 1.2999,
+      "step": 31804
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.558085714285715e-06,
+      "loss": 1.2369,
+      "step": 31805
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.557817142857143e-06,
+      "loss": 1.1226,
+      "step": 31806
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.557548571428572e-06,
+      "loss": 1.1434,
+      "step": 31807
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5572800000000004e-06,
+      "loss": 1.06,
+      "step": 31808
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.557011428571429e-06,
+      "loss": 1.3002,
+      "step": 31809
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.556742857142858e-06,
+      "loss": 1.3142,
+      "step": 31810
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.556474285714286e-06,
+      "loss": 0.6765,
+      "step": 31811
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.556205714285715e-06,
+      "loss": 1.0815,
+      "step": 31812
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.555937142857143e-06,
+      "loss": 1.0209,
+      "step": 31813
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.555668571428571e-06,
+      "loss": 1.1711,
+      "step": 31814
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5554e-06,
+      "loss": 1.2489,
+      "step": 31815
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.555131428571428e-06,
+      "loss": 0.8861,
+      "step": 31816
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5548628571428566e-06,
+      "loss": 1.2006,
+      "step": 31817
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5545942857142864e-06,
+      "loss": 1.0167,
+      "step": 31818
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.554325714285715e-06,
+      "loss": 1.1561,
+      "step": 31819
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.554057142857144e-06,
+      "loss": 1.3061,
+      "step": 31820
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.553788571428572e-06,
+      "loss": 1.1493,
+      "step": 31821
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.55352e-06,
+      "loss": 1.2359,
+      "step": 31822
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.553251428571429e-06,
+      "loss": 0.9813,
+      "step": 31823
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.552982857142857e-06,
+      "loss": 1.3501,
+      "step": 31824
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.552714285714286e-06,
+      "loss": 1.2503,
+      "step": 31825
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.552445714285714e-06,
+      "loss": 0.9306,
+      "step": 31826
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5521771428571426e-06,
+      "loss": 1.5265,
+      "step": 31827
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.551908571428572e-06,
+      "loss": 1.2283,
+      "step": 31828
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.55164e-06,
+      "loss": 1.1758,
+      "step": 31829
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.551371428571429e-06,
+      "loss": 0.7671,
+      "step": 31830
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.551102857142858e-06,
+      "loss": 0.6865,
+      "step": 31831
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.550834285714286e-06,
+      "loss": 1.4632,
+      "step": 31832
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.550565714285715e-06,
+      "loss": 1.131,
+      "step": 31833
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.550297142857143e-06,
+      "loss": 0.8724,
+      "step": 31834
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.550028571428572e-06,
+      "loss": 1.1555,
+      "step": 31835
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.54976e-06,
+      "loss": 1.2807,
+      "step": 31836
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5494914285714286e-06,
+      "loss": 1.0873,
+      "step": 31837
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5492228571428576e-06,
+      "loss": 1.078,
+      "step": 31838
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.548954285714286e-06,
+      "loss": 1.2469,
+      "step": 31839
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.548685714285714e-06,
+      "loss": 1.1714,
+      "step": 31840
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.548417142857143e-06,
+      "loss": 0.9547,
+      "step": 31841
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.548148571428571e-06,
+      "loss": 1.2243,
+      "step": 31842
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.54788e-06,
+      "loss": 1.3709,
+      "step": 31843
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.547611428571429e-06,
+      "loss": 0.9491,
+      "step": 31844
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.547342857142857e-06,
+      "loss": 1.234,
+      "step": 31845
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.547074285714286e-06,
+      "loss": 1.1168,
+      "step": 31846
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5468057142857145e-06,
+      "loss": 0.8946,
+      "step": 31847
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5465371428571436e-06,
+      "loss": 1.2246,
+      "step": 31848
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.546268571428572e-06,
+      "loss": 1.0721,
+      "step": 31849
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.546e-06,
+      "loss": 0.9125,
+      "step": 31850
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.545731428571429e-06,
+      "loss": 1.3744,
+      "step": 31851
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.545462857142857e-06,
+      "loss": 1.1877,
+      "step": 31852
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.545194285714286e-06,
+      "loss": 1.4889,
+      "step": 31853
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.544925714285714e-06,
+      "loss": 0.9236,
+      "step": 31854
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5446571428571425e-06,
+      "loss": 0.9302,
+      "step": 31855
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5443885714285715e-06,
+      "loss": 0.8314,
+      "step": 31856
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.54412e-06,
+      "loss": 0.8299,
+      "step": 31857
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.543851428571429e-06,
+      "loss": 1.6712,
+      "step": 31858
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.543582857142858e-06,
+      "loss": 0.8734,
+      "step": 31859
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.543314285714286e-06,
+      "loss": 0.8119,
+      "step": 31860
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.543045714285715e-06,
+      "loss": 1.1857,
+      "step": 31861
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.542777142857143e-06,
+      "loss": 1.1716,
+      "step": 31862
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.542508571428571e-06,
+      "loss": 0.958,
+      "step": 31863
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.54224e-06,
+      "loss": 1.1283,
+      "step": 31864
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5419714285714285e-06,
+      "loss": 0.7479,
+      "step": 31865
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5417028571428575e-06,
+      "loss": 0.719,
+      "step": 31866
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.541434285714286e-06,
+      "loss": 1.3349,
+      "step": 31867
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.541165714285714e-06,
+      "loss": 1.3462,
+      "step": 31868
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.540897142857143e-06,
+      "loss": 1.0681,
+      "step": 31869
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.540628571428571e-06,
+      "loss": 1.2379,
+      "step": 31870
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.540360000000001e-06,
+      "loss": 1.1358,
+      "step": 31871
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.540091428571429e-06,
+      "loss": 1.1017,
+      "step": 31872
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.539822857142857e-06,
+      "loss": 1.3629,
+      "step": 31873
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.539554285714286e-06,
+      "loss": 1.0872,
+      "step": 31874
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5392857142857145e-06,
+      "loss": 1.7075,
+      "step": 31875
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5390171428571435e-06,
+      "loss": 1.1908,
+      "step": 31876
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.538748571428572e-06,
+      "loss": 0.8402,
+      "step": 31877
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.53848e-06,
+      "loss": 1.4578,
+      "step": 31878
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.538211428571429e-06,
+      "loss": 1.2748,
+      "step": 31879
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.537942857142857e-06,
+      "loss": 0.8544,
+      "step": 31880
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.537674285714285e-06,
+      "loss": 1.127,
+      "step": 31881
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.537405714285714e-06,
+      "loss": 1.1873,
+      "step": 31882
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5371371428571424e-06,
+      "loss": 1.0137,
+      "step": 31883
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.536868571428572e-06,
+      "loss": 1.3502,
+      "step": 31884
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5366000000000005e-06,
+      "loss": 1.0213,
+      "step": 31885
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.536331428571429e-06,
+      "loss": 1.0981,
+      "step": 31886
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.536062857142858e-06,
+      "loss": 1.1319,
+      "step": 31887
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.535794285714286e-06,
+      "loss": 1.1331,
+      "step": 31888
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.535525714285715e-06,
+      "loss": 1.1204,
+      "step": 31889
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.535257142857143e-06,
+      "loss": 1.5079,
+      "step": 31890
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.534988571428571e-06,
+      "loss": 1.1227,
+      "step": 31891
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.53472e-06,
+      "loss": 1.1817,
+      "step": 31892
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5344514285714284e-06,
+      "loss": 1.1996,
+      "step": 31893
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5341828571428575e-06,
+      "loss": 1.4384,
+      "step": 31894
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.533914285714286e-06,
+      "loss": 0.8255,
+      "step": 31895
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.533645714285714e-06,
+      "loss": 1.42,
+      "step": 31896
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.533377142857144e-06,
+      "loss": 1.2636,
+      "step": 31897
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.533108571428572e-06,
+      "loss": 1.1559,
+      "step": 31898
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.532840000000001e-06,
+      "loss": 1.1711,
+      "step": 31899
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.532571428571429e-06,
+      "loss": 0.827,
+      "step": 31900
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.532302857142857e-06,
+      "loss": 1.5033,
+      "step": 31901
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.532034285714286e-06,
+      "loss": 1.1658,
+      "step": 31902
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5317657142857144e-06,
+      "loss": 1.3296,
+      "step": 31903
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.531497142857143e-06,
+      "loss": 0.992,
+      "step": 31904
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.531228571428572e-06,
+      "loss": 1.3782,
+      "step": 31905
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.53096e-06,
+      "loss": 1.125,
+      "step": 31906
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.530691428571429e-06,
+      "loss": 1.2708,
+      "step": 31907
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.530422857142857e-06,
+      "loss": 1.1897,
+      "step": 31908
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.530154285714285e-06,
+      "loss": 1.1618,
+      "step": 31909
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.529885714285714e-06,
+      "loss": 1.1141,
+      "step": 31910
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.529617142857143e-06,
+      "loss": 1.1451,
+      "step": 31911
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.529348571428572e-06,
+      "loss": 1.2436,
+      "step": 31912
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5290800000000004e-06,
+      "loss": 0.9128,
+      "step": 31913
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.528811428571429e-06,
+      "loss": 1.2509,
+      "step": 31914
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.528542857142858e-06,
+      "loss": 1.6298,
+      "step": 31915
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.528274285714286e-06,
+      "loss": 1.2362,
+      "step": 31916
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.528005714285715e-06,
+      "loss": 0.8242,
+      "step": 31917
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.527737142857143e-06,
+      "loss": 1.0676,
+      "step": 31918
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.527468571428571e-06,
+      "loss": 0.9668,
+      "step": 31919
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5272e-06,
+      "loss": 1.0017,
+      "step": 31920
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.526931428571428e-06,
+      "loss": 1.1386,
+      "step": 31921
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.526662857142857e-06,
+      "loss": 1.2152,
+      "step": 31922
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.526394285714286e-06,
+      "loss": 0.9373,
+      "step": 31923
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.526125714285715e-06,
+      "loss": 1.3725,
+      "step": 31924
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.525857142857144e-06,
+      "loss": 1.2695,
+      "step": 31925
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.525588571428572e-06,
+      "loss": 1.3582,
+      "step": 31926
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.52532e-06,
+      "loss": 1.0999,
+      "step": 31927
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.525051428571429e-06,
+      "loss": 1.1622,
+      "step": 31928
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.524782857142857e-06,
+      "loss": 0.8355,
+      "step": 31929
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.524514285714286e-06,
+      "loss": 1.3532,
+      "step": 31930
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.524245714285714e-06,
+      "loss": 0.9398,
+      "step": 31931
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.5239771428571426e-06,
+      "loss": 1.0179,
+      "step": 31932
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.523708571428572e-06,
+      "loss": 0.8344,
+      "step": 31933
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.52344e-06,
+      "loss": 0.9428,
+      "step": 31934
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.523171428571429e-06,
+      "loss": 1.0923,
+      "step": 31935
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.522902857142857e-06,
+      "loss": 1.1346,
+      "step": 31936
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 5.522634285714286e-06,
+      "loss": 1.2322,
+      "step": 31937
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.522365714285715e-06,
+      "loss": 1.0202,
+      "step": 31938
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.522097142857143e-06,
+      "loss": 1.0185,
+      "step": 31939
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.521828571428572e-06,
+      "loss": 1.3334,
+      "step": 31940
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.52156e-06,
+      "loss": 0.8726,
+      "step": 31941
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5212914285714286e-06,
+      "loss": 1.3124,
+      "step": 31942
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5210228571428576e-06,
+      "loss": 0.9389,
+      "step": 31943
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.520754285714286e-06,
+      "loss": 0.9054,
+      "step": 31944
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.520485714285714e-06,
+      "loss": 1.06,
+      "step": 31945
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.520217142857143e-06,
+      "loss": 0.7795,
+      "step": 31946
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.519948571428571e-06,
+      "loss": 1.6076,
+      "step": 31947
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.51968e-06,
+      "loss": 1.2623,
+      "step": 31948
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.519411428571428e-06,
+      "loss": 1.3178,
+      "step": 31949
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.519142857142857e-06,
+      "loss": 1.0832,
+      "step": 31950
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.518874285714286e-06,
+      "loss": 0.9848,
+      "step": 31951
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5186057142857145e-06,
+      "loss": 1.4403,
+      "step": 31952
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5183371428571436e-06,
+      "loss": 1.1065,
+      "step": 31953
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.518068571428572e-06,
+      "loss": 1.2143,
+      "step": 31954
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5178e-06,
+      "loss": 1.296,
+      "step": 31955
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.517531428571429e-06,
+      "loss": 1.3761,
+      "step": 31956
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.517262857142857e-06,
+      "loss": 1.0521,
+      "step": 31957
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.516994285714286e-06,
+      "loss": 1.3128,
+      "step": 31958
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.516725714285714e-06,
+      "loss": 0.7969,
+      "step": 31959
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5164571428571425e-06,
+      "loss": 1.3673,
+      "step": 31960
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5161885714285715e-06,
+      "loss": 0.965,
+      "step": 31961
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.51592e-06,
+      "loss": 1.1215,
+      "step": 31962
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.515651428571429e-06,
+      "loss": 1.2456,
+      "step": 31963
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.515382857142858e-06,
+      "loss": 1.3421,
+      "step": 31964
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.515114285714286e-06,
+      "loss": 0.8505,
+      "step": 31965
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.514845714285715e-06,
+      "loss": 1.2823,
+      "step": 31966
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.514577142857143e-06,
+      "loss": 0.8845,
+      "step": 31967
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.514308571428571e-06,
+      "loss": 1.0464,
+      "step": 31968
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.51404e-06,
+      "loss": 0.9808,
+      "step": 31969
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5137714285714285e-06,
+      "loss": 1.4407,
+      "step": 31970
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5135028571428575e-06,
+      "loss": 1.1401,
+      "step": 31971
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.513234285714286e-06,
+      "loss": 0.9519,
+      "step": 31972
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.512965714285714e-06,
+      "loss": 1.3388,
+      "step": 31973
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.512697142857143e-06,
+      "loss": 1.2458,
+      "step": 31974
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.512428571428571e-06,
+      "loss": 1.2987,
+      "step": 31975
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.51216e-06,
+      "loss": 1.1319,
+      "step": 31976
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.511891428571429e-06,
+      "loss": 0.9081,
+      "step": 31977
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.511622857142857e-06,
+      "loss": 1.5837,
+      "step": 31978
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.511354285714286e-06,
+      "loss": 0.9341,
+      "step": 31979
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5110857142857145e-06,
+      "loss": 1.2213,
+      "step": 31980
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5108171428571435e-06,
+      "loss": 0.9365,
+      "step": 31981
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.510548571428572e-06,
+      "loss": 1.0481,
+      "step": 31982
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.51028e-06,
+      "loss": 0.9504,
+      "step": 31983
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.510011428571429e-06,
+      "loss": 1.1603,
+      "step": 31984
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.509742857142857e-06,
+      "loss": 1.1195,
+      "step": 31985
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.509474285714286e-06,
+      "loss": 0.8811,
+      "step": 31986
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.509205714285714e-06,
+      "loss": 1.25,
+      "step": 31987
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5089371428571424e-06,
+      "loss": 0.6956,
+      "step": 31988
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5086685714285715e-06,
+      "loss": 1.0169,
+      "step": 31989
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5084000000000005e-06,
+      "loss": 1.0906,
+      "step": 31990
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.508131428571429e-06,
+      "loss": 1.2906,
+      "step": 31991
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.507862857142858e-06,
+      "loss": 1.2357,
+      "step": 31992
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.507594285714286e-06,
+      "loss": 0.9292,
+      "step": 31993
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.507325714285715e-06,
+      "loss": 1.0744,
+      "step": 31994
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.507057142857143e-06,
+      "loss": 1.1743,
+      "step": 31995
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.506788571428571e-06,
+      "loss": 1.5003,
+      "step": 31996
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.50652e-06,
+      "loss": 0.9224,
+      "step": 31997
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5062514285714284e-06,
+      "loss": 1.5093,
+      "step": 31998
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5059828571428575e-06,
+      "loss": 0.731,
+      "step": 31999
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.505714285714286e-06,
+      "loss": 1.6624,
+      "step": 32000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.505445714285714e-06,
+      "loss": 1.1277,
+      "step": 32001
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.505177142857143e-06,
+      "loss": 1.116,
+      "step": 32002
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.504908571428572e-06,
+      "loss": 0.9591,
+      "step": 32003
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.504640000000001e-06,
+      "loss": 0.8109,
+      "step": 32004
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.504371428571429e-06,
+      "loss": 1.292,
+      "step": 32005
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.504102857142857e-06,
+      "loss": 1.2517,
+      "step": 32006
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.503834285714286e-06,
+      "loss": 1.6896,
+      "step": 32007
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5035657142857144e-06,
+      "loss": 1.1862,
+      "step": 32008
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5032971428571435e-06,
+      "loss": 1.2923,
+      "step": 32009
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.503028571428572e-06,
+      "loss": 1.3707,
+      "step": 32010
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.50276e-06,
+      "loss": 0.9525,
+      "step": 32011
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.502491428571429e-06,
+      "loss": 1.2031,
+      "step": 32012
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.502222857142857e-06,
+      "loss": 1.1071,
+      "step": 32013
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.501954285714285e-06,
+      "loss": 0.9408,
+      "step": 32014
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.501685714285714e-06,
+      "loss": 1.1926,
+      "step": 32015
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.501417142857143e-06,
+      "loss": 1.466,
+      "step": 32016
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.501148571428572e-06,
+      "loss": 1.1846,
+      "step": 32017
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.5008800000000004e-06,
+      "loss": 1.0301,
+      "step": 32018
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.500611428571429e-06,
+      "loss": 1.3535,
+      "step": 32019
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.500342857142858e-06,
+      "loss": 0.8789,
+      "step": 32020
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.500074285714286e-06,
+      "loss": 1.1685,
+      "step": 32021
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.499805714285715e-06,
+      "loss": 1.0842,
+      "step": 32022
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.499537142857143e-06,
+      "loss": 1.4656,
+      "step": 32023
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.499268571428571e-06,
+      "loss": 0.7211,
+      "step": 32024
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.499e-06,
+      "loss": 1.283,
+      "step": 32025
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.498731428571428e-06,
+      "loss": 1.0293,
+      "step": 32026
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.498462857142857e-06,
+      "loss": 1.4298,
+      "step": 32027
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.498194285714286e-06,
+      "loss": 1.135,
+      "step": 32028
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.497925714285715e-06,
+      "loss": 1.0754,
+      "step": 32029
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.497657142857144e-06,
+      "loss": 1.3884,
+      "step": 32030
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.497388571428572e-06,
+      "loss": 0.9609,
+      "step": 32031
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.49712e-06,
+      "loss": 0.9754,
+      "step": 32032
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.496851428571429e-06,
+      "loss": 0.8976,
+      "step": 32033
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.496582857142857e-06,
+      "loss": 1.1237,
+      "step": 32034
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.496314285714286e-06,
+      "loss": 1.1732,
+      "step": 32035
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.496045714285714e-06,
+      "loss": 1.0285,
+      "step": 32036
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4957771428571426e-06,
+      "loss": 0.9796,
+      "step": 32037
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.495508571428572e-06,
+      "loss": 1.3751,
+      "step": 32038
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.49524e-06,
+      "loss": 1.411,
+      "step": 32039
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.494971428571429e-06,
+      "loss": 1.1373,
+      "step": 32040
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.494702857142857e-06,
+      "loss": 1.1107,
+      "step": 32041
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.494434285714286e-06,
+      "loss": 1.302,
+      "step": 32042
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.494165714285715e-06,
+      "loss": 1.2173,
+      "step": 32043
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.493897142857143e-06,
+      "loss": 1.0206,
+      "step": 32044
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.493628571428572e-06,
+      "loss": 0.9312,
+      "step": 32045
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.49336e-06,
+      "loss": 1.2558,
+      "step": 32046
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4930914285714286e-06,
+      "loss": 0.8094,
+      "step": 32047
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4928228571428576e-06,
+      "loss": 0.9767,
+      "step": 32048
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.492554285714286e-06,
+      "loss": 1.0278,
+      "step": 32049
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.492285714285715e-06,
+      "loss": 1.151,
+      "step": 32050
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.492017142857143e-06,
+      "loss": 1.2109,
+      "step": 32051
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.491748571428571e-06,
+      "loss": 1.0566,
+      "step": 32052
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.49148e-06,
+      "loss": 1.0596,
+      "step": 32053
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.491211428571428e-06,
+      "loss": 1.4502,
+      "step": 32054
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.490942857142857e-06,
+      "loss": 1.0484,
+      "step": 32055
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.490674285714286e-06,
+      "loss": 0.9906,
+      "step": 32056
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4904057142857145e-06,
+      "loss": 0.8696,
+      "step": 32057
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4901371428571436e-06,
+      "loss": 0.9406,
+      "step": 32058
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.489868571428572e-06,
+      "loss": 1.3586,
+      "step": 32059
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4896e-06,
+      "loss": 1.1099,
+      "step": 32060
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.489331428571429e-06,
+      "loss": 0.9524,
+      "step": 32061
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.489062857142857e-06,
+      "loss": 1.3334,
+      "step": 32062
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.488794285714286e-06,
+      "loss": 1.2553,
+      "step": 32063
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.488525714285714e-06,
+      "loss": 1.3179,
+      "step": 32064
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4882571428571425e-06,
+      "loss": 1.2084,
+      "step": 32065
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4879885714285715e-06,
+      "loss": 1.0668,
+      "step": 32066
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.48772e-06,
+      "loss": 0.9904,
+      "step": 32067
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.487451428571429e-06,
+      "loss": 0.9747,
+      "step": 32068
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.487182857142858e-06,
+      "loss": 0.9073,
+      "step": 32069
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.486914285714286e-06,
+      "loss": 1.0711,
+      "step": 32070
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.486645714285715e-06,
+      "loss": 1.1323,
+      "step": 32071
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.486377142857143e-06,
+      "loss": 1.4465,
+      "step": 32072
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.486108571428572e-06,
+      "loss": 1.2503,
+      "step": 32073
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.48584e-06,
+      "loss": 1.4349,
+      "step": 32074
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4855714285714285e-06,
+      "loss": 1.2042,
+      "step": 32075
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4853028571428575e-06,
+      "loss": 0.9922,
+      "step": 32076
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.485034285714286e-06,
+      "loss": 1.3541,
+      "step": 32077
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.484765714285714e-06,
+      "loss": 1.2301,
+      "step": 32078
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.484497142857143e-06,
+      "loss": 0.8364,
+      "step": 32079
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.484228571428571e-06,
+      "loss": 1.0331,
+      "step": 32080
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.48396e-06,
+      "loss": 0.7841,
+      "step": 32081
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.483691428571429e-06,
+      "loss": 1.0926,
+      "step": 32082
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.483422857142857e-06,
+      "loss": 1.0176,
+      "step": 32083
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.483154285714286e-06,
+      "loss": 1.0548,
+      "step": 32084
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4828857142857145e-06,
+      "loss": 0.74,
+      "step": 32085
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4826171428571435e-06,
+      "loss": 1.0925,
+      "step": 32086
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.482348571428572e-06,
+      "loss": 1.3069,
+      "step": 32087
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.48208e-06,
+      "loss": 1.6864,
+      "step": 32088
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.481811428571429e-06,
+      "loss": 1.4704,
+      "step": 32089
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.481542857142857e-06,
+      "loss": 1.5283,
+      "step": 32090
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.481274285714286e-06,
+      "loss": 1.482,
+      "step": 32091
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.481005714285714e-06,
+      "loss": 0.8519,
+      "step": 32092
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4807371428571424e-06,
+      "loss": 1.0839,
+      "step": 32093
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4804685714285715e-06,
+      "loss": 1.1476,
+      "step": 32094
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4802000000000005e-06,
+      "loss": 1.1926,
+      "step": 32095
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4799314285714295e-06,
+      "loss": 0.8498,
+      "step": 32096
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.479662857142858e-06,
+      "loss": 1.288,
+      "step": 32097
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.479394285714286e-06,
+      "loss": 0.9402,
+      "step": 32098
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.479125714285715e-06,
+      "loss": 1.132,
+      "step": 32099
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.478857142857143e-06,
+      "loss": 1.1381,
+      "step": 32100
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.478588571428571e-06,
+      "loss": 1.2018,
+      "step": 32101
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.47832e-06,
+      "loss": 1.4865,
+      "step": 32102
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4780514285714284e-06,
+      "loss": 0.7491,
+      "step": 32103
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.4777828571428575e-06,
+      "loss": 0.8038,
+      "step": 32104
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.477514285714286e-06,
+      "loss": 0.9934,
+      "step": 32105
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.477245714285714e-06,
+      "loss": 1.2373,
+      "step": 32106
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.476977142857143e-06,
+      "loss": 1.3649,
+      "step": 32107
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.476708571428572e-06,
+      "loss": 1.1156,
+      "step": 32108
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.476440000000001e-06,
+      "loss": 1.2921,
+      "step": 32109
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.476171428571429e-06,
+      "loss": 1.448,
+      "step": 32110
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.475902857142857e-06,
+      "loss": 1.0925,
+      "step": 32111
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 5.475634285714286e-06,
+      "loss": 1.1533,
+      "step": 32112
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4753657142857144e-06,
+      "loss": 1.1803,
+      "step": 32113
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4750971428571435e-06,
+      "loss": 1.6632,
+      "step": 32114
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.474828571428572e-06,
+      "loss": 1.0136,
+      "step": 32115
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.47456e-06,
+      "loss": 1.4673,
+      "step": 32116
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.474291428571429e-06,
+      "loss": 0.7002,
+      "step": 32117
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.474022857142857e-06,
+      "loss": 0.9007,
+      "step": 32118
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.473754285714285e-06,
+      "loss": 1.0488,
+      "step": 32119
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.473485714285714e-06,
+      "loss": 0.947,
+      "step": 32120
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.473217142857143e-06,
+      "loss": 1.0301,
+      "step": 32121
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.472948571428572e-06,
+      "loss": 1.4006,
+      "step": 32122
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4726800000000004e-06,
+      "loss": 1.1313,
+      "step": 32123
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.472411428571429e-06,
+      "loss": 1.0565,
+      "step": 32124
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.472142857142858e-06,
+      "loss": 1.4118,
+      "step": 32125
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.471874285714286e-06,
+      "loss": 1.1973,
+      "step": 32126
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.471605714285715e-06,
+      "loss": 1.1226,
+      "step": 32127
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.471337142857143e-06,
+      "loss": 1.2344,
+      "step": 32128
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.471068571428571e-06,
+      "loss": 1.1207,
+      "step": 32129
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4708e-06,
+      "loss": 1.1439,
+      "step": 32130
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.470531428571428e-06,
+      "loss": 1.5427,
+      "step": 32131
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.470262857142857e-06,
+      "loss": 1.0329,
+      "step": 32132
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.469994285714286e-06,
+      "loss": 1.3032,
+      "step": 32133
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.469725714285715e-06,
+      "loss": 1.0549,
+      "step": 32134
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.469457142857144e-06,
+      "loss": 0.7614,
+      "step": 32135
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.469188571428572e-06,
+      "loss": 0.9443,
+      "step": 32136
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.468920000000001e-06,
+      "loss": 1.2008,
+      "step": 32137
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.468651428571429e-06,
+      "loss": 1.0975,
+      "step": 32138
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.468382857142857e-06,
+      "loss": 1.2,
+      "step": 32139
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.468114285714286e-06,
+      "loss": 1.0523,
+      "step": 32140
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.467845714285714e-06,
+      "loss": 1.0753,
+      "step": 32141
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4675771428571426e-06,
+      "loss": 1.2672,
+      "step": 32142
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4673085714285716e-06,
+      "loss": 1.2938,
+      "step": 32143
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.46704e-06,
+      "loss": 1.084,
+      "step": 32144
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.466771428571429e-06,
+      "loss": 0.8081,
+      "step": 32145
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.466502857142857e-06,
+      "loss": 1.2911,
+      "step": 32146
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.466234285714286e-06,
+      "loss": 0.9745,
+      "step": 32147
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.465965714285715e-06,
+      "loss": 1.3687,
+      "step": 32148
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.465697142857143e-06,
+      "loss": 1.2362,
+      "step": 32149
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.465428571428572e-06,
+      "loss": 0.819,
+      "step": 32150
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.46516e-06,
+      "loss": 1.5628,
+      "step": 32151
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4648914285714285e-06,
+      "loss": 1.0407,
+      "step": 32152
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4646228571428576e-06,
+      "loss": 1.4225,
+      "step": 32153
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.464354285714286e-06,
+      "loss": 1.296,
+      "step": 32154
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.464085714285715e-06,
+      "loss": 0.8643,
+      "step": 32155
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.463817142857143e-06,
+      "loss": 1.0843,
+      "step": 32156
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.463548571428571e-06,
+      "loss": 1.3843,
+      "step": 32157
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.46328e-06,
+      "loss": 1.1845,
+      "step": 32158
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.463011428571428e-06,
+      "loss": 1.2474,
+      "step": 32159
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.462742857142857e-06,
+      "loss": 1.0052,
+      "step": 32160
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.462474285714286e-06,
+      "loss": 1.2747,
+      "step": 32161
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4622057142857145e-06,
+      "loss": 1.3918,
+      "step": 32162
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4619371428571436e-06,
+      "loss": 0.9098,
+      "step": 32163
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.461668571428572e-06,
+      "loss": 1.201,
+      "step": 32164
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4614e-06,
+      "loss": 1.0474,
+      "step": 32165
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.461131428571429e-06,
+      "loss": 1.0244,
+      "step": 32166
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.460862857142857e-06,
+      "loss": 0.9779,
+      "step": 32167
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.460594285714286e-06,
+      "loss": 1.6359,
+      "step": 32168
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.460325714285714e-06,
+      "loss": 1.7469,
+      "step": 32169
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4600571428571425e-06,
+      "loss": 1.1183,
+      "step": 32170
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4597885714285715e-06,
+      "loss": 1.0524,
+      "step": 32171
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.45952e-06,
+      "loss": 1.2039,
+      "step": 32172
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.459251428571429e-06,
+      "loss": 1.5333,
+      "step": 32173
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.458982857142858e-06,
+      "loss": 0.6045,
+      "step": 32174
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.458714285714286e-06,
+      "loss": 1.0946,
+      "step": 32175
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.458445714285715e-06,
+      "loss": 1.2577,
+      "step": 32176
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.458177142857143e-06,
+      "loss": 1.0465,
+      "step": 32177
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.457908571428572e-06,
+      "loss": 1.1837,
+      "step": 32178
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.45764e-06,
+      "loss": 1.056,
+      "step": 32179
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4573714285714285e-06,
+      "loss": 0.9535,
+      "step": 32180
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4571028571428575e-06,
+      "loss": 1.1008,
+      "step": 32181
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.456834285714286e-06,
+      "loss": 1.109,
+      "step": 32182
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.456565714285715e-06,
+      "loss": 1.2543,
+      "step": 32183
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.456297142857143e-06,
+      "loss": 1.595,
+      "step": 32184
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.456028571428571e-06,
+      "loss": 0.993,
+      "step": 32185
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.45576e-06,
+      "loss": 1.3137,
+      "step": 32186
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.455491428571429e-06,
+      "loss": 1.106,
+      "step": 32187
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.455222857142857e-06,
+      "loss": 1.1141,
+      "step": 32188
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.454954285714286e-06,
+      "loss": 0.5752,
+      "step": 32189
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4546857142857145e-06,
+      "loss": 1.1414,
+      "step": 32190
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4544171428571435e-06,
+      "loss": 1.2462,
+      "step": 32191
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.454148571428572e-06,
+      "loss": 0.7299,
+      "step": 32192
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.45388e-06,
+      "loss": 1.171,
+      "step": 32193
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.453611428571429e-06,
+      "loss": 1.2743,
+      "step": 32194
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.453342857142857e-06,
+      "loss": 1.2759,
+      "step": 32195
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.453074285714286e-06,
+      "loss": 1.3805,
+      "step": 32196
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.452805714285714e-06,
+      "loss": 1.1452,
+      "step": 32197
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4525371428571424e-06,
+      "loss": 1.0672,
+      "step": 32198
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4522685714285715e-06,
+      "loss": 0.888,
+      "step": 32199
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4520000000000005e-06,
+      "loss": 1.3349,
+      "step": 32200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4517314285714295e-06,
+      "loss": 0.9729,
+      "step": 32201
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.451462857142858e-06,
+      "loss": 1.1592,
+      "step": 32202
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.451194285714286e-06,
+      "loss": 1.7186,
+      "step": 32203
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.450925714285715e-06,
+      "loss": 0.9242,
+      "step": 32204
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.450657142857143e-06,
+      "loss": 1.1566,
+      "step": 32205
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.450388571428571e-06,
+      "loss": 1.0158,
+      "step": 32206
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.45012e-06,
+      "loss": 0.8725,
+      "step": 32207
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4498514285714284e-06,
+      "loss": 1.543,
+      "step": 32208
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4495828571428575e-06,
+      "loss": 1.4123,
+      "step": 32209
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.449314285714286e-06,
+      "loss": 0.7589,
+      "step": 32210
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.449045714285714e-06,
+      "loss": 1.0176,
+      "step": 32211
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.448777142857143e-06,
+      "loss": 0.9824,
+      "step": 32212
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.448508571428572e-06,
+      "loss": 1.2587,
+      "step": 32213
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.448240000000001e-06,
+      "loss": 1.4841,
+      "step": 32214
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.447971428571429e-06,
+      "loss": 1.2199,
+      "step": 32215
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.447702857142857e-06,
+      "loss": 1.1943,
+      "step": 32216
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.447434285714286e-06,
+      "loss": 1.2263,
+      "step": 32217
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4471657142857144e-06,
+      "loss": 1.1222,
+      "step": 32218
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4468971428571435e-06,
+      "loss": 1.1832,
+      "step": 32219
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.446628571428572e-06,
+      "loss": 1.2221,
+      "step": 32220
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.44636e-06,
+      "loss": 1.2098,
+      "step": 32221
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.446091428571429e-06,
+      "loss": 1.3498,
+      "step": 32222
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.445822857142857e-06,
+      "loss": 1.4482,
+      "step": 32223
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.445554285714286e-06,
+      "loss": 1.0105,
+      "step": 32224
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.445285714285714e-06,
+      "loss": 1.041,
+      "step": 32225
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.445017142857143e-06,
+      "loss": 0.9299,
+      "step": 32226
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.444748571428572e-06,
+      "loss": 1.2714,
+      "step": 32227
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4444800000000004e-06,
+      "loss": 0.912,
+      "step": 32228
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.444211428571429e-06,
+      "loss": 1.2509,
+      "step": 32229
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.443942857142858e-06,
+      "loss": 1.0486,
+      "step": 32230
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.443674285714286e-06,
+      "loss": 0.9419,
+      "step": 32231
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.443405714285715e-06,
+      "loss": 1.2594,
+      "step": 32232
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.443137142857143e-06,
+      "loss": 1.4541,
+      "step": 32233
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.442868571428571e-06,
+      "loss": 1.1296,
+      "step": 32234
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4426e-06,
+      "loss": 1.221,
+      "step": 32235
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.442331428571428e-06,
+      "loss": 1.4608,
+      "step": 32236
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.442062857142857e-06,
+      "loss": 1.0043,
+      "step": 32237
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.441794285714286e-06,
+      "loss": 1.4426,
+      "step": 32238
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.441525714285715e-06,
+      "loss": 1.0877,
+      "step": 32239
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.441257142857144e-06,
+      "loss": 0.7293,
+      "step": 32240
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.440988571428572e-06,
+      "loss": 0.8437,
+      "step": 32241
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.440720000000001e-06,
+      "loss": 1.3095,
+      "step": 32242
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.440451428571429e-06,
+      "loss": 0.9264,
+      "step": 32243
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.440182857142857e-06,
+      "loss": 1.1549,
+      "step": 32244
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.439914285714286e-06,
+      "loss": 1.1831,
+      "step": 32245
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.439645714285714e-06,
+      "loss": 1.2468,
+      "step": 32246
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.439377142857143e-06,
+      "loss": 1.1541,
+      "step": 32247
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4391085714285716e-06,
+      "loss": 1.2798,
+      "step": 32248
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.43884e-06,
+      "loss": 1.0521,
+      "step": 32249
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.438571428571429e-06,
+      "loss": 1.0892,
+      "step": 32250
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.438302857142857e-06,
+      "loss": 1.0894,
+      "step": 32251
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.438034285714286e-06,
+      "loss": 1.515,
+      "step": 32252
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.437765714285715e-06,
+      "loss": 1.1423,
+      "step": 32253
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.437497142857143e-06,
+      "loss": 0.8449,
+      "step": 32254
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.437228571428572e-06,
+      "loss": 1.3358,
+      "step": 32255
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.43696e-06,
+      "loss": 0.9457,
+      "step": 32256
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4366914285714285e-06,
+      "loss": 1.1144,
+      "step": 32257
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4364228571428576e-06,
+      "loss": 1.0894,
+      "step": 32258
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.436154285714286e-06,
+      "loss": 0.9891,
+      "step": 32259
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.435885714285715e-06,
+      "loss": 1.0778,
+      "step": 32260
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.435617142857143e-06,
+      "loss": 1.2768,
+      "step": 32261
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.435348571428571e-06,
+      "loss": 0.804,
+      "step": 32262
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.43508e-06,
+      "loss": 1.1829,
+      "step": 32263
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.434811428571428e-06,
+      "loss": 1.3234,
+      "step": 32264
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.434542857142857e-06,
+      "loss": 1.1917,
+      "step": 32265
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.434274285714286e-06,
+      "loss": 1.1926,
+      "step": 32266
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4340057142857145e-06,
+      "loss": 1.4347,
+      "step": 32267
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4337371428571436e-06,
+      "loss": 1.3219,
+      "step": 32268
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.433468571428572e-06,
+      "loss": 1.5143,
+      "step": 32269
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.433200000000001e-06,
+      "loss": 1.3387,
+      "step": 32270
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.432931428571429e-06,
+      "loss": 1.318,
+      "step": 32271
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.432662857142857e-06,
+      "loss": 1.4385,
+      "step": 32272
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.432394285714286e-06,
+      "loss": 1.3976,
+      "step": 32273
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.432125714285714e-06,
+      "loss": 1.0254,
+      "step": 32274
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4318571428571425e-06,
+      "loss": 0.7072,
+      "step": 32275
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4315885714285715e-06,
+      "loss": 1.2761,
+      "step": 32276
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.43132e-06,
+      "loss": 0.8811,
+      "step": 32277
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.431051428571429e-06,
+      "loss": 0.9999,
+      "step": 32278
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.430782857142858e-06,
+      "loss": 1.3444,
+      "step": 32279
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.430514285714286e-06,
+      "loss": 0.9603,
+      "step": 32280
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.430245714285715e-06,
+      "loss": 1.2484,
+      "step": 32281
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.429977142857143e-06,
+      "loss": 1.5,
+      "step": 32282
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.429708571428572e-06,
+      "loss": 0.8896,
+      "step": 32283
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.42944e-06,
+      "loss": 0.7687,
+      "step": 32284
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4291714285714285e-06,
+      "loss": 0.5674,
+      "step": 32285
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.4289028571428575e-06,
+      "loss": 0.918,
+      "step": 32286
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 5.428634285714286e-06,
+      "loss": 1.4222,
+      "step": 32287
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.428365714285715e-06,
+      "loss": 1.0424,
+      "step": 32288
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.428097142857143e-06,
+      "loss": 1.168,
+      "step": 32289
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.427828571428571e-06,
+      "loss": 0.9296,
+      "step": 32290
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.42756e-06,
+      "loss": 1.3147,
+      "step": 32291
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.427291428571429e-06,
+      "loss": 1.4859,
+      "step": 32292
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.427022857142857e-06,
+      "loss": 0.8189,
+      "step": 32293
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.426754285714286e-06,
+      "loss": 1.0785,
+      "step": 32294
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4264857142857145e-06,
+      "loss": 1.3219,
+      "step": 32295
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4262171428571435e-06,
+      "loss": 0.9926,
+      "step": 32296
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.425948571428572e-06,
+      "loss": 1.0827,
+      "step": 32297
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.42568e-06,
+      "loss": 0.9176,
+      "step": 32298
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.425411428571429e-06,
+      "loss": 1.0452,
+      "step": 32299
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.425142857142857e-06,
+      "loss": 1.035,
+      "step": 32300
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.424874285714286e-06,
+      "loss": 1.0462,
+      "step": 32301
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.424605714285714e-06,
+      "loss": 1.0641,
+      "step": 32302
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4243371428571424e-06,
+      "loss": 1.0507,
+      "step": 32303
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4240685714285715e-06,
+      "loss": 1.2709,
+      "step": 32304
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4238000000000005e-06,
+      "loss": 1.5624,
+      "step": 32305
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4235314285714295e-06,
+      "loss": 1.1684,
+      "step": 32306
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.423262857142858e-06,
+      "loss": 1.092,
+      "step": 32307
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.422994285714286e-06,
+      "loss": 1.4891,
+      "step": 32308
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.422725714285715e-06,
+      "loss": 1.4978,
+      "step": 32309
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.422457142857143e-06,
+      "loss": 1.3885,
+      "step": 32310
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.422188571428572e-06,
+      "loss": 1.2279,
+      "step": 32311
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.42192e-06,
+      "loss": 1.1726,
+      "step": 32312
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4216514285714284e-06,
+      "loss": 1.2006,
+      "step": 32313
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4213828571428575e-06,
+      "loss": 0.9324,
+      "step": 32314
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.421114285714286e-06,
+      "loss": 1.0664,
+      "step": 32315
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.420845714285714e-06,
+      "loss": 0.9477,
+      "step": 32316
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.420577142857143e-06,
+      "loss": 0.7864,
+      "step": 32317
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.420308571428571e-06,
+      "loss": 0.8166,
+      "step": 32318
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.420040000000001e-06,
+      "loss": 1.0695,
+      "step": 32319
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.419771428571429e-06,
+      "loss": 1.0119,
+      "step": 32320
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.419502857142857e-06,
+      "loss": 1.2835,
+      "step": 32321
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.419234285714286e-06,
+      "loss": 1.1108,
+      "step": 32322
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4189657142857144e-06,
+      "loss": 1.5484,
+      "step": 32323
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4186971428571435e-06,
+      "loss": 1.0298,
+      "step": 32324
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.418428571428572e-06,
+      "loss": 1.0174,
+      "step": 32325
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.41816e-06,
+      "loss": 1.2861,
+      "step": 32326
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.417891428571429e-06,
+      "loss": 1.4091,
+      "step": 32327
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.417622857142857e-06,
+      "loss": 0.8312,
+      "step": 32328
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.417354285714286e-06,
+      "loss": 1.184,
+      "step": 32329
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.417085714285714e-06,
+      "loss": 1.4447,
+      "step": 32330
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.416817142857142e-06,
+      "loss": 1.549,
+      "step": 32331
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.416548571428572e-06,
+      "loss": 1.0734,
+      "step": 32332
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4162800000000004e-06,
+      "loss": 1.3227,
+      "step": 32333
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4160114285714295e-06,
+      "loss": 0.9814,
+      "step": 32334
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.415742857142858e-06,
+      "loss": 0.9971,
+      "step": 32335
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.415474285714286e-06,
+      "loss": 1.1108,
+      "step": 32336
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.415205714285715e-06,
+      "loss": 1.2401,
+      "step": 32337
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.414937142857143e-06,
+      "loss": 1.2529,
+      "step": 32338
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.414668571428571e-06,
+      "loss": 1.4815,
+      "step": 32339
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4144e-06,
+      "loss": 0.9615,
+      "step": 32340
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.414131428571428e-06,
+      "loss": 0.6828,
+      "step": 32341
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.413862857142857e-06,
+      "loss": 1.1934,
+      "step": 32342
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.413594285714286e-06,
+      "loss": 1.2363,
+      "step": 32343
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.413325714285714e-06,
+      "loss": 1.1742,
+      "step": 32344
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.413057142857144e-06,
+      "loss": 1.1868,
+      "step": 32345
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.412788571428572e-06,
+      "loss": 0.8353,
+      "step": 32346
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.412520000000001e-06,
+      "loss": 1.0808,
+      "step": 32347
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.412251428571429e-06,
+      "loss": 1.2802,
+      "step": 32348
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.411982857142857e-06,
+      "loss": 0.9114,
+      "step": 32349
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.411714285714286e-06,
+      "loss": 1.0341,
+      "step": 32350
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.411445714285714e-06,
+      "loss": 0.8234,
+      "step": 32351
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.411177142857143e-06,
+      "loss": 1.141,
+      "step": 32352
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4109085714285716e-06,
+      "loss": 1.3919,
+      "step": 32353
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.41064e-06,
+      "loss": 1.1869,
+      "step": 32354
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.410371428571429e-06,
+      "loss": 1.0136,
+      "step": 32355
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.410102857142857e-06,
+      "loss": 0.9844,
+      "step": 32356
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.409834285714285e-06,
+      "loss": 1.1596,
+      "step": 32357
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.409565714285715e-06,
+      "loss": 1.5164,
+      "step": 32358
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.409297142857143e-06,
+      "loss": 1.2531,
+      "step": 32359
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.409028571428572e-06,
+      "loss": 0.7095,
+      "step": 32360
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.40876e-06,
+      "loss": 1.0412,
+      "step": 32361
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4084914285714285e-06,
+      "loss": 1.1287,
+      "step": 32362
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4082228571428576e-06,
+      "loss": 1.058,
+      "step": 32363
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.407954285714286e-06,
+      "loss": 1.3794,
+      "step": 32364
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.407685714285715e-06,
+      "loss": 0.9181,
+      "step": 32365
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.407417142857143e-06,
+      "loss": 1.1508,
+      "step": 32366
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.407148571428571e-06,
+      "loss": 0.9997,
+      "step": 32367
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.40688e-06,
+      "loss": 1.3516,
+      "step": 32368
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.406611428571428e-06,
+      "loss": 1.0325,
+      "step": 32369
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.406342857142857e-06,
+      "loss": 1.1794,
+      "step": 32370
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4060742857142855e-06,
+      "loss": 1.2379,
+      "step": 32371
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4058057142857145e-06,
+      "loss": 0.9789,
+      "step": 32372
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4055371428571436e-06,
+      "loss": 1.1401,
+      "step": 32373
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.405268571428572e-06,
+      "loss": 1.6704,
+      "step": 32374
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.405000000000001e-06,
+      "loss": 1.2418,
+      "step": 32375
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.404731428571429e-06,
+      "loss": 0.899,
+      "step": 32376
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.404462857142857e-06,
+      "loss": 1.0836,
+      "step": 32377
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.404194285714286e-06,
+      "loss": 1.2389,
+      "step": 32378
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.403925714285714e-06,
+      "loss": 1.5226,
+      "step": 32379
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4036571428571425e-06,
+      "loss": 1.3234,
+      "step": 32380
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4033885714285715e-06,
+      "loss": 0.4873,
+      "step": 32381
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.40312e-06,
+      "loss": 1.1269,
+      "step": 32382
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.402851428571429e-06,
+      "loss": 1.3298,
+      "step": 32383
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.402582857142857e-06,
+      "loss": 1.3471,
+      "step": 32384
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.402314285714286e-06,
+      "loss": 1.2093,
+      "step": 32385
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.402045714285715e-06,
+      "loss": 1.0295,
+      "step": 32386
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.401777142857143e-06,
+      "loss": 1.2483,
+      "step": 32387
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.401508571428572e-06,
+      "loss": 1.153,
+      "step": 32388
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.40124e-06,
+      "loss": 1.1387,
+      "step": 32389
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4009714285714285e-06,
+      "loss": 0.6769,
+      "step": 32390
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.4007028571428575e-06,
+      "loss": 1.2945,
+      "step": 32391
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.400434285714286e-06,
+      "loss": 1.0817,
+      "step": 32392
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.400165714285715e-06,
+      "loss": 1.2822,
+      "step": 32393
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.399897142857143e-06,
+      "loss": 0.6399,
+      "step": 32394
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.399628571428571e-06,
+      "loss": 1.3187,
+      "step": 32395
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.39936e-06,
+      "loss": 1.2576,
+      "step": 32396
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.399091428571428e-06,
+      "loss": 1.3698,
+      "step": 32397
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.398822857142858e-06,
+      "loss": 1.6157,
+      "step": 32398
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.398554285714286e-06,
+      "loss": 1.1042,
+      "step": 32399
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3982857142857145e-06,
+      "loss": 1.3932,
+      "step": 32400
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3980171428571435e-06,
+      "loss": 0.9742,
+      "step": 32401
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.397748571428572e-06,
+      "loss": 0.9996,
+      "step": 32402
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.39748e-06,
+      "loss": 0.8665,
+      "step": 32403
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.397211428571429e-06,
+      "loss": 0.5279,
+      "step": 32404
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.396942857142857e-06,
+      "loss": 1.2474,
+      "step": 32405
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.396674285714286e-06,
+      "loss": 1.4696,
+      "step": 32406
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.396405714285714e-06,
+      "loss": 1.0329,
+      "step": 32407
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3961371428571424e-06,
+      "loss": 0.6748,
+      "step": 32408
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3958685714285715e-06,
+      "loss": 1.1408,
+      "step": 32409
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3956e-06,
+      "loss": 1.1454,
+      "step": 32410
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3953314285714295e-06,
+      "loss": 0.9634,
+      "step": 32411
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.395062857142858e-06,
+      "loss": 1.1637,
+      "step": 32412
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.394794285714286e-06,
+      "loss": 0.8337,
+      "step": 32413
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.394525714285715e-06,
+      "loss": 1.3265,
+      "step": 32414
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.394257142857143e-06,
+      "loss": 1.0926,
+      "step": 32415
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.393988571428572e-06,
+      "loss": 0.7901,
+      "step": 32416
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.39372e-06,
+      "loss": 1.1293,
+      "step": 32417
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3934514285714284e-06,
+      "loss": 0.9941,
+      "step": 32418
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3931828571428575e-06,
+      "loss": 1.2399,
+      "step": 32419
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.392914285714286e-06,
+      "loss": 1.3662,
+      "step": 32420
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.392645714285715e-06,
+      "loss": 0.8181,
+      "step": 32421
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.392377142857143e-06,
+      "loss": 1.032,
+      "step": 32422
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.392108571428571e-06,
+      "loss": 0.7585,
+      "step": 32423
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.39184e-06,
+      "loss": 1.1975,
+      "step": 32424
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.391571428571429e-06,
+      "loss": 1.4264,
+      "step": 32425
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.391302857142857e-06,
+      "loss": 1.2513,
+      "step": 32426
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.391034285714286e-06,
+      "loss": 1.321,
+      "step": 32427
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3907657142857144e-06,
+      "loss": 1.3033,
+      "step": 32428
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3904971428571435e-06,
+      "loss": 1.2122,
+      "step": 32429
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.390228571428572e-06,
+      "loss": 1.1475,
+      "step": 32430
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.38996e-06,
+      "loss": 0.6456,
+      "step": 32431
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.389691428571429e-06,
+      "loss": 0.9694,
+      "step": 32432
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.389422857142857e-06,
+      "loss": 0.6947,
+      "step": 32433
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.389154285714286e-06,
+      "loss": 1.5376,
+      "step": 32434
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.388885714285714e-06,
+      "loss": 0.8936,
+      "step": 32435
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.388617142857142e-06,
+      "loss": 0.6325,
+      "step": 32436
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.388348571428571e-06,
+      "loss": 1.572,
+      "step": 32437
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3880800000000004e-06,
+      "loss": 1.1781,
+      "step": 32438
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3878114285714295e-06,
+      "loss": 1.1022,
+      "step": 32439
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.387542857142858e-06,
+      "loss": 1.5585,
+      "step": 32440
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.387274285714286e-06,
+      "loss": 1.3186,
+      "step": 32441
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.387005714285715e-06,
+      "loss": 1.1235,
+      "step": 32442
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.386737142857143e-06,
+      "loss": 1.1957,
+      "step": 32443
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.386468571428571e-06,
+      "loss": 1.3974,
+      "step": 32444
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3862e-06,
+      "loss": 0.8679,
+      "step": 32445
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.385931428571428e-06,
+      "loss": 1.0578,
+      "step": 32446
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.385662857142857e-06,
+      "loss": 1.0477,
+      "step": 32447
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.385394285714286e-06,
+      "loss": 1.4114,
+      "step": 32448
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.385125714285714e-06,
+      "loss": 1.6334,
+      "step": 32449
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.384857142857143e-06,
+      "loss": 0.6054,
+      "step": 32450
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.384588571428572e-06,
+      "loss": 0.9466,
+      "step": 32451
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.384320000000001e-06,
+      "loss": 1.3925,
+      "step": 32452
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.384051428571429e-06,
+      "loss": 1.6881,
+      "step": 32453
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.383782857142857e-06,
+      "loss": 1.04,
+      "step": 32454
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.383514285714286e-06,
+      "loss": 1.1597,
+      "step": 32455
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.383245714285714e-06,
+      "loss": 1.5199,
+      "step": 32456
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.382977142857143e-06,
+      "loss": 1.219,
+      "step": 32457
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.3827085714285716e-06,
+      "loss": 1.068,
+      "step": 32458
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.38244e-06,
+      "loss": 1.2686,
+      "step": 32459
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.382171428571429e-06,
+      "loss": 1.555,
+      "step": 32460
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.381902857142857e-06,
+      "loss": 0.7021,
+      "step": 32461
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 5.381634285714286e-06,
+      "loss": 0.9174,
+      "step": 32462
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.381365714285714e-06,
+      "loss": 1.3153,
+      "step": 32463
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.381097142857143e-06,
+      "loss": 1.1692,
+      "step": 32464
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.380828571428572e-06,
+      "loss": 1.5331,
+      "step": 32465
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.38056e-06,
+      "loss": 0.8683,
+      "step": 32466
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3802914285714285e-06,
+      "loss": 1.168,
+      "step": 32467
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3800228571428576e-06,
+      "loss": 1.413,
+      "step": 32468
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.379754285714286e-06,
+      "loss": 0.9526,
+      "step": 32469
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.379485714285715e-06,
+      "loss": 0.6099,
+      "step": 32470
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.379217142857143e-06,
+      "loss": 1.4135,
+      "step": 32471
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.378948571428571e-06,
+      "loss": 1.042,
+      "step": 32472
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.37868e-06,
+      "loss": 1.081,
+      "step": 32473
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.378411428571428e-06,
+      "loss": 0.8524,
+      "step": 32474
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.378142857142857e-06,
+      "loss": 0.7971,
+      "step": 32475
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3778742857142855e-06,
+      "loss": 1.0725,
+      "step": 32476
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3776057142857145e-06,
+      "loss": 1.1592,
+      "step": 32477
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3773371428571436e-06,
+      "loss": 1.3179,
+      "step": 32478
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.377068571428572e-06,
+      "loss": 1.1233,
+      "step": 32479
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.376800000000001e-06,
+      "loss": 1.1801,
+      "step": 32480
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.376531428571429e-06,
+      "loss": 0.907,
+      "step": 32481
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.376262857142857e-06,
+      "loss": 1.1632,
+      "step": 32482
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.375994285714286e-06,
+      "loss": 1.2579,
+      "step": 32483
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.375725714285714e-06,
+      "loss": 1.3039,
+      "step": 32484
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.375457142857143e-06,
+      "loss": 1.5091,
+      "step": 32485
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3751885714285715e-06,
+      "loss": 1.0257,
+      "step": 32486
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.37492e-06,
+      "loss": 1.4001,
+      "step": 32487
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.374651428571429e-06,
+      "loss": 0.8752,
+      "step": 32488
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.374382857142857e-06,
+      "loss": 1.4692,
+      "step": 32489
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.374114285714286e-06,
+      "loss": 0.8853,
+      "step": 32490
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.373845714285715e-06,
+      "loss": 0.7768,
+      "step": 32491
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.373577142857143e-06,
+      "loss": 0.967,
+      "step": 32492
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.373308571428572e-06,
+      "loss": 0.9007,
+      "step": 32493
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.37304e-06,
+      "loss": 0.9909,
+      "step": 32494
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3727714285714285e-06,
+      "loss": 1.3735,
+      "step": 32495
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3725028571428575e-06,
+      "loss": 1.0728,
+      "step": 32496
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.372234285714286e-06,
+      "loss": 1.1382,
+      "step": 32497
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.371965714285715e-06,
+      "loss": 0.9198,
+      "step": 32498
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.371697142857143e-06,
+      "loss": 0.9035,
+      "step": 32499
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.371428571428571e-06,
+      "loss": 1.036,
+      "step": 32500
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.37116e-06,
+      "loss": 1.4851,
+      "step": 32501
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.370891428571428e-06,
+      "loss": 1.1288,
+      "step": 32502
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.370622857142857e-06,
+      "loss": 0.9651,
+      "step": 32503
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.370354285714286e-06,
+      "loss": 1.4396,
+      "step": 32504
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3700857142857145e-06,
+      "loss": 0.9897,
+      "step": 32505
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3698171428571435e-06,
+      "loss": 0.7165,
+      "step": 32506
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.369548571428572e-06,
+      "loss": 1.038,
+      "step": 32507
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.369280000000001e-06,
+      "loss": 1.1782,
+      "step": 32508
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.369011428571429e-06,
+      "loss": 1.1178,
+      "step": 32509
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.368742857142857e-06,
+      "loss": 1.2711,
+      "step": 32510
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.368474285714286e-06,
+      "loss": 1.1805,
+      "step": 32511
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.368205714285714e-06,
+      "loss": 1.3071,
+      "step": 32512
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3679371428571424e-06,
+      "loss": 1.232,
+      "step": 32513
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3676685714285715e-06,
+      "loss": 1.215,
+      "step": 32514
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3674e-06,
+      "loss": 0.9455,
+      "step": 32515
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.367131428571429e-06,
+      "loss": 1.196,
+      "step": 32516
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.366862857142858e-06,
+      "loss": 0.8035,
+      "step": 32517
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.366594285714286e-06,
+      "loss": 0.8992,
+      "step": 32518
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.366325714285715e-06,
+      "loss": 1.1169,
+      "step": 32519
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.366057142857143e-06,
+      "loss": 1.0076,
+      "step": 32520
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.365788571428572e-06,
+      "loss": 1.0678,
+      "step": 32521
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.36552e-06,
+      "loss": 1.251,
+      "step": 32522
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3652514285714284e-06,
+      "loss": 1.0949,
+      "step": 32523
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3649828571428575e-06,
+      "loss": 1.4186,
+      "step": 32524
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.364714285714286e-06,
+      "loss": 1.2605,
+      "step": 32525
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.364445714285715e-06,
+      "loss": 1.2457,
+      "step": 32526
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.364177142857143e-06,
+      "loss": 1.5571,
+      "step": 32527
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.363908571428571e-06,
+      "loss": 1.0382,
+      "step": 32528
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.36364e-06,
+      "loss": 1.2891,
+      "step": 32529
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.363371428571429e-06,
+      "loss": 1.2077,
+      "step": 32530
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.363102857142857e-06,
+      "loss": 1.8567,
+      "step": 32531
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.362834285714286e-06,
+      "loss": 1.0162,
+      "step": 32532
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3625657142857144e-06,
+      "loss": 1.095,
+      "step": 32533
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3622971428571435e-06,
+      "loss": 1.3286,
+      "step": 32534
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.362028571428572e-06,
+      "loss": 1.3565,
+      "step": 32535
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.36176e-06,
+      "loss": 0.7414,
+      "step": 32536
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.361491428571429e-06,
+      "loss": 1.0697,
+      "step": 32537
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.361222857142857e-06,
+      "loss": 1.1535,
+      "step": 32538
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.360954285714286e-06,
+      "loss": 1.1463,
+      "step": 32539
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.360685714285714e-06,
+      "loss": 0.9476,
+      "step": 32540
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.360417142857142e-06,
+      "loss": 0.5431,
+      "step": 32541
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.360148571428571e-06,
+      "loss": 1.1936,
+      "step": 32542
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3598800000000004e-06,
+      "loss": 1.0149,
+      "step": 32543
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3596114285714294e-06,
+      "loss": 0.8732,
+      "step": 32544
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.359342857142858e-06,
+      "loss": 1.0573,
+      "step": 32545
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.359074285714286e-06,
+      "loss": 1.181,
+      "step": 32546
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.358805714285715e-06,
+      "loss": 1.1804,
+      "step": 32547
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.358537142857143e-06,
+      "loss": 0.6781,
+      "step": 32548
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.358268571428572e-06,
+      "loss": 1.7208,
+      "step": 32549
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.358e-06,
+      "loss": 1.4562,
+      "step": 32550
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.357731428571428e-06,
+      "loss": 1.3114,
+      "step": 32551
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.357462857142857e-06,
+      "loss": 1.324,
+      "step": 32552
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3571942857142856e-06,
+      "loss": 1.0957,
+      "step": 32553
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.356925714285714e-06,
+      "loss": 1.3376,
+      "step": 32554
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.356657142857143e-06,
+      "loss": 1.4894,
+      "step": 32555
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.356388571428572e-06,
+      "loss": 0.9657,
+      "step": 32556
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.356120000000001e-06,
+      "loss": 0.9257,
+      "step": 32557
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.355851428571429e-06,
+      "loss": 1.2498,
+      "step": 32558
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.355582857142857e-06,
+      "loss": 1.3356,
+      "step": 32559
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.355314285714286e-06,
+      "loss": 1.6029,
+      "step": 32560
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.355045714285714e-06,
+      "loss": 1.4622,
+      "step": 32561
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.354777142857143e-06,
+      "loss": 1.277,
+      "step": 32562
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3545085714285716e-06,
+      "loss": 0.9965,
+      "step": 32563
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.35424e-06,
+      "loss": 1.0717,
+      "step": 32564
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.353971428571429e-06,
+      "loss": 0.8633,
+      "step": 32565
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.353702857142857e-06,
+      "loss": 0.9736,
+      "step": 32566
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.353434285714286e-06,
+      "loss": 1.3404,
+      "step": 32567
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.353165714285714e-06,
+      "loss": 0.9545,
+      "step": 32568
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.352897142857143e-06,
+      "loss": 1.1307,
+      "step": 32569
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.352628571428572e-06,
+      "loss": 0.9409,
+      "step": 32570
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.35236e-06,
+      "loss": 0.651,
+      "step": 32571
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.352091428571429e-06,
+      "loss": 1.0755,
+      "step": 32572
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3518228571428576e-06,
+      "loss": 0.7377,
+      "step": 32573
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.351554285714286e-06,
+      "loss": 1.351,
+      "step": 32574
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.351285714285715e-06,
+      "loss": 1.179,
+      "step": 32575
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.351017142857143e-06,
+      "loss": 0.8272,
+      "step": 32576
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.350748571428571e-06,
+      "loss": 0.9928,
+      "step": 32577
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.35048e-06,
+      "loss": 1.4775,
+      "step": 32578
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.350211428571428e-06,
+      "loss": 1.0096,
+      "step": 32579
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.349942857142857e-06,
+      "loss": 1.0191,
+      "step": 32580
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3496742857142855e-06,
+      "loss": 1.2671,
+      "step": 32581
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3494057142857145e-06,
+      "loss": 1.0136,
+      "step": 32582
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3491371428571436e-06,
+      "loss": 0.8159,
+      "step": 32583
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.348868571428572e-06,
+      "loss": 1.2981,
+      "step": 32584
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.348600000000001e-06,
+      "loss": 0.8356,
+      "step": 32585
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.348331428571429e-06,
+      "loss": 0.995,
+      "step": 32586
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.348062857142857e-06,
+      "loss": 1.1078,
+      "step": 32587
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.347794285714286e-06,
+      "loss": 1.6761,
+      "step": 32588
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.347525714285714e-06,
+      "loss": 0.9066,
+      "step": 32589
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.347257142857143e-06,
+      "loss": 1.4382,
+      "step": 32590
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3469885714285715e-06,
+      "loss": 0.8448,
+      "step": 32591
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.34672e-06,
+      "loss": 1.3546,
+      "step": 32592
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.346451428571429e-06,
+      "loss": 1.3809,
+      "step": 32593
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.346182857142857e-06,
+      "loss": 1.0446,
+      "step": 32594
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.345914285714286e-06,
+      "loss": 1.0668,
+      "step": 32595
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.345645714285715e-06,
+      "loss": 1.2108,
+      "step": 32596
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.345377142857143e-06,
+      "loss": 1.7806,
+      "step": 32597
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.345108571428572e-06,
+      "loss": 1.2969,
+      "step": 32598
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.34484e-06,
+      "loss": 1.0522,
+      "step": 32599
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3445714285714285e-06,
+      "loss": 0.9426,
+      "step": 32600
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3443028571428575e-06,
+      "loss": 1.059,
+      "step": 32601
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.344034285714286e-06,
+      "loss": 1.5562,
+      "step": 32602
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.343765714285715e-06,
+      "loss": 1.4467,
+      "step": 32603
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.343497142857143e-06,
+      "loss": 0.8463,
+      "step": 32604
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.343228571428571e-06,
+      "loss": 1.1184,
+      "step": 32605
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.34296e-06,
+      "loss": 0.8263,
+      "step": 32606
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.342691428571428e-06,
+      "loss": 1.2138,
+      "step": 32607
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.342422857142857e-06,
+      "loss": 1.2663,
+      "step": 32608
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.342154285714286e-06,
+      "loss": 0.6763,
+      "step": 32609
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3418857142857145e-06,
+      "loss": 1.074,
+      "step": 32610
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3416171428571435e-06,
+      "loss": 0.9992,
+      "step": 32611
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.341348571428572e-06,
+      "loss": 0.8733,
+      "step": 32612
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.341080000000001e-06,
+      "loss": 0.7545,
+      "step": 32613
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.340811428571429e-06,
+      "loss": 1.0836,
+      "step": 32614
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.340542857142857e-06,
+      "loss": 1.1251,
+      "step": 32615
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.340274285714286e-06,
+      "loss": 1.335,
+      "step": 32616
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.340005714285714e-06,
+      "loss": 1.0166,
+      "step": 32617
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3397371428571424e-06,
+      "loss": 1.1228,
+      "step": 32618
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3394685714285715e-06,
+      "loss": 1.1609,
+      "step": 32619
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3392e-06,
+      "loss": 1.5475,
+      "step": 32620
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.338931428571429e-06,
+      "loss": 1.0464,
+      "step": 32621
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.338662857142858e-06,
+      "loss": 1.3484,
+      "step": 32622
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.338394285714286e-06,
+      "loss": 1.1876,
+      "step": 32623
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.338125714285715e-06,
+      "loss": 1.6461,
+      "step": 32624
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.337857142857143e-06,
+      "loss": 0.917,
+      "step": 32625
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.337588571428572e-06,
+      "loss": 1.2822,
+      "step": 32626
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.33732e-06,
+      "loss": 0.8283,
+      "step": 32627
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3370514285714284e-06,
+      "loss": 0.9618,
+      "step": 32628
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.3367828571428575e-06,
+      "loss": 0.9187,
+      "step": 32629
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.336514285714286e-06,
+      "loss": 1.0256,
+      "step": 32630
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.336245714285715e-06,
+      "loss": 0.7676,
+      "step": 32631
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.335977142857143e-06,
+      "loss": 1.4038,
+      "step": 32632
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.335708571428571e-06,
+      "loss": 1.3909,
+      "step": 32633
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.33544e-06,
+      "loss": 0.6987,
+      "step": 32634
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.335171428571429e-06,
+      "loss": 1.0275,
+      "step": 32635
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.334902857142858e-06,
+      "loss": 1.4775,
+      "step": 32636
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 5.334634285714286e-06,
+      "loss": 1.2087,
+      "step": 32637
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3343657142857144e-06,
+      "loss": 1.0477,
+      "step": 32638
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3340971428571435e-06,
+      "loss": 1.4406,
+      "step": 32639
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.333828571428572e-06,
+      "loss": 1.0292,
+      "step": 32640
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.33356e-06,
+      "loss": 0.6659,
+      "step": 32641
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.333291428571429e-06,
+      "loss": 0.9701,
+      "step": 32642
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.333022857142857e-06,
+      "loss": 1.1827,
+      "step": 32643
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.332754285714286e-06,
+      "loss": 0.8829,
+      "step": 32644
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.332485714285714e-06,
+      "loss": 0.9951,
+      "step": 32645
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.332217142857142e-06,
+      "loss": 0.7562,
+      "step": 32646
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.331948571428571e-06,
+      "loss": 0.9832,
+      "step": 32647
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3316800000000004e-06,
+      "loss": 1.6375,
+      "step": 32648
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3314114285714294e-06,
+      "loss": 1.1066,
+      "step": 32649
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.331142857142858e-06,
+      "loss": 0.9068,
+      "step": 32650
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.330874285714286e-06,
+      "loss": 1.4471,
+      "step": 32651
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.330605714285715e-06,
+      "loss": 0.9441,
+      "step": 32652
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.330337142857143e-06,
+      "loss": 1.0688,
+      "step": 32653
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.330068571428572e-06,
+      "loss": 1.0228,
+      "step": 32654
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3298e-06,
+      "loss": 0.8075,
+      "step": 32655
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.329531428571428e-06,
+      "loss": 1.5327,
+      "step": 32656
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.329262857142857e-06,
+      "loss": 1.4547,
+      "step": 32657
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3289942857142856e-06,
+      "loss": 1.5808,
+      "step": 32658
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.328725714285715e-06,
+      "loss": 1.0314,
+      "step": 32659
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.328457142857143e-06,
+      "loss": 1.1681,
+      "step": 32660
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.328188571428572e-06,
+      "loss": 1.388,
+      "step": 32661
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.327920000000001e-06,
+      "loss": 1.1844,
+      "step": 32662
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.327651428571429e-06,
+      "loss": 1.407,
+      "step": 32663
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.327382857142857e-06,
+      "loss": 1.2726,
+      "step": 32664
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.327114285714286e-06,
+      "loss": 0.9047,
+      "step": 32665
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.326845714285714e-06,
+      "loss": 1.3696,
+      "step": 32666
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.326577142857143e-06,
+      "loss": 1.3901,
+      "step": 32667
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3263085714285716e-06,
+      "loss": 1.2805,
+      "step": 32668
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.32604e-06,
+      "loss": 0.9899,
+      "step": 32669
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.325771428571429e-06,
+      "loss": 1.249,
+      "step": 32670
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.325502857142857e-06,
+      "loss": 0.7384,
+      "step": 32671
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.325234285714286e-06,
+      "loss": 1.1962,
+      "step": 32672
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.324965714285714e-06,
+      "loss": 1.3628,
+      "step": 32673
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.324697142857143e-06,
+      "loss": 1.2099,
+      "step": 32674
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.324428571428572e-06,
+      "loss": 0.9674,
+      "step": 32675
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.32416e-06,
+      "loss": 0.9838,
+      "step": 32676
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.323891428571429e-06,
+      "loss": 1.2206,
+      "step": 32677
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3236228571428576e-06,
+      "loss": 0.8184,
+      "step": 32678
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.323354285714286e-06,
+      "loss": 0.9528,
+      "step": 32679
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.323085714285715e-06,
+      "loss": 0.5891,
+      "step": 32680
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.322817142857143e-06,
+      "loss": 1.2851,
+      "step": 32681
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.322548571428572e-06,
+      "loss": 1.1112,
+      "step": 32682
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.32228e-06,
+      "loss": 1.2913,
+      "step": 32683
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.322011428571428e-06,
+      "loss": 1.3306,
+      "step": 32684
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.321742857142857e-06,
+      "loss": 1.2759,
+      "step": 32685
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3214742857142855e-06,
+      "loss": 1.4718,
+      "step": 32686
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3212057142857145e-06,
+      "loss": 1.4164,
+      "step": 32687
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3209371428571436e-06,
+      "loss": 1.011,
+      "step": 32688
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.320668571428572e-06,
+      "loss": 1.2431,
+      "step": 32689
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.320400000000001e-06,
+      "loss": 1.2132,
+      "step": 32690
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.320131428571429e-06,
+      "loss": 0.9814,
+      "step": 32691
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.319862857142857e-06,
+      "loss": 1.0298,
+      "step": 32692
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.319594285714286e-06,
+      "loss": 1.5806,
+      "step": 32693
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.319325714285714e-06,
+      "loss": 0.9971,
+      "step": 32694
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.319057142857143e-06,
+      "loss": 1.0426,
+      "step": 32695
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3187885714285715e-06,
+      "loss": 1.2245,
+      "step": 32696
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.31852e-06,
+      "loss": 1.001,
+      "step": 32697
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.318251428571429e-06,
+      "loss": 1.5262,
+      "step": 32698
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.317982857142857e-06,
+      "loss": 0.9886,
+      "step": 32699
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.317714285714286e-06,
+      "loss": 1.2625,
+      "step": 32700
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.317445714285715e-06,
+      "loss": 0.6822,
+      "step": 32701
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.317177142857143e-06,
+      "loss": 0.9135,
+      "step": 32702
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.316908571428572e-06,
+      "loss": 1.5929,
+      "step": 32703
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.31664e-06,
+      "loss": 0.9661,
+      "step": 32704
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3163714285714285e-06,
+      "loss": 1.2391,
+      "step": 32705
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3161028571428575e-06,
+      "loss": 1.6202,
+      "step": 32706
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.315834285714286e-06,
+      "loss": 0.8944,
+      "step": 32707
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.315565714285715e-06,
+      "loss": 1.0499,
+      "step": 32708
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.315297142857143e-06,
+      "loss": 0.8236,
+      "step": 32709
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.315028571428571e-06,
+      "loss": 0.7165,
+      "step": 32710
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.31476e-06,
+      "loss": 1.1056,
+      "step": 32711
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.314491428571428e-06,
+      "loss": 1.2081,
+      "step": 32712
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.314222857142857e-06,
+      "loss": 0.9585,
+      "step": 32713
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.313954285714286e-06,
+      "loss": 1.2772,
+      "step": 32714
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3136857142857145e-06,
+      "loss": 0.8563,
+      "step": 32715
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3134171428571435e-06,
+      "loss": 1.0806,
+      "step": 32716
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.313148571428572e-06,
+      "loss": 0.798,
+      "step": 32717
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.312880000000001e-06,
+      "loss": 1.3925,
+      "step": 32718
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.312611428571429e-06,
+      "loss": 1.4537,
+      "step": 32719
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.312342857142857e-06,
+      "loss": 1.232,
+      "step": 32720
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.312074285714286e-06,
+      "loss": 0.8116,
+      "step": 32721
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.311805714285714e-06,
+      "loss": 1.3535,
+      "step": 32722
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.311537142857143e-06,
+      "loss": 0.7933,
+      "step": 32723
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3112685714285715e-06,
+      "loss": 1.2553,
+      "step": 32724
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.311e-06,
+      "loss": 0.8841,
+      "step": 32725
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.310731428571429e-06,
+      "loss": 1.1705,
+      "step": 32726
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.310462857142858e-06,
+      "loss": 1.0419,
+      "step": 32727
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.310194285714286e-06,
+      "loss": 1.2778,
+      "step": 32728
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.309925714285715e-06,
+      "loss": 1.2499,
+      "step": 32729
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.309657142857143e-06,
+      "loss": 1.2756,
+      "step": 32730
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.309388571428572e-06,
+      "loss": 1.6208,
+      "step": 32731
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.30912e-06,
+      "loss": 1.1926,
+      "step": 32732
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3088514285714284e-06,
+      "loss": 1.2123,
+      "step": 32733
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3085828571428575e-06,
+      "loss": 1.2251,
+      "step": 32734
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.308314285714286e-06,
+      "loss": 1.0319,
+      "step": 32735
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.308045714285715e-06,
+      "loss": 1.5333,
+      "step": 32736
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.307777142857143e-06,
+      "loss": 0.9714,
+      "step": 32737
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.307508571428571e-06,
+      "loss": 0.9127,
+      "step": 32738
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.30724e-06,
+      "loss": 0.9368,
+      "step": 32739
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.306971428571429e-06,
+      "loss": 1.3658,
+      "step": 32740
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.306702857142858e-06,
+      "loss": 1.2128,
+      "step": 32741
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.306434285714286e-06,
+      "loss": 0.948,
+      "step": 32742
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3061657142857144e-06,
+      "loss": 0.8659,
+      "step": 32743
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3058971428571435e-06,
+      "loss": 0.8925,
+      "step": 32744
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.305628571428572e-06,
+      "loss": 0.9665,
+      "step": 32745
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.305360000000001e-06,
+      "loss": 1.2181,
+      "step": 32746
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.305091428571429e-06,
+      "loss": 1.1502,
+      "step": 32747
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.304822857142857e-06,
+      "loss": 1.1555,
+      "step": 32748
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.304554285714286e-06,
+      "loss": 1.4375,
+      "step": 32749
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.304285714285714e-06,
+      "loss": 1.42,
+      "step": 32750
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.304017142857142e-06,
+      "loss": 1.5723,
+      "step": 32751
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.303748571428571e-06,
+      "loss": 0.955,
+      "step": 32752
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3034800000000004e-06,
+      "loss": 1.0808,
+      "step": 32753
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3032114285714294e-06,
+      "loss": 1.393,
+      "step": 32754
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.302942857142858e-06,
+      "loss": 0.7973,
+      "step": 32755
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.302674285714286e-06,
+      "loss": 1.7058,
+      "step": 32756
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.302405714285715e-06,
+      "loss": 0.8016,
+      "step": 32757
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.302137142857143e-06,
+      "loss": 0.9419,
+      "step": 32758
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.301868571428572e-06,
+      "loss": 1.1841,
+      "step": 32759
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3016e-06,
+      "loss": 1.6073,
+      "step": 32760
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.301331428571428e-06,
+      "loss": 1.3089,
+      "step": 32761
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.301062857142857e-06,
+      "loss": 1.0378,
+      "step": 32762
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.3007942857142856e-06,
+      "loss": 0.8258,
+      "step": 32763
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.300525714285715e-06,
+      "loss": 0.9275,
+      "step": 32764
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.300257142857143e-06,
+      "loss": 1.1722,
+      "step": 32765
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.299988571428571e-06,
+      "loss": 0.6877,
+      "step": 32766
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.299720000000001e-06,
+      "loss": 1.2094,
+      "step": 32767
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.299451428571429e-06,
+      "loss": 1.0943,
+      "step": 32768
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.299182857142858e-06,
+      "loss": 1.3436,
+      "step": 32769
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.298914285714286e-06,
+      "loss": 1.5157,
+      "step": 32770
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.298645714285714e-06,
+      "loss": 1.075,
+      "step": 32771
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.298377142857143e-06,
+      "loss": 1.0227,
+      "step": 32772
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.2981085714285716e-06,
+      "loss": 1.14,
+      "step": 32773
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.29784e-06,
+      "loss": 1.2929,
+      "step": 32774
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.297571428571429e-06,
+      "loss": 1.2091,
+      "step": 32775
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.297302857142857e-06,
+      "loss": 0.9748,
+      "step": 32776
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.297034285714286e-06,
+      "loss": 0.7876,
+      "step": 32777
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.296765714285714e-06,
+      "loss": 1.5485,
+      "step": 32778
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.296497142857142e-06,
+      "loss": 1.3637,
+      "step": 32779
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.296228571428572e-06,
+      "loss": 0.9964,
+      "step": 32780
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.29596e-06,
+      "loss": 1.1678,
+      "step": 32781
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.295691428571429e-06,
+      "loss": 0.7813,
+      "step": 32782
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.2954228571428576e-06,
+      "loss": 1.1999,
+      "step": 32783
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.295154285714286e-06,
+      "loss": 1.2358,
+      "step": 32784
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.294885714285715e-06,
+      "loss": 1.4223,
+      "step": 32785
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.294617142857143e-06,
+      "loss": 1.0672,
+      "step": 32786
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.294348571428572e-06,
+      "loss": 1.5318,
+      "step": 32787
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.29408e-06,
+      "loss": 1.2032,
+      "step": 32788
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.293811428571428e-06,
+      "loss": 1.328,
+      "step": 32789
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.293542857142857e-06,
+      "loss": 0.8695,
+      "step": 32790
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.2932742857142855e-06,
+      "loss": 1.2309,
+      "step": 32791
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.293005714285714e-06,
+      "loss": 1.454,
+      "step": 32792
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.2927371428571436e-06,
+      "loss": 1.4218,
+      "step": 32793
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.292468571428572e-06,
+      "loss": 0.9465,
+      "step": 32794
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.292200000000001e-06,
+      "loss": 1.0924,
+      "step": 32795
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.291931428571429e-06,
+      "loss": 1.1073,
+      "step": 32796
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.291662857142857e-06,
+      "loss": 0.8779,
+      "step": 32797
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.291394285714286e-06,
+      "loss": 1.3239,
+      "step": 32798
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.291125714285714e-06,
+      "loss": 0.7613,
+      "step": 32799
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.290857142857143e-06,
+      "loss": 1.0689,
+      "step": 32800
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.2905885714285715e-06,
+      "loss": 1.2465,
+      "step": 32801
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.29032e-06,
+      "loss": 1.3329,
+      "step": 32802
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.290051428571429e-06,
+      "loss": 1.0319,
+      "step": 32803
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.289782857142857e-06,
+      "loss": 0.8843,
+      "step": 32804
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.289514285714286e-06,
+      "loss": 1.4199,
+      "step": 32805
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.289245714285715e-06,
+      "loss": 1.219,
+      "step": 32806
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.288977142857143e-06,
+      "loss": 1.1621,
+      "step": 32807
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.288708571428572e-06,
+      "loss": 0.976,
+      "step": 32808
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.28844e-06,
+      "loss": 1.2838,
+      "step": 32809
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.288171428571429e-06,
+      "loss": 1.2845,
+      "step": 32810
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.2879028571428575e-06,
+      "loss": 1.0017,
+      "step": 32811
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 5.287634285714286e-06,
+      "loss": 1.325,
+      "step": 32812
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.287365714285715e-06,
+      "loss": 1.189,
+      "step": 32813
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.287097142857143e-06,
+      "loss": 0.8258,
+      "step": 32814
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.286828571428571e-06,
+      "loss": 0.9259,
+      "step": 32815
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.28656e-06,
+      "loss": 1.1552,
+      "step": 32816
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.286291428571428e-06,
+      "loss": 1.2642,
+      "step": 32817
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.286022857142857e-06,
+      "loss": 1.5778,
+      "step": 32818
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2857542857142855e-06,
+      "loss": 1.4617,
+      "step": 32819
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2854857142857145e-06,
+      "loss": 1.209,
+      "step": 32820
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2852171428571435e-06,
+      "loss": 1.0678,
+      "step": 32821
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.284948571428572e-06,
+      "loss": 1.6999,
+      "step": 32822
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.284680000000001e-06,
+      "loss": 1.4172,
+      "step": 32823
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.284411428571429e-06,
+      "loss": 1.1825,
+      "step": 32824
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.284142857142857e-06,
+      "loss": 1.3062,
+      "step": 32825
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.283874285714286e-06,
+      "loss": 1.1554,
+      "step": 32826
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.283605714285714e-06,
+      "loss": 0.9902,
+      "step": 32827
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.283337142857143e-06,
+      "loss": 1.5769,
+      "step": 32828
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2830685714285715e-06,
+      "loss": 0.6686,
+      "step": 32829
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2828e-06,
+      "loss": 0.5807,
+      "step": 32830
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.282531428571429e-06,
+      "loss": 0.9106,
+      "step": 32831
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.282262857142857e-06,
+      "loss": 1.5178,
+      "step": 32832
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.281994285714287e-06,
+      "loss": 1.0868,
+      "step": 32833
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.281725714285715e-06,
+      "loss": 0.9517,
+      "step": 32834
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.281457142857143e-06,
+      "loss": 1.2474,
+      "step": 32835
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.281188571428572e-06,
+      "loss": 1.0784,
+      "step": 32836
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.28092e-06,
+      "loss": 1.2255,
+      "step": 32837
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2806514285714284e-06,
+      "loss": 0.9657,
+      "step": 32838
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2803828571428575e-06,
+      "loss": 1.1601,
+      "step": 32839
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.280114285714286e-06,
+      "loss": 1.146,
+      "step": 32840
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.279845714285715e-06,
+      "loss": 1.2264,
+      "step": 32841
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.279577142857143e-06,
+      "loss": 1.6126,
+      "step": 32842
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.279308571428571e-06,
+      "loss": 0.8429,
+      "step": 32843
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.27904e-06,
+      "loss": 1.0808,
+      "step": 32844
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.278771428571428e-06,
+      "loss": 1.1758,
+      "step": 32845
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.278502857142858e-06,
+      "loss": 0.8092,
+      "step": 32846
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.278234285714286e-06,
+      "loss": 1.2544,
+      "step": 32847
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2779657142857144e-06,
+      "loss": 0.9018,
+      "step": 32848
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2776971428571434e-06,
+      "loss": 1.4477,
+      "step": 32849
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.277428571428572e-06,
+      "loss": 1.161,
+      "step": 32850
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.277160000000001e-06,
+      "loss": 1.0575,
+      "step": 32851
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.276891428571429e-06,
+      "loss": 0.9714,
+      "step": 32852
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.276622857142857e-06,
+      "loss": 1.3595,
+      "step": 32853
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.276354285714286e-06,
+      "loss": 1.1113,
+      "step": 32854
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.276085714285714e-06,
+      "loss": 1.3351,
+      "step": 32855
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.275817142857142e-06,
+      "loss": 1.23,
+      "step": 32856
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.275548571428571e-06,
+      "loss": 0.9175,
+      "step": 32857
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2752799999999996e-06,
+      "loss": 0.9712,
+      "step": 32858
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2750114285714294e-06,
+      "loss": 1.1581,
+      "step": 32859
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.274742857142858e-06,
+      "loss": 1.1828,
+      "step": 32860
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.274474285714286e-06,
+      "loss": 0.7054,
+      "step": 32861
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.274205714285715e-06,
+      "loss": 1.1777,
+      "step": 32862
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.273937142857143e-06,
+      "loss": 1.0482,
+      "step": 32863
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.273668571428572e-06,
+      "loss": 0.7337,
+      "step": 32864
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2734e-06,
+      "loss": 1.001,
+      "step": 32865
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.273131428571428e-06,
+      "loss": 1.0883,
+      "step": 32866
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.272862857142857e-06,
+      "loss": 1.1442,
+      "step": 32867
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2725942857142856e-06,
+      "loss": 1.295,
+      "step": 32868
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.272325714285715e-06,
+      "loss": 1.3902,
+      "step": 32869
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.272057142857143e-06,
+      "loss": 1.8405,
+      "step": 32870
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.271788571428571e-06,
+      "loss": 1.1907,
+      "step": 32871
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.27152e-06,
+      "loss": 1.507,
+      "step": 32872
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.271251428571429e-06,
+      "loss": 0.9299,
+      "step": 32873
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.270982857142858e-06,
+      "loss": 0.8587,
+      "step": 32874
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.270714285714286e-06,
+      "loss": 0.9389,
+      "step": 32875
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.270445714285714e-06,
+      "loss": 0.9204,
+      "step": 32876
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.270177142857143e-06,
+      "loss": 1.4769,
+      "step": 32877
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2699085714285716e-06,
+      "loss": 1.1867,
+      "step": 32878
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.26964e-06,
+      "loss": 1.0109,
+      "step": 32879
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.269371428571429e-06,
+      "loss": 1.1706,
+      "step": 32880
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.269102857142857e-06,
+      "loss": 0.9651,
+      "step": 32881
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.268834285714286e-06,
+      "loss": 1.2591,
+      "step": 32882
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.268565714285714e-06,
+      "loss": 1.2081,
+      "step": 32883
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.268297142857142e-06,
+      "loss": 1.3582,
+      "step": 32884
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.268028571428571e-06,
+      "loss": 1.7327,
+      "step": 32885
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.26776e-06,
+      "loss": 1.0277,
+      "step": 32886
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.267491428571429e-06,
+      "loss": 1.227,
+      "step": 32887
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2672228571428576e-06,
+      "loss": 1.0248,
+      "step": 32888
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.266954285714286e-06,
+      "loss": 1.326,
+      "step": 32889
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.266685714285715e-06,
+      "loss": 1.1068,
+      "step": 32890
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.266417142857143e-06,
+      "loss": 0.8156,
+      "step": 32891
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.266148571428572e-06,
+      "loss": 1.1962,
+      "step": 32892
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.26588e-06,
+      "loss": 1.2166,
+      "step": 32893
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.265611428571428e-06,
+      "loss": 1.5696,
+      "step": 32894
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.265342857142857e-06,
+      "loss": 1.063,
+      "step": 32895
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2650742857142855e-06,
+      "loss": 1.3085,
+      "step": 32896
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2648057142857145e-06,
+      "loss": 1.2682,
+      "step": 32897
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.264537142857143e-06,
+      "loss": 0.9487,
+      "step": 32898
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.264268571428572e-06,
+      "loss": 0.923,
+      "step": 32899
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.264000000000001e-06,
+      "loss": 1.136,
+      "step": 32900
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.263731428571429e-06,
+      "loss": 1.0582,
+      "step": 32901
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.263462857142857e-06,
+      "loss": 0.9535,
+      "step": 32902
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.263194285714286e-06,
+      "loss": 1.0994,
+      "step": 32903
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.262925714285714e-06,
+      "loss": 1.2142,
+      "step": 32904
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.262657142857143e-06,
+      "loss": 1.1946,
+      "step": 32905
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2623885714285715e-06,
+      "loss": 1.3041,
+      "step": 32906
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.26212e-06,
+      "loss": 1.4194,
+      "step": 32907
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.261851428571429e-06,
+      "loss": 0.9186,
+      "step": 32908
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.261582857142857e-06,
+      "loss": 1.1129,
+      "step": 32909
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.261314285714286e-06,
+      "loss": 1.1165,
+      "step": 32910
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.261045714285714e-06,
+      "loss": 1.4812,
+      "step": 32911
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.260777142857143e-06,
+      "loss": 0.9438,
+      "step": 32912
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.260508571428572e-06,
+      "loss": 0.8423,
+      "step": 32913
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.26024e-06,
+      "loss": 0.8912,
+      "step": 32914
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.259971428571429e-06,
+      "loss": 0.9654,
+      "step": 32915
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2597028571428575e-06,
+      "loss": 1.3386,
+      "step": 32916
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.259434285714286e-06,
+      "loss": 1.1741,
+      "step": 32917
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.259165714285715e-06,
+      "loss": 1.1131,
+      "step": 32918
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.258897142857143e-06,
+      "loss": 0.8991,
+      "step": 32919
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.258628571428572e-06,
+      "loss": 1.1454,
+      "step": 32920
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.25836e-06,
+      "loss": 0.7289,
+      "step": 32921
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.258091428571428e-06,
+      "loss": 1.1258,
+      "step": 32922
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.257822857142857e-06,
+      "loss": 1.4397,
+      "step": 32923
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2575542857142855e-06,
+      "loss": 1.6583,
+      "step": 32924
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2572857142857145e-06,
+      "loss": 1.2138,
+      "step": 32925
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2570171428571435e-06,
+      "loss": 1.3192,
+      "step": 32926
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.256748571428572e-06,
+      "loss": 1.5468,
+      "step": 32927
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.256480000000001e-06,
+      "loss": 0.9569,
+      "step": 32928
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.256211428571429e-06,
+      "loss": 1.1333,
+      "step": 32929
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.255942857142857e-06,
+      "loss": 1.0801,
+      "step": 32930
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.255674285714286e-06,
+      "loss": 1.399,
+      "step": 32931
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.255405714285714e-06,
+      "loss": 0.8455,
+      "step": 32932
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.255137142857143e-06,
+      "loss": 1.0563,
+      "step": 32933
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2548685714285715e-06,
+      "loss": 1.0371,
+      "step": 32934
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2546e-06,
+      "loss": 1.0529,
+      "step": 32935
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.254331428571429e-06,
+      "loss": 1.0003,
+      "step": 32936
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.254062857142857e-06,
+      "loss": 0.9199,
+      "step": 32937
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.253794285714286e-06,
+      "loss": 1.4282,
+      "step": 32938
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.253525714285715e-06,
+      "loss": 1.0824,
+      "step": 32939
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.253257142857143e-06,
+      "loss": 1.3625,
+      "step": 32940
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.252988571428572e-06,
+      "loss": 1.2627,
+      "step": 32941
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.25272e-06,
+      "loss": 1.253,
+      "step": 32942
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2524514285714284e-06,
+      "loss": 0.7931,
+      "step": 32943
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2521828571428575e-06,
+      "loss": 1.0823,
+      "step": 32944
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.251914285714286e-06,
+      "loss": 1.2701,
+      "step": 32945
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.251645714285715e-06,
+      "loss": 1.2381,
+      "step": 32946
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.251377142857143e-06,
+      "loss": 1.3775,
+      "step": 32947
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.251108571428571e-06,
+      "loss": 0.8012,
+      "step": 32948
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.25084e-06,
+      "loss": 0.9268,
+      "step": 32949
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.250571428571428e-06,
+      "loss": 1.3245,
+      "step": 32950
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.250302857142857e-06,
+      "loss": 1.2709,
+      "step": 32951
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.250034285714286e-06,
+      "loss": 1.3374,
+      "step": 32952
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2497657142857144e-06,
+      "loss": 1.0381,
+      "step": 32953
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2494971428571434e-06,
+      "loss": 1.2449,
+      "step": 32954
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.249228571428572e-06,
+      "loss": 1.2966,
+      "step": 32955
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.248960000000001e-06,
+      "loss": 1.1432,
+      "step": 32956
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.248691428571429e-06,
+      "loss": 1.055,
+      "step": 32957
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.248422857142857e-06,
+      "loss": 1.2135,
+      "step": 32958
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.248154285714286e-06,
+      "loss": 1.1227,
+      "step": 32959
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.247885714285714e-06,
+      "loss": 0.6848,
+      "step": 32960
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.247617142857143e-06,
+      "loss": 0.7616,
+      "step": 32961
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.247348571428571e-06,
+      "loss": 1.1495,
+      "step": 32962
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2470799999999996e-06,
+      "loss": 1.2205,
+      "step": 32963
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.246811428571429e-06,
+      "loss": 1.2745,
+      "step": 32964
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.246542857142858e-06,
+      "loss": 1.0139,
+      "step": 32965
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.246274285714286e-06,
+      "loss": 0.9952,
+      "step": 32966
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.246005714285715e-06,
+      "loss": 0.9765,
+      "step": 32967
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.245737142857143e-06,
+      "loss": 1.0103,
+      "step": 32968
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.245468571428572e-06,
+      "loss": 0.9556,
+      "step": 32969
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2452e-06,
+      "loss": 0.8155,
+      "step": 32970
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.244931428571428e-06,
+      "loss": 1.0306,
+      "step": 32971
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.244662857142857e-06,
+      "loss": 1.2641,
+      "step": 32972
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2443942857142856e-06,
+      "loss": 1.0991,
+      "step": 32973
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.244125714285715e-06,
+      "loss": 0.8716,
+      "step": 32974
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.243857142857143e-06,
+      "loss": 1.1435,
+      "step": 32975
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.243588571428571e-06,
+      "loss": 0.8105,
+      "step": 32976
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.24332e-06,
+      "loss": 0.9892,
+      "step": 32977
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.243051428571429e-06,
+      "loss": 1.1247,
+      "step": 32978
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.242782857142858e-06,
+      "loss": 1.1244,
+      "step": 32979
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.242514285714286e-06,
+      "loss": 0.8224,
+      "step": 32980
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.242245714285714e-06,
+      "loss": 1.4722,
+      "step": 32981
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.241977142857143e-06,
+      "loss": 0.9558,
+      "step": 32982
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.2417085714285716e-06,
+      "loss": 0.8907,
+      "step": 32983
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.241440000000001e-06,
+      "loss": 0.8802,
+      "step": 32984
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.241171428571429e-06,
+      "loss": 1.5904,
+      "step": 32985
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.240902857142857e-06,
+      "loss": 1.1385,
+      "step": 32986
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.240634285714286e-06,
+      "loss": 0.8117,
+      "step": 32987
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.240365714285714e-06,
+      "loss": 1.3627,
+      "step": 32988
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.240097142857142e-06,
+      "loss": 1.5287,
+      "step": 32989
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.239828571428571e-06,
+      "loss": 1.0256,
+      "step": 32990
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.23956e-06,
+      "loss": 1.7836,
+      "step": 32991
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.239291428571429e-06,
+      "loss": 1.0459,
+      "step": 32992
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2390228571428576e-06,
+      "loss": 1.0627,
+      "step": 32993
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.238754285714286e-06,
+      "loss": 0.9497,
+      "step": 32994
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.238485714285715e-06,
+      "loss": 1.0472,
+      "step": 32995
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.238217142857143e-06,
+      "loss": 1.2356,
+      "step": 32996
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.237948571428572e-06,
+      "loss": 0.7586,
+      "step": 32997
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.23768e-06,
+      "loss": 1.6359,
+      "step": 32998
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.237411428571428e-06,
+      "loss": 1.7053,
+      "step": 32999
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.237142857142857e-06,
+      "loss": 1.2115,
+      "step": 33000
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2368742857142855e-06,
+      "loss": 1.2372,
+      "step": 33001
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2366057142857145e-06,
+      "loss": 1.0656,
+      "step": 33002
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.236337142857143e-06,
+      "loss": 1.2063,
+      "step": 33003
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.236068571428572e-06,
+      "loss": 0.9524,
+      "step": 33004
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.235800000000001e-06,
+      "loss": 1.2229,
+      "step": 33005
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.235531428571429e-06,
+      "loss": 1.4971,
+      "step": 33006
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.235262857142858e-06,
+      "loss": 1.2196,
+      "step": 33007
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.234994285714286e-06,
+      "loss": 1.6999,
+      "step": 33008
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.234725714285714e-06,
+      "loss": 1.1529,
+      "step": 33009
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.234457142857143e-06,
+      "loss": 0.9411,
+      "step": 33010
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2341885714285715e-06,
+      "loss": 1.09,
+      "step": 33011
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.23392e-06,
+      "loss": 1.5999,
+      "step": 33012
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.233651428571429e-06,
+      "loss": 1.0477,
+      "step": 33013
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.233382857142857e-06,
+      "loss": 1.4148,
+      "step": 33014
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.233114285714286e-06,
+      "loss": 1.1573,
+      "step": 33015
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.232845714285714e-06,
+      "loss": 1.1593,
+      "step": 33016
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.232577142857143e-06,
+      "loss": 1.0727,
+      "step": 33017
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.232308571428572e-06,
+      "loss": 1.2421,
+      "step": 33018
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.23204e-06,
+      "loss": 1.1697,
+      "step": 33019
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.231771428571429e-06,
+      "loss": 1.0736,
+      "step": 33020
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2315028571428575e-06,
+      "loss": 0.7585,
+      "step": 33021
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.231234285714286e-06,
+      "loss": 1.1358,
+      "step": 33022
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.230965714285715e-06,
+      "loss": 1.2148,
+      "step": 33023
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.230697142857143e-06,
+      "loss": 0.9748,
+      "step": 33024
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.230428571428572e-06,
+      "loss": 1.1491,
+      "step": 33025
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.23016e-06,
+      "loss": 0.9495,
+      "step": 33026
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.229891428571428e-06,
+      "loss": 0.8499,
+      "step": 33027
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.229622857142857e-06,
+      "loss": 1.4914,
+      "step": 33028
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2293542857142855e-06,
+      "loss": 1.193,
+      "step": 33029
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2290857142857145e-06,
+      "loss": 1.0608,
+      "step": 33030
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2288171428571435e-06,
+      "loss": 1.1129,
+      "step": 33031
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.228548571428572e-06,
+      "loss": 1.2994,
+      "step": 33032
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.228280000000001e-06,
+      "loss": 0.7375,
+      "step": 33033
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.228011428571429e-06,
+      "loss": 1.0179,
+      "step": 33034
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.227742857142857e-06,
+      "loss": 1.0854,
+      "step": 33035
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.227474285714286e-06,
+      "loss": 1.1354,
+      "step": 33036
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.227205714285714e-06,
+      "loss": 1.4884,
+      "step": 33037
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.226937142857143e-06,
+      "loss": 1.0736,
+      "step": 33038
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2266685714285715e-06,
+      "loss": 1.4293,
+      "step": 33039
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2264e-06,
+      "loss": 1.0686,
+      "step": 33040
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.226131428571429e-06,
+      "loss": 1.4327,
+      "step": 33041
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.225862857142857e-06,
+      "loss": 1.0502,
+      "step": 33042
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.225594285714286e-06,
+      "loss": 0.9973,
+      "step": 33043
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.225325714285715e-06,
+      "loss": 1.1319,
+      "step": 33044
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.225057142857143e-06,
+      "loss": 1.0695,
+      "step": 33045
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.224788571428572e-06,
+      "loss": 0.8837,
+      "step": 33046
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.22452e-06,
+      "loss": 0.9633,
+      "step": 33047
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.224251428571429e-06,
+      "loss": 1.1811,
+      "step": 33048
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2239828571428575e-06,
+      "loss": 1.3679,
+      "step": 33049
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.223714285714286e-06,
+      "loss": 1.1081,
+      "step": 33050
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.223445714285715e-06,
+      "loss": 1.2773,
+      "step": 33051
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.223177142857143e-06,
+      "loss": 0.973,
+      "step": 33052
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.222908571428571e-06,
+      "loss": 1.34,
+      "step": 33053
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.22264e-06,
+      "loss": 1.3244,
+      "step": 33054
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.222371428571428e-06,
+      "loss": 1.0723,
+      "step": 33055
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.222102857142857e-06,
+      "loss": 1.318,
+      "step": 33056
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.221834285714286e-06,
+      "loss": 1.2289,
+      "step": 33057
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2215657142857144e-06,
+      "loss": 0.9708,
+      "step": 33058
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2212971428571434e-06,
+      "loss": 1.681,
+      "step": 33059
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.221028571428572e-06,
+      "loss": 1.0617,
+      "step": 33060
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.220760000000001e-06,
+      "loss": 1.2197,
+      "step": 33061
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.220491428571429e-06,
+      "loss": 0.8515,
+      "step": 33062
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.220222857142857e-06,
+      "loss": 1.2613,
+      "step": 33063
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.219954285714286e-06,
+      "loss": 0.794,
+      "step": 33064
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.219685714285714e-06,
+      "loss": 0.825,
+      "step": 33065
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.219417142857143e-06,
+      "loss": 1.3415,
+      "step": 33066
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.219148571428571e-06,
+      "loss": 1.0841,
+      "step": 33067
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2188799999999996e-06,
+      "loss": 0.9554,
+      "step": 33068
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.218611428571429e-06,
+      "loss": 1.2242,
+      "step": 33069
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.218342857142858e-06,
+      "loss": 1.1097,
+      "step": 33070
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.218074285714287e-06,
+      "loss": 0.9866,
+      "step": 33071
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.217805714285715e-06,
+      "loss": 0.9145,
+      "step": 33072
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.217537142857143e-06,
+      "loss": 0.7907,
+      "step": 33073
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.217268571428572e-06,
+      "loss": 1.2344,
+      "step": 33074
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.217e-06,
+      "loss": 1.0872,
+      "step": 33075
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.216731428571428e-06,
+      "loss": 1.4667,
+      "step": 33076
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.216462857142857e-06,
+      "loss": 1.013,
+      "step": 33077
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2161942857142856e-06,
+      "loss": 1.0334,
+      "step": 33078
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.215925714285715e-06,
+      "loss": 0.7426,
+      "step": 33079
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.215657142857143e-06,
+      "loss": 1.0374,
+      "step": 33080
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.215388571428571e-06,
+      "loss": 0.9063,
+      "step": 33081
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.21512e-06,
+      "loss": 1.3274,
+      "step": 33082
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.214851428571429e-06,
+      "loss": 1.446,
+      "step": 33083
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.214582857142858e-06,
+      "loss": 1.409,
+      "step": 33084
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.214314285714286e-06,
+      "loss": 1.1213,
+      "step": 33085
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.214045714285714e-06,
+      "loss": 1.0111,
+      "step": 33086
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.213777142857143e-06,
+      "loss": 1.0218,
+      "step": 33087
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2135085714285716e-06,
+      "loss": 0.8083,
+      "step": 33088
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.213240000000001e-06,
+      "loss": 1.1207,
+      "step": 33089
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.212971428571429e-06,
+      "loss": 0.7046,
+      "step": 33090
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.212702857142857e-06,
+      "loss": 0.9439,
+      "step": 33091
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.212434285714286e-06,
+      "loss": 1.0664,
+      "step": 33092
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.212165714285714e-06,
+      "loss": 1.4793,
+      "step": 33093
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.211897142857143e-06,
+      "loss": 1.0562,
+      "step": 33094
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.211628571428571e-06,
+      "loss": 1.1784,
+      "step": 33095
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.21136e-06,
+      "loss": 1.0848,
+      "step": 33096
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.211091428571429e-06,
+      "loss": 1.4218,
+      "step": 33097
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2108228571428576e-06,
+      "loss": 1.0702,
+      "step": 33098
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.210554285714286e-06,
+      "loss": 1.3203,
+      "step": 33099
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.210285714285715e-06,
+      "loss": 1.2189,
+      "step": 33100
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.210017142857143e-06,
+      "loss": 0.857,
+      "step": 33101
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.209748571428572e-06,
+      "loss": 1.0665,
+      "step": 33102
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.20948e-06,
+      "loss": 1.3067,
+      "step": 33103
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.209211428571428e-06,
+      "loss": 1.4046,
+      "step": 33104
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.208942857142857e-06,
+      "loss": 0.8453,
+      "step": 33105
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2086742857142855e-06,
+      "loss": 1.2722,
+      "step": 33106
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2084057142857145e-06,
+      "loss": 0.8972,
+      "step": 33107
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.208137142857143e-06,
+      "loss": 1.0233,
+      "step": 33108
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.207868571428572e-06,
+      "loss": 1.0134,
+      "step": 33109
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.207600000000001e-06,
+      "loss": 1.3164,
+      "step": 33110
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.207331428571429e-06,
+      "loss": 1.1921,
+      "step": 33111
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.207062857142858e-06,
+      "loss": 1.1514,
+      "step": 33112
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.206794285714286e-06,
+      "loss": 1.0615,
+      "step": 33113
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.206525714285714e-06,
+      "loss": 0.9634,
+      "step": 33114
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.206257142857143e-06,
+      "loss": 1.056,
+      "step": 33115
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2059885714285715e-06,
+      "loss": 1.0637,
+      "step": 33116
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.20572e-06,
+      "loss": 0.8919,
+      "step": 33117
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.205451428571429e-06,
+      "loss": 0.7195,
+      "step": 33118
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.205182857142857e-06,
+      "loss": 1.028,
+      "step": 33119
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.204914285714286e-06,
+      "loss": 0.8229,
+      "step": 33120
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.204645714285714e-06,
+      "loss": 1.0486,
+      "step": 33121
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.204377142857143e-06,
+      "loss": 0.8573,
+      "step": 33122
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.204108571428572e-06,
+      "loss": 0.7745,
+      "step": 33123
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.20384e-06,
+      "loss": 0.9826,
+      "step": 33124
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.203571428571429e-06,
+      "loss": 0.9321,
+      "step": 33125
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2033028571428575e-06,
+      "loss": 1.2017,
+      "step": 33126
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.203034285714286e-06,
+      "loss": 1.117,
+      "step": 33127
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.202765714285715e-06,
+      "loss": 0.8523,
+      "step": 33128
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.202497142857143e-06,
+      "loss": 1.3186,
+      "step": 33129
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.202228571428572e-06,
+      "loss": 0.8778,
+      "step": 33130
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.20196e-06,
+      "loss": 1.1504,
+      "step": 33131
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.201691428571428e-06,
+      "loss": 1.0809,
+      "step": 33132
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.201422857142857e-06,
+      "loss": 1.149,
+      "step": 33133
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2011542857142855e-06,
+      "loss": 0.8378,
+      "step": 33134
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2008857142857145e-06,
+      "loss": 1.2606,
+      "step": 33135
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2006171428571435e-06,
+      "loss": 1.7796,
+      "step": 33136
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.200348571428572e-06,
+      "loss": 1.088,
+      "step": 33137
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.200080000000001e-06,
+      "loss": 1.4426,
+      "step": 33138
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.199811428571429e-06,
+      "loss": 1.2267,
+      "step": 33139
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.199542857142857e-06,
+      "loss": 1.1878,
+      "step": 33140
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.199274285714286e-06,
+      "loss": 1.2642,
+      "step": 33141
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.199005714285714e-06,
+      "loss": 1.5706,
+      "step": 33142
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.198737142857143e-06,
+      "loss": 0.8847,
+      "step": 33143
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.1984685714285715e-06,
+      "loss": 1.2305,
+      "step": 33144
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.1982e-06,
+      "loss": 1.4981,
+      "step": 33145
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.197931428571429e-06,
+      "loss": 1.2691,
+      "step": 33146
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.197662857142857e-06,
+      "loss": 1.423,
+      "step": 33147
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.197394285714286e-06,
+      "loss": 0.9612,
+      "step": 33148
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.197125714285715e-06,
+      "loss": 1.7864,
+      "step": 33149
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.196857142857143e-06,
+      "loss": 0.7143,
+      "step": 33150
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.196588571428572e-06,
+      "loss": 0.9044,
+      "step": 33151
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.19632e-06,
+      "loss": 1.1262,
+      "step": 33152
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.196051428571429e-06,
+      "loss": 1.1721,
+      "step": 33153
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.1957828571428574e-06,
+      "loss": 1.006,
+      "step": 33154
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.195514285714286e-06,
+      "loss": 1.2425,
+      "step": 33155
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.195245714285715e-06,
+      "loss": 1.0703,
+      "step": 33156
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.194977142857143e-06,
+      "loss": 1.374,
+      "step": 33157
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.194708571428572e-06,
+      "loss": 0.9791,
+      "step": 33158
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.19444e-06,
+      "loss": 1.6056,
+      "step": 33159
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.194171428571428e-06,
+      "loss": 0.9774,
+      "step": 33160
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.193902857142857e-06,
+      "loss": 1.3553,
+      "step": 33161
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.193634285714286e-06,
+      "loss": 1.4178,
+      "step": 33162
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.193365714285714e-06,
+      "loss": 0.9484,
+      "step": 33163
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1930971428571434e-06,
+      "loss": 1.1529,
+      "step": 33164
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.192828571428572e-06,
+      "loss": 1.1869,
+      "step": 33165
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.192560000000001e-06,
+      "loss": 0.8162,
+      "step": 33166
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.192291428571429e-06,
+      "loss": 0.9309,
+      "step": 33167
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.192022857142857e-06,
+      "loss": 0.855,
+      "step": 33168
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.191754285714286e-06,
+      "loss": 1.1486,
+      "step": 33169
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.191485714285714e-06,
+      "loss": 1.1194,
+      "step": 33170
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.191217142857143e-06,
+      "loss": 1.3316,
+      "step": 33171
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.190948571428571e-06,
+      "loss": 1.2192,
+      "step": 33172
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1906799999999996e-06,
+      "loss": 0.7677,
+      "step": 33173
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.190411428571429e-06,
+      "loss": 1.1416,
+      "step": 33174
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.190142857142858e-06,
+      "loss": 1.5787,
+      "step": 33175
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.189874285714287e-06,
+      "loss": 1.2737,
+      "step": 33176
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.189605714285715e-06,
+      "loss": 1.4356,
+      "step": 33177
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.189337142857143e-06,
+      "loss": 1.2883,
+      "step": 33178
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.189068571428572e-06,
+      "loss": 1.0868,
+      "step": 33179
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1888e-06,
+      "loss": 1.0484,
+      "step": 33180
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.188531428571429e-06,
+      "loss": 1.417,
+      "step": 33181
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.188262857142857e-06,
+      "loss": 1.2373,
+      "step": 33182
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1879942857142856e-06,
+      "loss": 0.9647,
+      "step": 33183
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.187725714285715e-06,
+      "loss": 1.3882,
+      "step": 33184
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.187457142857143e-06,
+      "loss": 1.2834,
+      "step": 33185
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.187188571428571e-06,
+      "loss": 1.2503,
+      "step": 33186
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.18692e-06,
+      "loss": 1.139,
+      "step": 33187
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.186651428571429e-06,
+      "loss": 1.6464,
+      "step": 33188
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.186382857142858e-06,
+      "loss": 0.8566,
+      "step": 33189
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.186114285714286e-06,
+      "loss": 1.4013,
+      "step": 33190
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.185845714285714e-06,
+      "loss": 1.1582,
+      "step": 33191
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.185577142857143e-06,
+      "loss": 0.8672,
+      "step": 33192
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1853085714285716e-06,
+      "loss": 1.2548,
+      "step": 33193
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.185040000000001e-06,
+      "loss": 1.1821,
+      "step": 33194
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.184771428571429e-06,
+      "loss": 0.7982,
+      "step": 33195
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.184502857142857e-06,
+      "loss": 1.2597,
+      "step": 33196
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.184234285714286e-06,
+      "loss": 1.0896,
+      "step": 33197
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.183965714285714e-06,
+      "loss": 1.1662,
+      "step": 33198
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.183697142857143e-06,
+      "loss": 1.2547,
+      "step": 33199
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.183428571428571e-06,
+      "loss": 1.3879,
+      "step": 33200
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.18316e-06,
+      "loss": 1.4019,
+      "step": 33201
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.182891428571429e-06,
+      "loss": 1.187,
+      "step": 33202
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1826228571428576e-06,
+      "loss": 1.5337,
+      "step": 33203
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.182354285714286e-06,
+      "loss": 1.0722,
+      "step": 33204
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.182085714285715e-06,
+      "loss": 1.1293,
+      "step": 33205
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.181817142857143e-06,
+      "loss": 1.1378,
+      "step": 33206
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.181548571428572e-06,
+      "loss": 0.9733,
+      "step": 33207
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.18128e-06,
+      "loss": 1.0995,
+      "step": 33208
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.181011428571428e-06,
+      "loss": 1.4117,
+      "step": 33209
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.180742857142857e-06,
+      "loss": 0.8669,
+      "step": 33210
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1804742857142855e-06,
+      "loss": 0.9749,
+      "step": 33211
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1802057142857145e-06,
+      "loss": 0.8766,
+      "step": 33212
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.179937142857143e-06,
+      "loss": 1.159,
+      "step": 33213
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.179668571428571e-06,
+      "loss": 1.0503,
+      "step": 33214
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.179400000000001e-06,
+      "loss": 1.2092,
+      "step": 33215
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.179131428571429e-06,
+      "loss": 0.9188,
+      "step": 33216
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.178862857142858e-06,
+      "loss": 1.1541,
+      "step": 33217
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.178594285714286e-06,
+      "loss": 0.8423,
+      "step": 33218
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.178325714285714e-06,
+      "loss": 1.0386,
+      "step": 33219
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.178057142857143e-06,
+      "loss": 1.2492,
+      "step": 33220
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1777885714285715e-06,
+      "loss": 1.0375,
+      "step": 33221
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1775200000000005e-06,
+      "loss": 0.9289,
+      "step": 33222
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.177251428571429e-06,
+      "loss": 1.559,
+      "step": 33223
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.176982857142857e-06,
+      "loss": 0.9671,
+      "step": 33224
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.176714285714286e-06,
+      "loss": 1.1088,
+      "step": 33225
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.176445714285714e-06,
+      "loss": 1.0845,
+      "step": 33226
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.176177142857142e-06,
+      "loss": 0.8201,
+      "step": 33227
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.175908571428572e-06,
+      "loss": 1.5975,
+      "step": 33228
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.17564e-06,
+      "loss": 1.282,
+      "step": 33229
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.175371428571429e-06,
+      "loss": 1.2414,
+      "step": 33230
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1751028571428575e-06,
+      "loss": 0.9832,
+      "step": 33231
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.174834285714286e-06,
+      "loss": 0.6801,
+      "step": 33232
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.174565714285715e-06,
+      "loss": 1.4333,
+      "step": 33233
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.174297142857143e-06,
+      "loss": 1.1366,
+      "step": 33234
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.174028571428572e-06,
+      "loss": 0.7904,
+      "step": 33235
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.17376e-06,
+      "loss": 1.03,
+      "step": 33236
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.173491428571428e-06,
+      "loss": 0.9812,
+      "step": 33237
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.173222857142857e-06,
+      "loss": 1.0599,
+      "step": 33238
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1729542857142855e-06,
+      "loss": 0.6445,
+      "step": 33239
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1726857142857145e-06,
+      "loss": 0.9648,
+      "step": 33240
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1724171428571435e-06,
+      "loss": 1.1951,
+      "step": 33241
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.172148571428572e-06,
+      "loss": 1.0942,
+      "step": 33242
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.171880000000001e-06,
+      "loss": 1.3047,
+      "step": 33243
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.171611428571429e-06,
+      "loss": 1.0365,
+      "step": 33244
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.171342857142858e-06,
+      "loss": 1.0528,
+      "step": 33245
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.171074285714286e-06,
+      "loss": 1.2052,
+      "step": 33246
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.170805714285714e-06,
+      "loss": 1.4449,
+      "step": 33247
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.170537142857143e-06,
+      "loss": 1.2637,
+      "step": 33248
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1702685714285715e-06,
+      "loss": 1.1036,
+      "step": 33249
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.17e-06,
+      "loss": 0.9208,
+      "step": 33250
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.169731428571429e-06,
+      "loss": 1.145,
+      "step": 33251
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.169462857142857e-06,
+      "loss": 0.6136,
+      "step": 33252
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.169194285714286e-06,
+      "loss": 0.8837,
+      "step": 33253
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.168925714285715e-06,
+      "loss": 1.534,
+      "step": 33254
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.168657142857143e-06,
+      "loss": 1.1127,
+      "step": 33255
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.168388571428572e-06,
+      "loss": 1.0146,
+      "step": 33256
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.16812e-06,
+      "loss": 0.8343,
+      "step": 33257
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.167851428571429e-06,
+      "loss": 0.6985,
+      "step": 33258
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1675828571428574e-06,
+      "loss": 1.344,
+      "step": 33259
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.167314285714286e-06,
+      "loss": 0.8162,
+      "step": 33260
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.167045714285715e-06,
+      "loss": 1.1956,
+      "step": 33261
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.166777142857143e-06,
+      "loss": 1.1031,
+      "step": 33262
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.166508571428572e-06,
+      "loss": 0.9016,
+      "step": 33263
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.16624e-06,
+      "loss": 0.7873,
+      "step": 33264
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.165971428571428e-06,
+      "loss": 1.1336,
+      "step": 33265
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.165702857142857e-06,
+      "loss": 1.1813,
+      "step": 33266
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.165434285714285e-06,
+      "loss": 0.8359,
+      "step": 33267
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.165165714285715e-06,
+      "loss": 1.4043,
+      "step": 33268
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1648971428571434e-06,
+      "loss": 0.8554,
+      "step": 33269
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.164628571428572e-06,
+      "loss": 0.7901,
+      "step": 33270
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.164360000000001e-06,
+      "loss": 1.3892,
+      "step": 33271
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.164091428571429e-06,
+      "loss": 1.1454,
+      "step": 33272
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.163822857142857e-06,
+      "loss": 1.1347,
+      "step": 33273
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.163554285714286e-06,
+      "loss": 1.1616,
+      "step": 33274
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.163285714285714e-06,
+      "loss": 1.2528,
+      "step": 33275
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.163017142857143e-06,
+      "loss": 1.1721,
+      "step": 33276
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.162748571428571e-06,
+      "loss": 0.8791,
+      "step": 33277
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1624799999999996e-06,
+      "loss": 1.4171,
+      "step": 33278
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.162211428571429e-06,
+      "loss": 1.0868,
+      "step": 33279
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.161942857142857e-06,
+      "loss": 1.0378,
+      "step": 33280
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.161674285714287e-06,
+      "loss": 1.1954,
+      "step": 33281
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.161405714285715e-06,
+      "loss": 1.1467,
+      "step": 33282
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.161137142857143e-06,
+      "loss": 1.0851,
+      "step": 33283
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.160868571428572e-06,
+      "loss": 1.271,
+      "step": 33284
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1606e-06,
+      "loss": 1.1241,
+      "step": 33285
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.160331428571429e-06,
+      "loss": 0.882,
+      "step": 33286
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.160062857142857e-06,
+      "loss": 0.5371,
+      "step": 33287
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1597942857142856e-06,
+      "loss": 1.0799,
+      "step": 33288
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.159525714285715e-06,
+      "loss": 1.3708,
+      "step": 33289
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.159257142857143e-06,
+      "loss": 1.5523,
+      "step": 33290
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.158988571428571e-06,
+      "loss": 1.4665,
+      "step": 33291
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.15872e-06,
+      "loss": 1.264,
+      "step": 33292
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.158451428571428e-06,
+      "loss": 1.1499,
+      "step": 33293
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.158182857142858e-06,
+      "loss": 0.9011,
+      "step": 33294
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.157914285714286e-06,
+      "loss": 1.2776,
+      "step": 33295
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.157645714285714e-06,
+      "loss": 1.1614,
+      "step": 33296
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.157377142857143e-06,
+      "loss": 1.2107,
+      "step": 33297
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1571085714285716e-06,
+      "loss": 0.928,
+      "step": 33298
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.156840000000001e-06,
+      "loss": 1.2312,
+      "step": 33299
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.156571428571429e-06,
+      "loss": 1.4292,
+      "step": 33300
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.156302857142857e-06,
+      "loss": 1.1155,
+      "step": 33301
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.156034285714286e-06,
+      "loss": 1.3291,
+      "step": 33302
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.155765714285714e-06,
+      "loss": 1.0939,
+      "step": 33303
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.155497142857143e-06,
+      "loss": 1.2903,
+      "step": 33304
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.155228571428571e-06,
+      "loss": 0.9988,
+      "step": 33305
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1549599999999995e-06,
+      "loss": 1.5187,
+      "step": 33306
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.154691428571429e-06,
+      "loss": 0.9736,
+      "step": 33307
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1544228571428576e-06,
+      "loss": 1.1918,
+      "step": 33308
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.154154285714287e-06,
+      "loss": 0.8547,
+      "step": 33309
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.153885714285715e-06,
+      "loss": 0.9819,
+      "step": 33310
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.153617142857143e-06,
+      "loss": 1.114,
+      "step": 33311
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.153348571428572e-06,
+      "loss": 0.9415,
+      "step": 33312
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.15308e-06,
+      "loss": 1.259,
+      "step": 33313
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.152811428571428e-06,
+      "loss": 1.6047,
+      "step": 33314
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.152542857142857e-06,
+      "loss": 1.0899,
+      "step": 33315
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1522742857142855e-06,
+      "loss": 1.3024,
+      "step": 33316
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1520057142857145e-06,
+      "loss": 0.7312,
+      "step": 33317
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.151737142857143e-06,
+      "loss": 1.3787,
+      "step": 33318
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.151468571428571e-06,
+      "loss": 1.2653,
+      "step": 33319
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1512e-06,
+      "loss": 1.0333,
+      "step": 33320
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.150931428571429e-06,
+      "loss": 0.8303,
+      "step": 33321
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.150662857142858e-06,
+      "loss": 1.0856,
+      "step": 33322
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.150394285714286e-06,
+      "loss": 0.7912,
+      "step": 33323
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.150125714285714e-06,
+      "loss": 0.9245,
+      "step": 33324
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.149857142857143e-06,
+      "loss": 1.3696,
+      "step": 33325
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1495885714285715e-06,
+      "loss": 1.4862,
+      "step": 33326
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1493200000000005e-06,
+      "loss": 1.3239,
+      "step": 33327
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.149051428571429e-06,
+      "loss": 1.044,
+      "step": 33328
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.148782857142857e-06,
+      "loss": 0.7029,
+      "step": 33329
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.148514285714286e-06,
+      "loss": 1.1268,
+      "step": 33330
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.148245714285714e-06,
+      "loss": 0.8383,
+      "step": 33331
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.147977142857143e-06,
+      "loss": 1.3503,
+      "step": 33332
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.147708571428571e-06,
+      "loss": 0.8026,
+      "step": 33333
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.14744e-06,
+      "loss": 1.1213,
+      "step": 33334
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.147171428571429e-06,
+      "loss": 0.8759,
+      "step": 33335
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1469028571428575e-06,
+      "loss": 1.3185,
+      "step": 33336
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.146634285714286e-06,
+      "loss": 1.2892,
+      "step": 33337
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.146365714285715e-06,
+      "loss": 1.5607,
+      "step": 33338
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.146097142857143e-06,
+      "loss": 1.023,
+      "step": 33339
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.145828571428572e-06,
+      "loss": 0.6541,
+      "step": 33340
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.14556e-06,
+      "loss": 0.6442,
+      "step": 33341
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.145291428571428e-06,
+      "loss": 0.8013,
+      "step": 33342
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.145022857142857e-06,
+      "loss": 0.9579,
+      "step": 33343
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1447542857142855e-06,
+      "loss": 1.0407,
+      "step": 33344
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1444857142857145e-06,
+      "loss": 0.9471,
+      "step": 33345
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.144217142857143e-06,
+      "loss": 1.1548,
+      "step": 33346
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.143948571428572e-06,
+      "loss": 1.233,
+      "step": 33347
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.143680000000001e-06,
+      "loss": 1.2256,
+      "step": 33348
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.143411428571429e-06,
+      "loss": 1.174,
+      "step": 33349
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.143142857142858e-06,
+      "loss": 1.0933,
+      "step": 33350
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.142874285714286e-06,
+      "loss": 0.8413,
+      "step": 33351
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.142605714285714e-06,
+      "loss": 1.342,
+      "step": 33352
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.142337142857143e-06,
+      "loss": 1.0209,
+      "step": 33353
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1420685714285715e-06,
+      "loss": 1.0526,
+      "step": 33354
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1418e-06,
+      "loss": 1.2499,
+      "step": 33355
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.141531428571429e-06,
+      "loss": 1.5631,
+      "step": 33356
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.141262857142857e-06,
+      "loss": 1.0946,
+      "step": 33357
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.140994285714286e-06,
+      "loss": 1.0553,
+      "step": 33358
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.140725714285714e-06,
+      "loss": 1.2608,
+      "step": 33359
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.140457142857143e-06,
+      "loss": 1.517,
+      "step": 33360
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.140188571428572e-06,
+      "loss": 1.2992,
+      "step": 33361
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.13992e-06,
+      "loss": 0.9007,
+      "step": 33362
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.139651428571429e-06,
+      "loss": 1.1436,
+      "step": 33363
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1393828571428574e-06,
+      "loss": 1.1589,
+      "step": 33364
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.139114285714286e-06,
+      "loss": 0.8865,
+      "step": 33365
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.138845714285715e-06,
+      "loss": 1.1285,
+      "step": 33366
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.138577142857143e-06,
+      "loss": 1.3505,
+      "step": 33367
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.138308571428572e-06,
+      "loss": 0.9267,
+      "step": 33368
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.13804e-06,
+      "loss": 1.8598,
+      "step": 33369
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.137771428571428e-06,
+      "loss": 0.8139,
+      "step": 33370
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.137502857142857e-06,
+      "loss": 1.1217,
+      "step": 33371
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.137234285714285e-06,
+      "loss": 1.2963,
+      "step": 33372
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.136965714285714e-06,
+      "loss": 0.9713,
+      "step": 33373
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1366971428571434e-06,
+      "loss": 1.2067,
+      "step": 33374
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.136428571428572e-06,
+      "loss": 0.7399,
+      "step": 33375
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.136160000000001e-06,
+      "loss": 1.0353,
+      "step": 33376
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.135891428571429e-06,
+      "loss": 1.0077,
+      "step": 33377
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.135622857142857e-06,
+      "loss": 1.4015,
+      "step": 33378
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.135354285714286e-06,
+      "loss": 1.1245,
+      "step": 33379
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.135085714285714e-06,
+      "loss": 1.1664,
+      "step": 33380
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.134817142857143e-06,
+      "loss": 1.3737,
+      "step": 33381
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.134548571428571e-06,
+      "loss": 1.2005,
+      "step": 33382
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1342799999999996e-06,
+      "loss": 0.9638,
+      "step": 33383
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.134011428571429e-06,
+      "loss": 1.273,
+      "step": 33384
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.133742857142857e-06,
+      "loss": 1.4271,
+      "step": 33385
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.133474285714286e-06,
+      "loss": 1.6175,
+      "step": 33386
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.133205714285715e-06,
+      "loss": 1.2382,
+      "step": 33387
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.132937142857143e-06,
+      "loss": 0.859,
+      "step": 33388
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.132668571428572e-06,
+      "loss": 0.8148,
+      "step": 33389
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1324e-06,
+      "loss": 0.9983,
+      "step": 33390
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.132131428571429e-06,
+      "loss": 0.8339,
+      "step": 33391
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.131862857142857e-06,
+      "loss": 1.0757,
+      "step": 33392
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1315942857142856e-06,
+      "loss": 1.1048,
+      "step": 33393
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.131325714285715e-06,
+      "loss": 0.8016,
+      "step": 33394
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.131057142857143e-06,
+      "loss": 0.8756,
+      "step": 33395
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.130788571428572e-06,
+      "loss": 0.5458,
+      "step": 33396
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.13052e-06,
+      "loss": 1.1484,
+      "step": 33397
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.130251428571428e-06,
+      "loss": 0.8273,
+      "step": 33398
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.129982857142857e-06,
+      "loss": 1.0779,
+      "step": 33399
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.129714285714286e-06,
+      "loss": 1.5437,
+      "step": 33400
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.129445714285714e-06,
+      "loss": 1.2363,
+      "step": 33401
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.129177142857143e-06,
+      "loss": 1.1004,
+      "step": 33402
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1289085714285716e-06,
+      "loss": 0.7792,
+      "step": 33403
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.128640000000001e-06,
+      "loss": 0.8754,
+      "step": 33404
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.128371428571429e-06,
+      "loss": 1.5112,
+      "step": 33405
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.128102857142857e-06,
+      "loss": 1.1859,
+      "step": 33406
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.127834285714286e-06,
+      "loss": 0.9419,
+      "step": 33407
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.127565714285714e-06,
+      "loss": 1.4731,
+      "step": 33408
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.127297142857143e-06,
+      "loss": 1.1057,
+      "step": 33409
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.127028571428571e-06,
+      "loss": 1.1864,
+      "step": 33410
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1267599999999995e-06,
+      "loss": 1.054,
+      "step": 33411
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1264914285714285e-06,
+      "loss": 1.272,
+      "step": 33412
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1262228571428576e-06,
+      "loss": 1.1314,
+      "step": 33413
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.125954285714287e-06,
+      "loss": 0.9312,
+      "step": 33414
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.125685714285715e-06,
+      "loss": 1.3376,
+      "step": 33415
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.125417142857143e-06,
+      "loss": 1.1605,
+      "step": 33416
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.125148571428572e-06,
+      "loss": 1.1133,
+      "step": 33417
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.12488e-06,
+      "loss": 1.0357,
+      "step": 33418
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.124611428571429e-06,
+      "loss": 0.7447,
+      "step": 33419
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.124342857142857e-06,
+      "loss": 1.0835,
+      "step": 33420
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1240742857142855e-06,
+      "loss": 1.0952,
+      "step": 33421
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1238057142857145e-06,
+      "loss": 0.8784,
+      "step": 33422
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.123537142857143e-06,
+      "loss": 0.7173,
+      "step": 33423
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.123268571428571e-06,
+      "loss": 0.8552,
+      "step": 33424
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.123e-06,
+      "loss": 1.3146,
+      "step": 33425
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.122731428571429e-06,
+      "loss": 1.1952,
+      "step": 33426
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.122462857142858e-06,
+      "loss": 0.8851,
+      "step": 33427
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.122194285714286e-06,
+      "loss": 1.5622,
+      "step": 33428
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.121925714285714e-06,
+      "loss": 0.9461,
+      "step": 33429
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.121657142857143e-06,
+      "loss": 1.3201,
+      "step": 33430
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1213885714285715e-06,
+      "loss": 0.9875,
+      "step": 33431
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1211200000000005e-06,
+      "loss": 0.9815,
+      "step": 33432
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.120851428571429e-06,
+      "loss": 0.8991,
+      "step": 33433
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.120582857142857e-06,
+      "loss": 0.9187,
+      "step": 33434
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.120314285714286e-06,
+      "loss": 1.3628,
+      "step": 33435
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.120045714285714e-06,
+      "loss": 1.3298,
+      "step": 33436
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.119777142857143e-06,
+      "loss": 1.5407,
+      "step": 33437
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.119508571428571e-06,
+      "loss": 1.2092,
+      "step": 33438
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.11924e-06,
+      "loss": 1.2036,
+      "step": 33439
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.118971428571429e-06,
+      "loss": 1.3882,
+      "step": 33440
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1187028571428575e-06,
+      "loss": 1.1109,
+      "step": 33441
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.118434285714286e-06,
+      "loss": 0.7,
+      "step": 33442
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.118165714285715e-06,
+      "loss": 0.9467,
+      "step": 33443
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.117897142857143e-06,
+      "loss": 1.6064,
+      "step": 33444
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.117628571428572e-06,
+      "loss": 1.2258,
+      "step": 33445
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.11736e-06,
+      "loss": 1.108,
+      "step": 33446
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.117091428571428e-06,
+      "loss": 1.2065,
+      "step": 33447
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.116822857142857e-06,
+      "loss": 1.1816,
+      "step": 33448
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1165542857142855e-06,
+      "loss": 1.0782,
+      "step": 33449
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1162857142857145e-06,
+      "loss": 1.2387,
+      "step": 33450
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.116017142857143e-06,
+      "loss": 0.7885,
+      "step": 33451
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.115748571428572e-06,
+      "loss": 1.5904,
+      "step": 33452
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.115480000000001e-06,
+      "loss": 0.8969,
+      "step": 33453
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.115211428571429e-06,
+      "loss": 1.4484,
+      "step": 33454
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.114942857142858e-06,
+      "loss": 1.2997,
+      "step": 33455
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.114674285714286e-06,
+      "loss": 0.9409,
+      "step": 33456
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.114405714285714e-06,
+      "loss": 1.1435,
+      "step": 33457
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.114137142857143e-06,
+      "loss": 0.8609,
+      "step": 33458
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1138685714285715e-06,
+      "loss": 1.2355,
+      "step": 33459
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1136000000000005e-06,
+      "loss": 1.0905,
+      "step": 33460
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.113331428571429e-06,
+      "loss": 1.4806,
+      "step": 33461
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.113062857142857e-06,
+      "loss": 1.3576,
+      "step": 33462
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.112794285714286e-06,
+      "loss": 0.9125,
+      "step": 33463
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.112525714285714e-06,
+      "loss": 0.9511,
+      "step": 33464
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.112257142857143e-06,
+      "loss": 1.1133,
+      "step": 33465
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.111988571428572e-06,
+      "loss": 1.0121,
+      "step": 33466
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.11172e-06,
+      "loss": 1.1163,
+      "step": 33467
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.111451428571429e-06,
+      "loss": 1.1601,
+      "step": 33468
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1111828571428574e-06,
+      "loss": 1.0335,
+      "step": 33469
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.110914285714286e-06,
+      "loss": 1.4215,
+      "step": 33470
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.110645714285715e-06,
+      "loss": 0.8591,
+      "step": 33471
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.110377142857143e-06,
+      "loss": 1.4581,
+      "step": 33472
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.110108571428572e-06,
+      "loss": 0.819,
+      "step": 33473
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.10984e-06,
+      "loss": 0.9155,
+      "step": 33474
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.109571428571428e-06,
+      "loss": 1.4782,
+      "step": 33475
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.109302857142857e-06,
+      "loss": 0.9705,
+      "step": 33476
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.109034285714285e-06,
+      "loss": 1.6095,
+      "step": 33477
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.108765714285714e-06,
+      "loss": 1.1935,
+      "step": 33478
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1084971428571434e-06,
+      "loss": 1.1335,
+      "step": 33479
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.108228571428572e-06,
+      "loss": 1.3677,
+      "step": 33480
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.107960000000001e-06,
+      "loss": 1.0548,
+      "step": 33481
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.107691428571429e-06,
+      "loss": 1.4822,
+      "step": 33482
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.107422857142858e-06,
+      "loss": 0.8397,
+      "step": 33483
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.107154285714286e-06,
+      "loss": 0.8079,
+      "step": 33484
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.106885714285714e-06,
+      "loss": 1.0723,
+      "step": 33485
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.106617142857143e-06,
+      "loss": 1.5349,
+      "step": 33486
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.106348571428571e-06,
+      "loss": 1.1473,
+      "step": 33487
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1060799999999996e-06,
+      "loss": 1.0556,
+      "step": 33488
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.105811428571429e-06,
+      "loss": 1.0761,
+      "step": 33489
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.105542857142857e-06,
+      "loss": 1.4393,
+      "step": 33490
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.105274285714286e-06,
+      "loss": 1.5853,
+      "step": 33491
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.105005714285715e-06,
+      "loss": 1.2475,
+      "step": 33492
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.104737142857143e-06,
+      "loss": 0.7827,
+      "step": 33493
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.104468571428572e-06,
+      "loss": 0.9991,
+      "step": 33494
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1042e-06,
+      "loss": 1.2606,
+      "step": 33495
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.103931428571429e-06,
+      "loss": 0.9286,
+      "step": 33496
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.103662857142857e-06,
+      "loss": 1.0048,
+      "step": 33497
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1033942857142856e-06,
+      "loss": 1.2038,
+      "step": 33498
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.103125714285715e-06,
+      "loss": 1.1629,
+      "step": 33499
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.102857142857143e-06,
+      "loss": 1.0929,
+      "step": 33500
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.102588571428572e-06,
+      "loss": 1.1578,
+      "step": 33501
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.10232e-06,
+      "loss": 1.3788,
+      "step": 33502
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.102051428571428e-06,
+      "loss": 1.1632,
+      "step": 33503
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.101782857142857e-06,
+      "loss": 1.2581,
+      "step": 33504
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.101514285714286e-06,
+      "loss": 1.1326,
+      "step": 33505
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.101245714285715e-06,
+      "loss": 0.9025,
+      "step": 33506
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.100977142857143e-06,
+      "loss": 0.8219,
+      "step": 33507
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.1007085714285716e-06,
+      "loss": 1.0807,
+      "step": 33508
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.100440000000001e-06,
+      "loss": 1.3416,
+      "step": 33509
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.100171428571429e-06,
+      "loss": 1.1667,
+      "step": 33510
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.099902857142857e-06,
+      "loss": 1.1836,
+      "step": 33511
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.099634285714286e-06,
+      "loss": 1.1926,
+      "step": 33512
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.099365714285714e-06,
+      "loss": 1.6451,
+      "step": 33513
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.099097142857143e-06,
+      "loss": 1.6025,
+      "step": 33514
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.098828571428571e-06,
+      "loss": 1.5767,
+      "step": 33515
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0985599999999995e-06,
+      "loss": 1.3232,
+      "step": 33516
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0982914285714285e-06,
+      "loss": 1.2796,
+      "step": 33517
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0980228571428576e-06,
+      "loss": 0.9077,
+      "step": 33518
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.097754285714287e-06,
+      "loss": 1.1841,
+      "step": 33519
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.097485714285715e-06,
+      "loss": 0.903,
+      "step": 33520
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.097217142857143e-06,
+      "loss": 1.0554,
+      "step": 33521
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.096948571428572e-06,
+      "loss": 1.1574,
+      "step": 33522
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.09668e-06,
+      "loss": 1.3576,
+      "step": 33523
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.096411428571429e-06,
+      "loss": 1.2396,
+      "step": 33524
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.096142857142857e-06,
+      "loss": 1.0854,
+      "step": 33525
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0958742857142855e-06,
+      "loss": 1.0526,
+      "step": 33526
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0956057142857145e-06,
+      "loss": 0.7673,
+      "step": 33527
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.095337142857143e-06,
+      "loss": 1.1999,
+      "step": 33528
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.095068571428571e-06,
+      "loss": 0.9446,
+      "step": 33529
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0948e-06,
+      "loss": 1.1362,
+      "step": 33530
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.094531428571429e-06,
+      "loss": 1.4346,
+      "step": 33531
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.094262857142858e-06,
+      "loss": 1.1357,
+      "step": 33532
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.093994285714286e-06,
+      "loss": 1.2487,
+      "step": 33533
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.093725714285714e-06,
+      "loss": 0.7494,
+      "step": 33534
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.093457142857143e-06,
+      "loss": 0.9218,
+      "step": 33535
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0931885714285715e-06,
+      "loss": 1.3851,
+      "step": 33536
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0929200000000005e-06,
+      "loss": 1.5499,
+      "step": 33537
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.092651428571429e-06,
+      "loss": 0.9046,
+      "step": 33538
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.092382857142857e-06,
+      "loss": 1.2694,
+      "step": 33539
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.092114285714286e-06,
+      "loss": 0.8934,
+      "step": 33540
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.091845714285714e-06,
+      "loss": 1.5483,
+      "step": 33541
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.091577142857143e-06,
+      "loss": 1.0975,
+      "step": 33542
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.091308571428571e-06,
+      "loss": 1.7192,
+      "step": 33543
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.09104e-06,
+      "loss": 1.6215,
+      "step": 33544
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.090771428571429e-06,
+      "loss": 1.196,
+      "step": 33545
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0905028571428575e-06,
+      "loss": 1.1957,
+      "step": 33546
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0902342857142865e-06,
+      "loss": 1.3833,
+      "step": 33547
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.089965714285715e-06,
+      "loss": 1.2892,
+      "step": 33548
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.089697142857143e-06,
+      "loss": 0.8228,
+      "step": 33549
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.089428571428572e-06,
+      "loss": 1.1469,
+      "step": 33550
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.08916e-06,
+      "loss": 1.1863,
+      "step": 33551
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.088891428571428e-06,
+      "loss": 1.2892,
+      "step": 33552
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.088622857142857e-06,
+      "loss": 1.1893,
+      "step": 33553
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0883542857142855e-06,
+      "loss": 1.1343,
+      "step": 33554
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0880857142857145e-06,
+      "loss": 1.1492,
+      "step": 33555
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.087817142857143e-06,
+      "loss": 1.2657,
+      "step": 33556
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.087548571428572e-06,
+      "loss": 1.4189,
+      "step": 33557
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.087280000000001e-06,
+      "loss": 1.0795,
+      "step": 33558
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.087011428571429e-06,
+      "loss": 0.9389,
+      "step": 33559
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.086742857142858e-06,
+      "loss": 1.452,
+      "step": 33560
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.086474285714286e-06,
+      "loss": 1.7579,
+      "step": 33561
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.086205714285714e-06,
+      "loss": 1.0398,
+      "step": 33562
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.085937142857143e-06,
+      "loss": 1.2844,
+      "step": 33563
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0856685714285714e-06,
+      "loss": 0.9155,
+      "step": 33564
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0854000000000005e-06,
+      "loss": 1.3037,
+      "step": 33565
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.085131428571429e-06,
+      "loss": 1.0222,
+      "step": 33566
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.084862857142857e-06,
+      "loss": 1.788,
+      "step": 33567
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.084594285714286e-06,
+      "loss": 1.3402,
+      "step": 33568
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.084325714285714e-06,
+      "loss": 0.9345,
+      "step": 33569
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.084057142857143e-06,
+      "loss": 1.587,
+      "step": 33570
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.083788571428572e-06,
+      "loss": 1.4509,
+      "step": 33571
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.08352e-06,
+      "loss": 1.412,
+      "step": 33572
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.083251428571429e-06,
+      "loss": 1.3814,
+      "step": 33573
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0829828571428574e-06,
+      "loss": 0.9137,
+      "step": 33574
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.082714285714286e-06,
+      "loss": 1.3204,
+      "step": 33575
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.082445714285715e-06,
+      "loss": 1.1062,
+      "step": 33576
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.082177142857143e-06,
+      "loss": 1.0593,
+      "step": 33577
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.081908571428572e-06,
+      "loss": 1.0466,
+      "step": 33578
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.08164e-06,
+      "loss": 1.0913,
+      "step": 33579
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.081371428571428e-06,
+      "loss": 0.7797,
+      "step": 33580
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.081102857142857e-06,
+      "loss": 1.0014,
+      "step": 33581
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.080834285714285e-06,
+      "loss": 0.9414,
+      "step": 33582
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.080565714285714e-06,
+      "loss": 1.0317,
+      "step": 33583
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0802971428571434e-06,
+      "loss": 0.8195,
+      "step": 33584
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.080028571428572e-06,
+      "loss": 1.2146,
+      "step": 33585
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.079760000000001e-06,
+      "loss": 1.058,
+      "step": 33586
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.079491428571429e-06,
+      "loss": 1.0303,
+      "step": 33587
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.079222857142858e-06,
+      "loss": 1.1747,
+      "step": 33588
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.078954285714286e-06,
+      "loss": 1.1161,
+      "step": 33589
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.078685714285714e-06,
+      "loss": 1.2786,
+      "step": 33590
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.078417142857143e-06,
+      "loss": 0.8119,
+      "step": 33591
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.078148571428571e-06,
+      "loss": 1.3765,
+      "step": 33592
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.07788e-06,
+      "loss": 1.7263,
+      "step": 33593
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.077611428571429e-06,
+      "loss": 1.168,
+      "step": 33594
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.077342857142857e-06,
+      "loss": 1.2774,
+      "step": 33595
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.077074285714286e-06,
+      "loss": 1.0182,
+      "step": 33596
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.076805714285715e-06,
+      "loss": 1.3678,
+      "step": 33597
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.076537142857143e-06,
+      "loss": 1.0477,
+      "step": 33598
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.076268571428572e-06,
+      "loss": 0.9787,
+      "step": 33599
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.076e-06,
+      "loss": 1.4451,
+      "step": 33600
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.075731428571429e-06,
+      "loss": 0.9735,
+      "step": 33601
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.075462857142857e-06,
+      "loss": 1.1969,
+      "step": 33602
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0751942857142856e-06,
+      "loss": 1.2456,
+      "step": 33603
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.074925714285715e-06,
+      "loss": 1.2676,
+      "step": 33604
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.074657142857143e-06,
+      "loss": 1.2219,
+      "step": 33605
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.074388571428572e-06,
+      "loss": 1.1659,
+      "step": 33606
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.07412e-06,
+      "loss": 1.387,
+      "step": 33607
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.073851428571428e-06,
+      "loss": 1.0461,
+      "step": 33608
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.073582857142857e-06,
+      "loss": 0.7223,
+      "step": 33609
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.073314285714286e-06,
+      "loss": 1.1633,
+      "step": 33610
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.073045714285715e-06,
+      "loss": 1.1992,
+      "step": 33611
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.072777142857143e-06,
+      "loss": 1.0262,
+      "step": 33612
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0725085714285716e-06,
+      "loss": 1.0862,
+      "step": 33613
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.072240000000001e-06,
+      "loss": 1.2363,
+      "step": 33614
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.071971428571429e-06,
+      "loss": 1.1972,
+      "step": 33615
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.071702857142857e-06,
+      "loss": 0.8923,
+      "step": 33616
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.071434285714286e-06,
+      "loss": 1.153,
+      "step": 33617
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.071165714285714e-06,
+      "loss": 0.7311,
+      "step": 33618
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.070897142857143e-06,
+      "loss": 1.0887,
+      "step": 33619
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.070628571428571e-06,
+      "loss": 1.0492,
+      "step": 33620
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0703599999999995e-06,
+      "loss": 1.1245,
+      "step": 33621
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0700914285714285e-06,
+      "loss": 1.0607,
+      "step": 33622
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0698228571428576e-06,
+      "loss": 0.9613,
+      "step": 33623
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.069554285714287e-06,
+      "loss": 1.0739,
+      "step": 33624
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.069285714285715e-06,
+      "loss": 1.1503,
+      "step": 33625
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.069017142857143e-06,
+      "loss": 1.1291,
+      "step": 33626
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.068748571428572e-06,
+      "loss": 1.2178,
+      "step": 33627
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.06848e-06,
+      "loss": 0.933,
+      "step": 33628
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.068211428571429e-06,
+      "loss": 1.017,
+      "step": 33629
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.067942857142857e-06,
+      "loss": 0.9685,
+      "step": 33630
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0676742857142855e-06,
+      "loss": 1.2588,
+      "step": 33631
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0674057142857145e-06,
+      "loss": 1.1052,
+      "step": 33632
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.067137142857143e-06,
+      "loss": 1.2791,
+      "step": 33633
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.066868571428572e-06,
+      "loss": 0.847,
+      "step": 33634
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0666e-06,
+      "loss": 0.9787,
+      "step": 33635
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.066331428571429e-06,
+      "loss": 0.8483,
+      "step": 33636
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.066062857142858e-06,
+      "loss": 1.1139,
+      "step": 33637
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.065794285714286e-06,
+      "loss": 1.1526,
+      "step": 33638
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.065525714285714e-06,
+      "loss": 1.4394,
+      "step": 33639
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.065257142857143e-06,
+      "loss": 1.3421,
+      "step": 33640
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0649885714285715e-06,
+      "loss": 1.268,
+      "step": 33641
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0647200000000005e-06,
+      "loss": 0.8068,
+      "step": 33642
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.064451428571429e-06,
+      "loss": 1.2619,
+      "step": 33643
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.064182857142857e-06,
+      "loss": 1.5201,
+      "step": 33644
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.063914285714286e-06,
+      "loss": 1.0778,
+      "step": 33645
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.063645714285714e-06,
+      "loss": 0.9322,
+      "step": 33646
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.063377142857143e-06,
+      "loss": 1.3825,
+      "step": 33647
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.063108571428571e-06,
+      "loss": 1.2533,
+      "step": 33648
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.06284e-06,
+      "loss": 1.5701,
+      "step": 33649
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.062571428571429e-06,
+      "loss": 1.2397,
+      "step": 33650
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0623028571428575e-06,
+      "loss": 0.8164,
+      "step": 33651
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0620342857142865e-06,
+      "loss": 0.9875,
+      "step": 33652
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.061765714285715e-06,
+      "loss": 1.1669,
+      "step": 33653
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.061497142857143e-06,
+      "loss": 1.0241,
+      "step": 33654
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.061228571428572e-06,
+      "loss": 1.0992,
+      "step": 33655
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.06096e-06,
+      "loss": 0.7359,
+      "step": 33656
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.060691428571429e-06,
+      "loss": 1.1639,
+      "step": 33657
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.060422857142857e-06,
+      "loss": 0.9072,
+      "step": 33658
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0601542857142855e-06,
+      "loss": 0.8352,
+      "step": 33659
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0598857142857145e-06,
+      "loss": 0.7911,
+      "step": 33660
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.059617142857143e-06,
+      "loss": 1.0478,
+      "step": 33661
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.059348571428572e-06,
+      "loss": 1.0795,
+      "step": 33662
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.059080000000001e-06,
+      "loss": 1.0956,
+      "step": 33663
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.058811428571429e-06,
+      "loss": 1.3367,
+      "step": 33664
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.058542857142858e-06,
+      "loss": 0.6868,
+      "step": 33665
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.058274285714286e-06,
+      "loss": 0.7638,
+      "step": 33666
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.058005714285714e-06,
+      "loss": 1.2789,
+      "step": 33667
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.057737142857143e-06,
+      "loss": 1.2503,
+      "step": 33668
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0574685714285714e-06,
+      "loss": 0.9925,
+      "step": 33669
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0572000000000005e-06,
+      "loss": 0.7297,
+      "step": 33670
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.056931428571429e-06,
+      "loss": 1.3139,
+      "step": 33671
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.056662857142857e-06,
+      "loss": 0.8601,
+      "step": 33672
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.056394285714286e-06,
+      "loss": 0.9556,
+      "step": 33673
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.056125714285714e-06,
+      "loss": 0.9848,
+      "step": 33674
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.055857142857143e-06,
+      "loss": 1.0967,
+      "step": 33675
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.055588571428572e-06,
+      "loss": 1.0091,
+      "step": 33676
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.05532e-06,
+      "loss": 1.2804,
+      "step": 33677
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.055051428571429e-06,
+      "loss": 1.3583,
+      "step": 33678
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0547828571428574e-06,
+      "loss": 1.1607,
+      "step": 33679
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.0545142857142865e-06,
+      "loss": 1.1377,
+      "step": 33680
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.054245714285715e-06,
+      "loss": 1.5201,
+      "step": 33681
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.053977142857143e-06,
+      "loss": 0.904,
+      "step": 33682
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.053708571428572e-06,
+      "loss": 1.3784,
+      "step": 33683
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.05344e-06,
+      "loss": 1.2857,
+      "step": 33684
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.053171428571428e-06,
+      "loss": 0.9432,
+      "step": 33685
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.052902857142857e-06,
+      "loss": 0.7577,
+      "step": 33686
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.052634285714285e-06,
+      "loss": 1.4028,
+      "step": 33687
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.052365714285714e-06,
+      "loss": 1.5796,
+      "step": 33688
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0520971428571434e-06,
+      "loss": 1.1332,
+      "step": 33689
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.051828571428572e-06,
+      "loss": 1.1465,
+      "step": 33690
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.051560000000001e-06,
+      "loss": 0.716,
+      "step": 33691
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.051291428571429e-06,
+      "loss": 1.2089,
+      "step": 33692
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.051022857142858e-06,
+      "loss": 0.9812,
+      "step": 33693
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.050754285714286e-06,
+      "loss": 0.8243,
+      "step": 33694
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.050485714285714e-06,
+      "loss": 1.3513,
+      "step": 33695
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.050217142857143e-06,
+      "loss": 0.9742,
+      "step": 33696
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.049948571428571e-06,
+      "loss": 1.6065,
+      "step": 33697
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.04968e-06,
+      "loss": 1.3871,
+      "step": 33698
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.049411428571429e-06,
+      "loss": 1.0689,
+      "step": 33699
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.049142857142857e-06,
+      "loss": 1.4897,
+      "step": 33700
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.048874285714286e-06,
+      "loss": 0.8433,
+      "step": 33701
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.048605714285715e-06,
+      "loss": 0.9026,
+      "step": 33702
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.048337142857143e-06,
+      "loss": 1.0761,
+      "step": 33703
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.048068571428572e-06,
+      "loss": 1.0728,
+      "step": 33704
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0478e-06,
+      "loss": 1.1242,
+      "step": 33705
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.047531428571429e-06,
+      "loss": 1.1026,
+      "step": 33706
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.047262857142857e-06,
+      "loss": 1.0717,
+      "step": 33707
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0469942857142856e-06,
+      "loss": 1.0873,
+      "step": 33708
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.046725714285715e-06,
+      "loss": 1.2349,
+      "step": 33709
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.046457142857143e-06,
+      "loss": 1.8703,
+      "step": 33710
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.046188571428572e-06,
+      "loss": 1.4455,
+      "step": 33711
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.04592e-06,
+      "loss": 1.2613,
+      "step": 33712
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.045651428571428e-06,
+      "loss": 1.2688,
+      "step": 33713
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.045382857142857e-06,
+      "loss": 1.1677,
+      "step": 33714
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.045114285714286e-06,
+      "loss": 1.1281,
+      "step": 33715
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.044845714285715e-06,
+      "loss": 0.8922,
+      "step": 33716
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.044577142857143e-06,
+      "loss": 1.3732,
+      "step": 33717
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0443085714285716e-06,
+      "loss": 1.0884,
+      "step": 33718
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.044040000000001e-06,
+      "loss": 1.114,
+      "step": 33719
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.043771428571429e-06,
+      "loss": 1.1255,
+      "step": 33720
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.043502857142858e-06,
+      "loss": 1.4455,
+      "step": 33721
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.043234285714286e-06,
+      "loss": 0.8473,
+      "step": 33722
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.042965714285714e-06,
+      "loss": 0.8693,
+      "step": 33723
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.042697142857143e-06,
+      "loss": 1.5756,
+      "step": 33724
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.042428571428571e-06,
+      "loss": 1.1634,
+      "step": 33725
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0421599999999995e-06,
+      "loss": 1.1936,
+      "step": 33726
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0418914285714285e-06,
+      "loss": 0.9356,
+      "step": 33727
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.041622857142857e-06,
+      "loss": 1.0306,
+      "step": 33728
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.041354285714287e-06,
+      "loss": 1.0525,
+      "step": 33729
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.041085714285715e-06,
+      "loss": 1.2037,
+      "step": 33730
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.040817142857143e-06,
+      "loss": 1.0908,
+      "step": 33731
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.040548571428572e-06,
+      "loss": 0.9161,
+      "step": 33732
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.04028e-06,
+      "loss": 0.96,
+      "step": 33733
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.040011428571429e-06,
+      "loss": 0.8436,
+      "step": 33734
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.039742857142857e-06,
+      "loss": 1.2329,
+      "step": 33735
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0394742857142855e-06,
+      "loss": 1.3583,
+      "step": 33736
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0392057142857145e-06,
+      "loss": 1.0526,
+      "step": 33737
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.038937142857143e-06,
+      "loss": 1.3866,
+      "step": 33738
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.038668571428572e-06,
+      "loss": 1.1714,
+      "step": 33739
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0384e-06,
+      "loss": 1.4621,
+      "step": 33740
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.038131428571428e-06,
+      "loss": 1.1709,
+      "step": 33741
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.037862857142858e-06,
+      "loss": 0.9891,
+      "step": 33742
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.037594285714286e-06,
+      "loss": 0.9822,
+      "step": 33743
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.037325714285715e-06,
+      "loss": 1.263,
+      "step": 33744
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.037057142857143e-06,
+      "loss": 0.9228,
+      "step": 33745
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0367885714285715e-06,
+      "loss": 1.0625,
+      "step": 33746
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0365200000000005e-06,
+      "loss": 1.2021,
+      "step": 33747
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.036251428571429e-06,
+      "loss": 1.6712,
+      "step": 33748
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.035982857142857e-06,
+      "loss": 1.1857,
+      "step": 33749
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.035714285714286e-06,
+      "loss": 1.1288,
+      "step": 33750
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.035445714285714e-06,
+      "loss": 0.9719,
+      "step": 33751
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.035177142857143e-06,
+      "loss": 1.005,
+      "step": 33752
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.034908571428571e-06,
+      "loss": 1.6084,
+      "step": 33753
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0346399999999995e-06,
+      "loss": 0.8433,
+      "step": 33754
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.034371428571429e-06,
+      "loss": 1.4959,
+      "step": 33755
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0341028571428575e-06,
+      "loss": 1.0694,
+      "step": 33756
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0338342857142865e-06,
+      "loss": 1.1418,
+      "step": 33757
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.033565714285715e-06,
+      "loss": 1.0121,
+      "step": 33758
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.033297142857143e-06,
+      "loss": 1.2811,
+      "step": 33759
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.033028571428572e-06,
+      "loss": 1.3202,
+      "step": 33760
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.03276e-06,
+      "loss": 1.0885,
+      "step": 33761
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.032491428571429e-06,
+      "loss": 1.5197,
+      "step": 33762
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.032222857142857e-06,
+      "loss": 1.115,
+      "step": 33763
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0319542857142855e-06,
+      "loss": 0.8203,
+      "step": 33764
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0316857142857145e-06,
+      "loss": 0.9455,
+      "step": 33765
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.031417142857143e-06,
+      "loss": 1.4561,
+      "step": 33766
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.031148571428572e-06,
+      "loss": 1.2708,
+      "step": 33767
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.030880000000001e-06,
+      "loss": 1.0466,
+      "step": 33768
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.030611428571429e-06,
+      "loss": 1.1068,
+      "step": 33769
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.030342857142858e-06,
+      "loss": 1.3145,
+      "step": 33770
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.030074285714286e-06,
+      "loss": 0.9433,
+      "step": 33771
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.029805714285714e-06,
+      "loss": 1.2585,
+      "step": 33772
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.029537142857143e-06,
+      "loss": 1.2961,
+      "step": 33773
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0292685714285714e-06,
+      "loss": 1.3152,
+      "step": 33774
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0290000000000005e-06,
+      "loss": 0.8701,
+      "step": 33775
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.028731428571429e-06,
+      "loss": 0.9609,
+      "step": 33776
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.028462857142857e-06,
+      "loss": 1.4921,
+      "step": 33777
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.028194285714286e-06,
+      "loss": 0.962,
+      "step": 33778
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.027925714285714e-06,
+      "loss": 1.4966,
+      "step": 33779
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.027657142857143e-06,
+      "loss": 0.8611,
+      "step": 33780
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.027388571428571e-06,
+      "loss": 1.0306,
+      "step": 33781
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.02712e-06,
+      "loss": 1.4013,
+      "step": 33782
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.026851428571429e-06,
+      "loss": 0.8472,
+      "step": 33783
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0265828571428574e-06,
+      "loss": 1.3651,
+      "step": 33784
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0263142857142865e-06,
+      "loss": 0.969,
+      "step": 33785
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.026045714285715e-06,
+      "loss": 1.2183,
+      "step": 33786
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.025777142857143e-06,
+      "loss": 1.0111,
+      "step": 33787
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.025508571428572e-06,
+      "loss": 1.2527,
+      "step": 33788
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.02524e-06,
+      "loss": 1.0164,
+      "step": 33789
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.024971428571428e-06,
+      "loss": 1.6095,
+      "step": 33790
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.024702857142857e-06,
+      "loss": 1.1813,
+      "step": 33791
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.024434285714285e-06,
+      "loss": 1.0615,
+      "step": 33792
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.024165714285714e-06,
+      "loss": 1.0186,
+      "step": 33793
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.023897142857143e-06,
+      "loss": 0.9176,
+      "step": 33794
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.023628571428572e-06,
+      "loss": 0.8548,
+      "step": 33795
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.023360000000001e-06,
+      "loss": 0.559,
+      "step": 33796
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.023091428571429e-06,
+      "loss": 1.3516,
+      "step": 33797
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.022822857142858e-06,
+      "loss": 0.9293,
+      "step": 33798
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.022554285714286e-06,
+      "loss": 1.0738,
+      "step": 33799
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.022285714285714e-06,
+      "loss": 1.6731,
+      "step": 33800
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.022017142857143e-06,
+      "loss": 1.7074,
+      "step": 33801
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.021748571428571e-06,
+      "loss": 1.0616,
+      "step": 33802
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.02148e-06,
+      "loss": 0.8539,
+      "step": 33803
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.021211428571429e-06,
+      "loss": 0.8782,
+      "step": 33804
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.020942857142857e-06,
+      "loss": 0.9112,
+      "step": 33805
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.020674285714286e-06,
+      "loss": 1.304,
+      "step": 33806
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.020405714285714e-06,
+      "loss": 1.003,
+      "step": 33807
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.020137142857144e-06,
+      "loss": 1.0946,
+      "step": 33808
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.019868571428572e-06,
+      "loss": 1.3545,
+      "step": 33809
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0196e-06,
+      "loss": 0.6659,
+      "step": 33810
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.019331428571429e-06,
+      "loss": 1.0621,
+      "step": 33811
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.019062857142857e-06,
+      "loss": 1.605,
+      "step": 33812
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0187942857142856e-06,
+      "loss": 1.0236,
+      "step": 33813
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.018525714285715e-06,
+      "loss": 0.9701,
+      "step": 33814
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.018257142857143e-06,
+      "loss": 1.2643,
+      "step": 33815
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.017988571428572e-06,
+      "loss": 0.9566,
+      "step": 33816
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.01772e-06,
+      "loss": 1.2134,
+      "step": 33817
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.017451428571428e-06,
+      "loss": 1.2937,
+      "step": 33818
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.017182857142857e-06,
+      "loss": 1.1843,
+      "step": 33819
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.016914285714285e-06,
+      "loss": 1.4095,
+      "step": 33820
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.016645714285715e-06,
+      "loss": 1.5227,
+      "step": 33821
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.016377142857143e-06,
+      "loss": 1.5701,
+      "step": 33822
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0161085714285716e-06,
+      "loss": 1.1557,
+      "step": 33823
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.015840000000001e-06,
+      "loss": 0.9167,
+      "step": 33824
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.015571428571429e-06,
+      "loss": 1.3741,
+      "step": 33825
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.015302857142858e-06,
+      "loss": 1.3558,
+      "step": 33826
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.015034285714286e-06,
+      "loss": 1.1405,
+      "step": 33827
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.014765714285714e-06,
+      "loss": 1.5474,
+      "step": 33828
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.014497142857143e-06,
+      "loss": 0.8763,
+      "step": 33829
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.014228571428571e-06,
+      "loss": 1.3704,
+      "step": 33830
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.01396e-06,
+      "loss": 0.9013,
+      "step": 33831
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0136914285714285e-06,
+      "loss": 1.1607,
+      "step": 33832
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.013422857142857e-06,
+      "loss": 1.4345,
+      "step": 33833
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.013154285714286e-06,
+      "loss": 0.989,
+      "step": 33834
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.012885714285715e-06,
+      "loss": 1.0214,
+      "step": 33835
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.012617142857143e-06,
+      "loss": 1.6083,
+      "step": 33836
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.012348571428572e-06,
+      "loss": 1.0916,
+      "step": 33837
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.01208e-06,
+      "loss": 1.172,
+      "step": 33838
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.011811428571429e-06,
+      "loss": 0.9449,
+      "step": 33839
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.011542857142857e-06,
+      "loss": 0.8675,
+      "step": 33840
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0112742857142855e-06,
+      "loss": 1.3672,
+      "step": 33841
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0110057142857145e-06,
+      "loss": 1.3278,
+      "step": 33842
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.010737142857143e-06,
+      "loss": 1.2196,
+      "step": 33843
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.010468571428572e-06,
+      "loss": 1.4822,
+      "step": 33844
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0102e-06,
+      "loss": 1.2523,
+      "step": 33845
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.009931428571428e-06,
+      "loss": 0.895,
+      "step": 33846
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.009662857142857e-06,
+      "loss": 1.2604,
+      "step": 33847
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.009394285714286e-06,
+      "loss": 1.144,
+      "step": 33848
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.009125714285715e-06,
+      "loss": 1.0868,
+      "step": 33849
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.008857142857143e-06,
+      "loss": 1.1319,
+      "step": 33850
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0085885714285715e-06,
+      "loss": 1.4453,
+      "step": 33851
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0083200000000005e-06,
+      "loss": 1.2131,
+      "step": 33852
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.008051428571429e-06,
+      "loss": 1.0812,
+      "step": 33853
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.007782857142857e-06,
+      "loss": 1.0688,
+      "step": 33854
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.007514285714286e-06,
+      "loss": 0.9444,
+      "step": 33855
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.007245714285714e-06,
+      "loss": 1.5856,
+      "step": 33856
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.006977142857143e-06,
+      "loss": 1.0216,
+      "step": 33857
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.006708571428571e-06,
+      "loss": 1.0861,
+      "step": 33858
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0064399999999995e-06,
+      "loss": 1.7282,
+      "step": 33859
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0061714285714285e-06,
+      "loss": 1.1755,
+      "step": 33860
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0059028571428575e-06,
+      "loss": 0.6967,
+      "step": 33861
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.0056342857142865e-06,
+      "loss": 0.9959,
+      "step": 33862
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.005365714285715e-06,
+      "loss": 1.4288,
+      "step": 33863
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.005097142857143e-06,
+      "loss": 1.0762,
+      "step": 33864
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.004828571428572e-06,
+      "loss": 0.871,
+      "step": 33865
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.00456e-06,
+      "loss": 0.9259,
+      "step": 33866
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.004291428571429e-06,
+      "loss": 1.083,
+      "step": 33867
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.004022857142857e-06,
+      "loss": 1.1319,
+      "step": 33868
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.0037542857142854e-06,
+      "loss": 1.4,
+      "step": 33869
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.0034857142857145e-06,
+      "loss": 1.5462,
+      "step": 33870
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.003217142857143e-06,
+      "loss": 0.8423,
+      "step": 33871
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.002948571428572e-06,
+      "loss": 0.8356,
+      "step": 33872
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.00268e-06,
+      "loss": 1.1745,
+      "step": 33873
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.002411428571429e-06,
+      "loss": 1.3666,
+      "step": 33874
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.002142857142858e-06,
+      "loss": 0.9731,
+      "step": 33875
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.001874285714286e-06,
+      "loss": 1.1913,
+      "step": 33876
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.001605714285714e-06,
+      "loss": 1.3017,
+      "step": 33877
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.001337142857143e-06,
+      "loss": 1.3625,
+      "step": 33878
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.0010685714285714e-06,
+      "loss": 0.9696,
+      "step": 33879
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.0008000000000005e-06,
+      "loss": 1.4769,
+      "step": 33880
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.000531428571429e-06,
+      "loss": 1.3751,
+      "step": 33881
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.000262857142857e-06,
+      "loss": 0.9923,
+      "step": 33882
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.999994285714286e-06,
+      "loss": 1.0718,
+      "step": 33883
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.999725714285714e-06,
+      "loss": 1.0458,
+      "step": 33884
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.999457142857143e-06,
+      "loss": 1.0642,
+      "step": 33885
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.999188571428571e-06,
+      "loss": 0.9399,
+      "step": 33886
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.99892e-06,
+      "loss": 1.2485,
+      "step": 33887
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.998651428571429e-06,
+      "loss": 1.6671,
+      "step": 33888
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9983828571428574e-06,
+      "loss": 0.8326,
+      "step": 33889
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9981142857142865e-06,
+      "loss": 1.1035,
+      "step": 33890
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.997845714285715e-06,
+      "loss": 0.8873,
+      "step": 33891
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.997577142857143e-06,
+      "loss": 1.4166,
+      "step": 33892
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.997308571428572e-06,
+      "loss": 1.5669,
+      "step": 33893
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.99704e-06,
+      "loss": 0.8582,
+      "step": 33894
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.996771428571429e-06,
+      "loss": 1.2015,
+      "step": 33895
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.996502857142857e-06,
+      "loss": 0.9597,
+      "step": 33896
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.996234285714285e-06,
+      "loss": 1.0282,
+      "step": 33897
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.995965714285714e-06,
+      "loss": 1.0164,
+      "step": 33898
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.995697142857143e-06,
+      "loss": 0.932,
+      "step": 33899
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.995428571428572e-06,
+      "loss": 1.1631,
+      "step": 33900
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.995160000000001e-06,
+      "loss": 0.8559,
+      "step": 33901
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.994891428571429e-06,
+      "loss": 1.0437,
+      "step": 33902
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.994622857142858e-06,
+      "loss": 1.208,
+      "step": 33903
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.994354285714286e-06,
+      "loss": 1.2622,
+      "step": 33904
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.994085714285714e-06,
+      "loss": 1.1762,
+      "step": 33905
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.993817142857143e-06,
+      "loss": 1.63,
+      "step": 33906
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.993548571428571e-06,
+      "loss": 1.3916,
+      "step": 33907
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.99328e-06,
+      "loss": 1.0999,
+      "step": 33908
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.993011428571429e-06,
+      "loss": 1.0371,
+      "step": 33909
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.992742857142857e-06,
+      "loss": 1.4383,
+      "step": 33910
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.992474285714286e-06,
+      "loss": 1.0011,
+      "step": 33911
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.992205714285714e-06,
+      "loss": 0.9028,
+      "step": 33912
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.991937142857143e-06,
+      "loss": 1.4742,
+      "step": 33913
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.991668571428572e-06,
+      "loss": 1.1719,
+      "step": 33914
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9914e-06,
+      "loss": 1.0401,
+      "step": 33915
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.991131428571429e-06,
+      "loss": 1.5526,
+      "step": 33916
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.990862857142857e-06,
+      "loss": 0.82,
+      "step": 33917
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.990594285714286e-06,
+      "loss": 1.2835,
+      "step": 33918
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.990325714285715e-06,
+      "loss": 1.176,
+      "step": 33919
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.990057142857143e-06,
+      "loss": 1.0263,
+      "step": 33920
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.989788571428572e-06,
+      "loss": 1.2836,
+      "step": 33921
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.98952e-06,
+      "loss": 1.1659,
+      "step": 33922
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.989251428571428e-06,
+      "loss": 1.279,
+      "step": 33923
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.988982857142857e-06,
+      "loss": 1.5508,
+      "step": 33924
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.988714285714285e-06,
+      "loss": 1.1211,
+      "step": 33925
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.988445714285714e-06,
+      "loss": 1.1576,
+      "step": 33926
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.988177142857143e-06,
+      "loss": 0.7411,
+      "step": 33927
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9879085714285716e-06,
+      "loss": 1.0856,
+      "step": 33928
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.987640000000001e-06,
+      "loss": 1.4261,
+      "step": 33929
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.987371428571429e-06,
+      "loss": 1.2771,
+      "step": 33930
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.987102857142858e-06,
+      "loss": 1.1689,
+      "step": 33931
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.986834285714286e-06,
+      "loss": 1.0056,
+      "step": 33932
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.986565714285714e-06,
+      "loss": 0.8395,
+      "step": 33933
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.986297142857143e-06,
+      "loss": 0.99,
+      "step": 33934
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.986028571428571e-06,
+      "loss": 0.9861,
+      "step": 33935
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.98576e-06,
+      "loss": 0.8587,
+      "step": 33936
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9854914285714285e-06,
+      "loss": 0.7715,
+      "step": 33937
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.985222857142857e-06,
+      "loss": 1.2598,
+      "step": 33938
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.984954285714286e-06,
+      "loss": 1.3802,
+      "step": 33939
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.984685714285715e-06,
+      "loss": 1.2183,
+      "step": 33940
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.984417142857143e-06,
+      "loss": 1.3281,
+      "step": 33941
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.984148571428572e-06,
+      "loss": 0.6636,
+      "step": 33942
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.98388e-06,
+      "loss": 0.9152,
+      "step": 33943
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.983611428571429e-06,
+      "loss": 1.2183,
+      "step": 33944
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.983342857142857e-06,
+      "loss": 1.1485,
+      "step": 33945
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9830742857142855e-06,
+      "loss": 1.0713,
+      "step": 33946
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9828057142857145e-06,
+      "loss": 1.2585,
+      "step": 33947
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.982537142857143e-06,
+      "loss": 1.2059,
+      "step": 33948
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.982268571428572e-06,
+      "loss": 1.1273,
+      "step": 33949
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.982e-06,
+      "loss": 1.7916,
+      "step": 33950
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.981731428571428e-06,
+      "loss": 1.0382,
+      "step": 33951
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.981462857142857e-06,
+      "loss": 1.0327,
+      "step": 33952
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.981194285714286e-06,
+      "loss": 0.9088,
+      "step": 33953
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.980925714285715e-06,
+      "loss": 1.2723,
+      "step": 33954
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.980657142857143e-06,
+      "loss": 1.4048,
+      "step": 33955
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9803885714285715e-06,
+      "loss": 1.1775,
+      "step": 33956
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9801200000000005e-06,
+      "loss": 1.476,
+      "step": 33957
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.979851428571429e-06,
+      "loss": 1.3509,
+      "step": 33958
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.979582857142858e-06,
+      "loss": 1.0922,
+      "step": 33959
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.979314285714286e-06,
+      "loss": 1.1493,
+      "step": 33960
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.979045714285714e-06,
+      "loss": 1.1963,
+      "step": 33961
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.978777142857143e-06,
+      "loss": 1.0832,
+      "step": 33962
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.978508571428571e-06,
+      "loss": 0.882,
+      "step": 33963
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9782399999999995e-06,
+      "loss": 0.8187,
+      "step": 33964
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9779714285714285e-06,
+      "loss": 0.9138,
+      "step": 33965
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9777028571428575e-06,
+      "loss": 1.6975,
+      "step": 33966
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9774342857142865e-06,
+      "loss": 1.2143,
+      "step": 33967
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.977165714285715e-06,
+      "loss": 1.2143,
+      "step": 33968
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.976897142857143e-06,
+      "loss": 1.1515,
+      "step": 33969
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.976628571428572e-06,
+      "loss": 1.3077,
+      "step": 33970
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.97636e-06,
+      "loss": 1.0716,
+      "step": 33971
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.976091428571429e-06,
+      "loss": 1.2187,
+      "step": 33972
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.975822857142857e-06,
+      "loss": 0.8751,
+      "step": 33973
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9755542857142854e-06,
+      "loss": 1.2977,
+      "step": 33974
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9752857142857145e-06,
+      "loss": 0.9175,
+      "step": 33975
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.975017142857143e-06,
+      "loss": 0.9056,
+      "step": 33976
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.974748571428572e-06,
+      "loss": 1.228,
+      "step": 33977
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.97448e-06,
+      "loss": 1.1432,
+      "step": 33978
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.974211428571429e-06,
+      "loss": 0.855,
+      "step": 33979
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.973942857142858e-06,
+      "loss": 1.0241,
+      "step": 33980
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.973674285714286e-06,
+      "loss": 1.5709,
+      "step": 33981
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.973405714285715e-06,
+      "loss": 1.3154,
+      "step": 33982
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.973137142857143e-06,
+      "loss": 0.7025,
+      "step": 33983
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9728685714285714e-06,
+      "loss": 1.2767,
+      "step": 33984
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9726000000000005e-06,
+      "loss": 0.8911,
+      "step": 33985
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.972331428571429e-06,
+      "loss": 0.86,
+      "step": 33986
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.972062857142857e-06,
+      "loss": 1.1557,
+      "step": 33987
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.971794285714286e-06,
+      "loss": 1.1738,
+      "step": 33988
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.971525714285714e-06,
+      "loss": 1.3718,
+      "step": 33989
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.971257142857143e-06,
+      "loss": 0.9413,
+      "step": 33990
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.970988571428571e-06,
+      "loss": 1.2279,
+      "step": 33991
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.97072e-06,
+      "loss": 1.264,
+      "step": 33992
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.970451428571429e-06,
+      "loss": 1.1382,
+      "step": 33993
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9701828571428574e-06,
+      "loss": 1.4088,
+      "step": 33994
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9699142857142865e-06,
+      "loss": 1.3911,
+      "step": 33995
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.969645714285715e-06,
+      "loss": 1.0248,
+      "step": 33996
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.969377142857143e-06,
+      "loss": 1.1008,
+      "step": 33997
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.969108571428572e-06,
+      "loss": 1.23,
+      "step": 33998
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.96884e-06,
+      "loss": 1.0458,
+      "step": 33999
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.968571428571429e-06,
+      "loss": 1.2058,
+      "step": 34000
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.968302857142857e-06,
+      "loss": 0.9824,
+      "step": 34001
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.968034285714285e-06,
+      "loss": 1.3403,
+      "step": 34002
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.967765714285714e-06,
+      "loss": 0.8526,
+      "step": 34003
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.967497142857143e-06,
+      "loss": 1.1028,
+      "step": 34004
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.967228571428572e-06,
+      "loss": 1.1411,
+      "step": 34005
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.966960000000001e-06,
+      "loss": 1.0617,
+      "step": 34006
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.966691428571429e-06,
+      "loss": 1.0339,
+      "step": 34007
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.966422857142858e-06,
+      "loss": 1.2551,
+      "step": 34008
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.966154285714286e-06,
+      "loss": 1.1714,
+      "step": 34009
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.965885714285714e-06,
+      "loss": 0.4852,
+      "step": 34010
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.965617142857143e-06,
+      "loss": 1.1757,
+      "step": 34011
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.965348571428571e-06,
+      "loss": 1.3578,
+      "step": 34012
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.96508e-06,
+      "loss": 1.2268,
+      "step": 34013
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.964811428571429e-06,
+      "loss": 1.4324,
+      "step": 34014
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.964542857142857e-06,
+      "loss": 1.0273,
+      "step": 34015
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.964274285714286e-06,
+      "loss": 1.056,
+      "step": 34016
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.964005714285714e-06,
+      "loss": 1.0511,
+      "step": 34017
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.963737142857143e-06,
+      "loss": 1.5813,
+      "step": 34018
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.963468571428572e-06,
+      "loss": 0.885,
+      "step": 34019
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9632e-06,
+      "loss": 1.1081,
+      "step": 34020
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.962931428571429e-06,
+      "loss": 1.2522,
+      "step": 34021
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.962662857142857e-06,
+      "loss": 1.1327,
+      "step": 34022
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.962394285714286e-06,
+      "loss": 1.0134,
+      "step": 34023
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.962125714285715e-06,
+      "loss": 0.9946,
+      "step": 34024
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.961857142857143e-06,
+      "loss": 1.0105,
+      "step": 34025
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.961588571428572e-06,
+      "loss": 1.5929,
+      "step": 34026
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.96132e-06,
+      "loss": 1.1875,
+      "step": 34027
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.961051428571428e-06,
+      "loss": 1.2391,
+      "step": 34028
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.960782857142857e-06,
+      "loss": 1.3208,
+      "step": 34029
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.960514285714285e-06,
+      "loss": 1.2428,
+      "step": 34030
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.960245714285714e-06,
+      "loss": 1.2558,
+      "step": 34031
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.959977142857143e-06,
+      "loss": 1.4362,
+      "step": 34032
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.9597085714285716e-06,
+      "loss": 1.2012,
+      "step": 34033
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.959440000000001e-06,
+      "loss": 0.8065,
+      "step": 34034
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.959171428571429e-06,
+      "loss": 1.2018,
+      "step": 34035
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.958902857142858e-06,
+      "loss": 1.168,
+      "step": 34036
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.958634285714286e-06,
+      "loss": 1.0554,
+      "step": 34037
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.958365714285714e-06,
+      "loss": 1.1342,
+      "step": 34038
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.958097142857143e-06,
+      "loss": 0.9355,
+      "step": 34039
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.957828571428571e-06,
+      "loss": 0.9568,
+      "step": 34040
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.95756e-06,
+      "loss": 0.9851,
+      "step": 34041
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9572914285714285e-06,
+      "loss": 1.1263,
+      "step": 34042
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.957022857142857e-06,
+      "loss": 0.7507,
+      "step": 34043
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.956754285714286e-06,
+      "loss": 1.2793,
+      "step": 34044
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.956485714285715e-06,
+      "loss": 1.5468,
+      "step": 34045
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.956217142857144e-06,
+      "loss": 0.768,
+      "step": 34046
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.955948571428572e-06,
+      "loss": 1.3816,
+      "step": 34047
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.95568e-06,
+      "loss": 1.1132,
+      "step": 34048
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.955411428571429e-06,
+      "loss": 1.2174,
+      "step": 34049
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.955142857142857e-06,
+      "loss": 1.4096,
+      "step": 34050
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9548742857142855e-06,
+      "loss": 1.3123,
+      "step": 34051
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9546057142857145e-06,
+      "loss": 1.3765,
+      "step": 34052
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.954337142857143e-06,
+      "loss": 1.1703,
+      "step": 34053
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.954068571428572e-06,
+      "loss": 1.1974,
+      "step": 34054
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9538e-06,
+      "loss": 0.8839,
+      "step": 34055
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.953531428571428e-06,
+      "loss": 1.2581,
+      "step": 34056
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.953262857142857e-06,
+      "loss": 1.2482,
+      "step": 34057
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.952994285714286e-06,
+      "loss": 1.1369,
+      "step": 34058
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.952725714285715e-06,
+      "loss": 0.7998,
+      "step": 34059
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.952457142857143e-06,
+      "loss": 0.9612,
+      "step": 34060
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9521885714285715e-06,
+      "loss": 1.0573,
+      "step": 34061
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9519200000000005e-06,
+      "loss": 0.9836,
+      "step": 34062
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.951651428571429e-06,
+      "loss": 0.8844,
+      "step": 34063
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.951382857142858e-06,
+      "loss": 1.2954,
+      "step": 34064
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.951114285714286e-06,
+      "loss": 1.728,
+      "step": 34065
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.950845714285714e-06,
+      "loss": 1.5122,
+      "step": 34066
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.950577142857143e-06,
+      "loss": 0.8638,
+      "step": 34067
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.950308571428571e-06,
+      "loss": 1.0536,
+      "step": 34068
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.95004e-06,
+      "loss": 1.0723,
+      "step": 34069
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9497714285714285e-06,
+      "loss": 1.2429,
+      "step": 34070
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9495028571428575e-06,
+      "loss": 1.3519,
+      "step": 34071
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9492342857142865e-06,
+      "loss": 1.3838,
+      "step": 34072
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.948965714285715e-06,
+      "loss": 1.3995,
+      "step": 34073
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.948697142857143e-06,
+      "loss": 1.3659,
+      "step": 34074
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.948428571428572e-06,
+      "loss": 1.2478,
+      "step": 34075
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.94816e-06,
+      "loss": 1.2205,
+      "step": 34076
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.947891428571429e-06,
+      "loss": 1.1598,
+      "step": 34077
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.947622857142857e-06,
+      "loss": 0.9949,
+      "step": 34078
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9473542857142854e-06,
+      "loss": 0.9755,
+      "step": 34079
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9470857142857145e-06,
+      "loss": 0.8695,
+      "step": 34080
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.946817142857143e-06,
+      "loss": 0.8139,
+      "step": 34081
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.946548571428572e-06,
+      "loss": 1.6015,
+      "step": 34082
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.94628e-06,
+      "loss": 1.0955,
+      "step": 34083
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.946011428571429e-06,
+      "loss": 0.9585,
+      "step": 34084
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.945742857142858e-06,
+      "loss": 1.1054,
+      "step": 34085
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.945474285714286e-06,
+      "loss": 0.8795,
+      "step": 34086
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.945205714285715e-06,
+      "loss": 1.2501,
+      "step": 34087
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.944937142857143e-06,
+      "loss": 1.2741,
+      "step": 34088
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9446685714285714e-06,
+      "loss": 1.0959,
+      "step": 34089
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9444000000000005e-06,
+      "loss": 1.2759,
+      "step": 34090
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.944131428571429e-06,
+      "loss": 1.1975,
+      "step": 34091
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.943862857142858e-06,
+      "loss": 0.9837,
+      "step": 34092
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.943594285714286e-06,
+      "loss": 0.8337,
+      "step": 34093
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.943325714285714e-06,
+      "loss": 1.4239,
+      "step": 34094
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.943057142857143e-06,
+      "loss": 0.7776,
+      "step": 34095
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.942788571428571e-06,
+      "loss": 1.3091,
+      "step": 34096
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.94252e-06,
+      "loss": 1.1979,
+      "step": 34097
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.942251428571429e-06,
+      "loss": 0.9345,
+      "step": 34098
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9419828571428574e-06,
+      "loss": 1.0151,
+      "step": 34099
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9417142857142865e-06,
+      "loss": 1.0615,
+      "step": 34100
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.941445714285715e-06,
+      "loss": 1.1366,
+      "step": 34101
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.941177142857143e-06,
+      "loss": 1.1895,
+      "step": 34102
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.940908571428572e-06,
+      "loss": 1.5044,
+      "step": 34103
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.94064e-06,
+      "loss": 1.0947,
+      "step": 34104
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.940371428571429e-06,
+      "loss": 1.4728,
+      "step": 34105
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.940102857142857e-06,
+      "loss": 1.202,
+      "step": 34106
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.939834285714285e-06,
+      "loss": 0.7948,
+      "step": 34107
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.939565714285714e-06,
+      "loss": 0.7948,
+      "step": 34108
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.939297142857143e-06,
+      "loss": 1.0202,
+      "step": 34109
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.939028571428572e-06,
+      "loss": 1.7086,
+      "step": 34110
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.938760000000001e-06,
+      "loss": 1.1865,
+      "step": 34111
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.938491428571429e-06,
+      "loss": 0.8042,
+      "step": 34112
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.938222857142858e-06,
+      "loss": 1.3908,
+      "step": 34113
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.937954285714286e-06,
+      "loss": 1.3775,
+      "step": 34114
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.937685714285714e-06,
+      "loss": 1.3227,
+      "step": 34115
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.937417142857143e-06,
+      "loss": 1.7083,
+      "step": 34116
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.937148571428571e-06,
+      "loss": 0.7693,
+      "step": 34117
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.93688e-06,
+      "loss": 1.2422,
+      "step": 34118
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.936611428571429e-06,
+      "loss": 0.7603,
+      "step": 34119
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.936342857142857e-06,
+      "loss": 0.9965,
+      "step": 34120
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.936074285714286e-06,
+      "loss": 1.1138,
+      "step": 34121
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.935805714285714e-06,
+      "loss": 1.1253,
+      "step": 34122
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.935537142857143e-06,
+      "loss": 1.3272,
+      "step": 34123
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.935268571428572e-06,
+      "loss": 0.9276,
+      "step": 34124
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.935e-06,
+      "loss": 1.159,
+      "step": 34125
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.934731428571429e-06,
+      "loss": 1.4603,
+      "step": 34126
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.934462857142857e-06,
+      "loss": 1.2343,
+      "step": 34127
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.934194285714286e-06,
+      "loss": 0.5579,
+      "step": 34128
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.933925714285715e-06,
+      "loss": 1.1532,
+      "step": 34129
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.933657142857143e-06,
+      "loss": 0.7579,
+      "step": 34130
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.933388571428572e-06,
+      "loss": 1.4132,
+      "step": 34131
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.93312e-06,
+      "loss": 1.3517,
+      "step": 34132
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.932851428571429e-06,
+      "loss": 0.8,
+      "step": 34133
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.932582857142857e-06,
+      "loss": 0.9833,
+      "step": 34134
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.932314285714285e-06,
+      "loss": 1.1185,
+      "step": 34135
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.932045714285714e-06,
+      "loss": 1.1669,
+      "step": 34136
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.931777142857143e-06,
+      "loss": 0.997,
+      "step": 34137
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9315085714285716e-06,
+      "loss": 1.5637,
+      "step": 34138
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.931240000000001e-06,
+      "loss": 0.6953,
+      "step": 34139
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.930971428571429e-06,
+      "loss": 1.2433,
+      "step": 34140
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.930702857142858e-06,
+      "loss": 1.4859,
+      "step": 34141
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.930434285714286e-06,
+      "loss": 0.8459,
+      "step": 34142
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.930165714285714e-06,
+      "loss": 1.4464,
+      "step": 34143
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.929897142857143e-06,
+      "loss": 1.0536,
+      "step": 34144
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.929628571428571e-06,
+      "loss": 0.6008,
+      "step": 34145
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.92936e-06,
+      "loss": 0.9713,
+      "step": 34146
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9290914285714285e-06,
+      "loss": 1.5361,
+      "step": 34147
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.928822857142857e-06,
+      "loss": 1.0001,
+      "step": 34148
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.928554285714286e-06,
+      "loss": 0.8057,
+      "step": 34149
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.928285714285715e-06,
+      "loss": 1.302,
+      "step": 34150
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.928017142857144e-06,
+      "loss": 0.9281,
+      "step": 34151
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.927748571428572e-06,
+      "loss": 0.9939,
+      "step": 34152
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.92748e-06,
+      "loss": 1.3349,
+      "step": 34153
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.927211428571429e-06,
+      "loss": 1.4013,
+      "step": 34154
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.926942857142857e-06,
+      "loss": 1.0943,
+      "step": 34155
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.926674285714286e-06,
+      "loss": 0.9903,
+      "step": 34156
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9264057142857145e-06,
+      "loss": 0.9365,
+      "step": 34157
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.926137142857143e-06,
+      "loss": 1.1027,
+      "step": 34158
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.925868571428572e-06,
+      "loss": 1.2101,
+      "step": 34159
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9256e-06,
+      "loss": 0.8483,
+      "step": 34160
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.925331428571428e-06,
+      "loss": 0.7206,
+      "step": 34161
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.925062857142857e-06,
+      "loss": 1.3018,
+      "step": 34162
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.924794285714286e-06,
+      "loss": 1.2933,
+      "step": 34163
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.924525714285715e-06,
+      "loss": 1.3432,
+      "step": 34164
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.924257142857143e-06,
+      "loss": 1.2179,
+      "step": 34165
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9239885714285715e-06,
+      "loss": 0.9224,
+      "step": 34166
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9237200000000005e-06,
+      "loss": 1.0102,
+      "step": 34167
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.923451428571429e-06,
+      "loss": 1.2788,
+      "step": 34168
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.923182857142858e-06,
+      "loss": 1.2049,
+      "step": 34169
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.922914285714286e-06,
+      "loss": 1.1791,
+      "step": 34170
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.922645714285714e-06,
+      "loss": 1.3871,
+      "step": 34171
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.922377142857143e-06,
+      "loss": 1.1205,
+      "step": 34172
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.922108571428571e-06,
+      "loss": 1.0099,
+      "step": 34173
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.92184e-06,
+      "loss": 0.8278,
+      "step": 34174
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9215714285714285e-06,
+      "loss": 0.9735,
+      "step": 34175
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.921302857142857e-06,
+      "loss": 1.2069,
+      "step": 34176
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9210342857142865e-06,
+      "loss": 1.2026,
+      "step": 34177
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.920765714285715e-06,
+      "loss": 0.8377,
+      "step": 34178
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.920497142857144e-06,
+      "loss": 0.8645,
+      "step": 34179
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.920228571428572e-06,
+      "loss": 1.3193,
+      "step": 34180
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.91996e-06,
+      "loss": 0.8538,
+      "step": 34181
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.919691428571429e-06,
+      "loss": 1.5038,
+      "step": 34182
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.919422857142857e-06,
+      "loss": 1.3077,
+      "step": 34183
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9191542857142854e-06,
+      "loss": 1.1465,
+      "step": 34184
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9188857142857145e-06,
+      "loss": 0.6457,
+      "step": 34185
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.918617142857143e-06,
+      "loss": 1.196,
+      "step": 34186
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.918348571428572e-06,
+      "loss": 1.2332,
+      "step": 34187
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.91808e-06,
+      "loss": 1.293,
+      "step": 34188
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.917811428571428e-06,
+      "loss": 0.9843,
+      "step": 34189
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.917542857142858e-06,
+      "loss": 0.9711,
+      "step": 34190
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.917274285714286e-06,
+      "loss": 1.3775,
+      "step": 34191
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.917005714285715e-06,
+      "loss": 1.1953,
+      "step": 34192
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.916737142857143e-06,
+      "loss": 1.4683,
+      "step": 34193
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9164685714285714e-06,
+      "loss": 1.307,
+      "step": 34194
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9162000000000005e-06,
+      "loss": 1.1615,
+      "step": 34195
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.915931428571429e-06,
+      "loss": 0.9988,
+      "step": 34196
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.915662857142858e-06,
+      "loss": 1.3218,
+      "step": 34197
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.915394285714286e-06,
+      "loss": 1.0252,
+      "step": 34198
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.915125714285714e-06,
+      "loss": 0.8241,
+      "step": 34199
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.914857142857143e-06,
+      "loss": 1.1866,
+      "step": 34200
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.914588571428571e-06,
+      "loss": 0.9432,
+      "step": 34201
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.914319999999999e-06,
+      "loss": 1.0044,
+      "step": 34202
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.914051428571429e-06,
+      "loss": 1.1421,
+      "step": 34203
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9137828571428574e-06,
+      "loss": 0.9313,
+      "step": 34204
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.9135142857142865e-06,
+      "loss": 1.1018,
+      "step": 34205
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.913245714285715e-06,
+      "loss": 1.1707,
+      "step": 34206
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.912977142857143e-06,
+      "loss": 1.1935,
+      "step": 34207
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.912708571428572e-06,
+      "loss": 0.8927,
+      "step": 34208
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.91244e-06,
+      "loss": 0.9906,
+      "step": 34209
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.912171428571429e-06,
+      "loss": 1.1107,
+      "step": 34210
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.911902857142857e-06,
+      "loss": 1.5985,
+      "step": 34211
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.911634285714285e-06,
+      "loss": 1.104,
+      "step": 34212
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.911365714285714e-06,
+      "loss": 0.82,
+      "step": 34213
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.911097142857143e-06,
+      "loss": 1.3927,
+      "step": 34214
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.910828571428572e-06,
+      "loss": 0.8124,
+      "step": 34215
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.910560000000001e-06,
+      "loss": 1.1094,
+      "step": 34216
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.910291428571429e-06,
+      "loss": 1.0557,
+      "step": 34217
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.910022857142858e-06,
+      "loss": 0.6511,
+      "step": 34218
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.909754285714286e-06,
+      "loss": 0.9924,
+      "step": 34219
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.909485714285715e-06,
+      "loss": 1.3306,
+      "step": 34220
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.909217142857143e-06,
+      "loss": 1.3932,
+      "step": 34221
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.908948571428571e-06,
+      "loss": 1.255,
+      "step": 34222
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.90868e-06,
+      "loss": 1.2337,
+      "step": 34223
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.908411428571429e-06,
+      "loss": 1.1599,
+      "step": 34224
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.908142857142857e-06,
+      "loss": 0.7376,
+      "step": 34225
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.907874285714286e-06,
+      "loss": 0.8684,
+      "step": 34226
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.907605714285714e-06,
+      "loss": 1.27,
+      "step": 34227
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.907337142857143e-06,
+      "loss": 1.1253,
+      "step": 34228
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.907068571428571e-06,
+      "loss": 0.8128,
+      "step": 34229
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.9068e-06,
+      "loss": 1.1202,
+      "step": 34230
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.906531428571429e-06,
+      "loss": 1.1108,
+      "step": 34231
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.906262857142857e-06,
+      "loss": 1.3137,
+      "step": 34232
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.905994285714286e-06,
+      "loss": 0.8114,
+      "step": 34233
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.905725714285715e-06,
+      "loss": 0.9139,
+      "step": 34234
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.905457142857143e-06,
+      "loss": 1.0195,
+      "step": 34235
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.905188571428572e-06,
+      "loss": 1.0845,
+      "step": 34236
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.90492e-06,
+      "loss": 1.0921,
+      "step": 34237
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.904651428571429e-06,
+      "loss": 0.9853,
+      "step": 34238
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.904382857142857e-06,
+      "loss": 1.1252,
+      "step": 34239
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.904114285714285e-06,
+      "loss": 1.1236,
+      "step": 34240
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.903845714285714e-06,
+      "loss": 1.2492,
+      "step": 34241
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.9035771428571425e-06,
+      "loss": 0.8648,
+      "step": 34242
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.903308571428572e-06,
+      "loss": 0.9999,
+      "step": 34243
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.903040000000001e-06,
+      "loss": 0.9186,
+      "step": 34244
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.902771428571429e-06,
+      "loss": 0.9089,
+      "step": 34245
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.902502857142858e-06,
+      "loss": 0.9699,
+      "step": 34246
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.902234285714286e-06,
+      "loss": 1.0135,
+      "step": 34247
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.901965714285714e-06,
+      "loss": 1.1935,
+      "step": 34248
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.901697142857143e-06,
+      "loss": 1.1536,
+      "step": 34249
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.901428571428571e-06,
+      "loss": 1.504,
+      "step": 34250
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.90116e-06,
+      "loss": 1.5827,
+      "step": 34251
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.9008914285714285e-06,
+      "loss": 1.4927,
+      "step": 34252
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.900622857142857e-06,
+      "loss": 0.9822,
+      "step": 34253
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.900354285714286e-06,
+      "loss": 1.2707,
+      "step": 34254
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.900085714285714e-06,
+      "loss": 1.1324,
+      "step": 34255
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.899817142857144e-06,
+      "loss": 1.2883,
+      "step": 34256
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.899548571428572e-06,
+      "loss": 1.2605,
+      "step": 34257
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.89928e-06,
+      "loss": 1.0431,
+      "step": 34258
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.899011428571429e-06,
+      "loss": 1.0439,
+      "step": 34259
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.898742857142857e-06,
+      "loss": 1.0157,
+      "step": 34260
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.898474285714286e-06,
+      "loss": 1.0494,
+      "step": 34261
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8982057142857145e-06,
+      "loss": 1.1858,
+      "step": 34262
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.897937142857143e-06,
+      "loss": 1.002,
+      "step": 34263
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.897668571428572e-06,
+      "loss": 1.5366,
+      "step": 34264
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8974e-06,
+      "loss": 1.0651,
+      "step": 34265
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.897131428571429e-06,
+      "loss": 1.2578,
+      "step": 34266
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.896862857142857e-06,
+      "loss": 1.0281,
+      "step": 34267
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.896594285714285e-06,
+      "loss": 1.1991,
+      "step": 34268
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.896325714285715e-06,
+      "loss": 1.4443,
+      "step": 34269
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.896057142857143e-06,
+      "loss": 1.3482,
+      "step": 34270
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8957885714285715e-06,
+      "loss": 1.0993,
+      "step": 34271
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8955200000000005e-06,
+      "loss": 1.15,
+      "step": 34272
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.895251428571429e-06,
+      "loss": 0.7716,
+      "step": 34273
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.894982857142858e-06,
+      "loss": 1.1204,
+      "step": 34274
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.894714285714286e-06,
+      "loss": 1.6668,
+      "step": 34275
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.894445714285714e-06,
+      "loss": 1.0513,
+      "step": 34276
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.894177142857143e-06,
+      "loss": 1.4111,
+      "step": 34277
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.893908571428571e-06,
+      "loss": 0.9705,
+      "step": 34278
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.89364e-06,
+      "loss": 1.1307,
+      "step": 34279
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8933714285714285e-06,
+      "loss": 1.0397,
+      "step": 34280
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.893102857142857e-06,
+      "loss": 0.9779,
+      "step": 34281
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.892834285714286e-06,
+      "loss": 1.0504,
+      "step": 34282
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.892565714285715e-06,
+      "loss": 1.014,
+      "step": 34283
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.892297142857144e-06,
+      "loss": 1.0656,
+      "step": 34284
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.892028571428572e-06,
+      "loss": 0.984,
+      "step": 34285
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.89176e-06,
+      "loss": 0.9511,
+      "step": 34286
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.891491428571429e-06,
+      "loss": 1.2192,
+      "step": 34287
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.891222857142857e-06,
+      "loss": 1.252,
+      "step": 34288
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8909542857142854e-06,
+      "loss": 0.8195,
+      "step": 34289
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8906857142857145e-06,
+      "loss": 0.7691,
+      "step": 34290
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.890417142857143e-06,
+      "loss": 1.0503,
+      "step": 34291
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.890148571428572e-06,
+      "loss": 1.2077,
+      "step": 34292
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.88988e-06,
+      "loss": 0.9054,
+      "step": 34293
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.889611428571428e-06,
+      "loss": 1.3557,
+      "step": 34294
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.889342857142857e-06,
+      "loss": 1.1245,
+      "step": 34295
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.889074285714286e-06,
+      "loss": 0.9238,
+      "step": 34296
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.888805714285715e-06,
+      "loss": 1.152,
+      "step": 34297
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.888537142857143e-06,
+      "loss": 1.0674,
+      "step": 34298
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8882685714285714e-06,
+      "loss": 0.8832,
+      "step": 34299
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8880000000000005e-06,
+      "loss": 0.9137,
+      "step": 34300
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.887731428571429e-06,
+      "loss": 0.9151,
+      "step": 34301
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.887462857142858e-06,
+      "loss": 1.1623,
+      "step": 34302
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.887194285714286e-06,
+      "loss": 1.3732,
+      "step": 34303
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.886925714285714e-06,
+      "loss": 1.4156,
+      "step": 34304
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.886657142857143e-06,
+      "loss": 0.9463,
+      "step": 34305
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.886388571428571e-06,
+      "loss": 1.0242,
+      "step": 34306
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.88612e-06,
+      "loss": 1.0049,
+      "step": 34307
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.885851428571428e-06,
+      "loss": 1.2299,
+      "step": 34308
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8855828571428574e-06,
+      "loss": 1.2423,
+      "step": 34309
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8853142857142865e-06,
+      "loss": 0.7834,
+      "step": 34310
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.885045714285715e-06,
+      "loss": 0.9285,
+      "step": 34311
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.884777142857143e-06,
+      "loss": 1.1283,
+      "step": 34312
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.884508571428572e-06,
+      "loss": 1.0723,
+      "step": 34313
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.88424e-06,
+      "loss": 1.2019,
+      "step": 34314
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.883971428571429e-06,
+      "loss": 1.2158,
+      "step": 34315
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.883702857142857e-06,
+      "loss": 0.8157,
+      "step": 34316
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.883434285714285e-06,
+      "loss": 1.249,
+      "step": 34317
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.883165714285714e-06,
+      "loss": 1.2184,
+      "step": 34318
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.882897142857143e-06,
+      "loss": 1.2531,
+      "step": 34319
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.882628571428572e-06,
+      "loss": 1.1942,
+      "step": 34320
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.88236e-06,
+      "loss": 0.9193,
+      "step": 34321
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.882091428571429e-06,
+      "loss": 1.2785,
+      "step": 34322
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.881822857142858e-06,
+      "loss": 1.1696,
+      "step": 34323
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.881554285714286e-06,
+      "loss": 0.8246,
+      "step": 34324
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.881285714285715e-06,
+      "loss": 0.7093,
+      "step": 34325
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.881017142857143e-06,
+      "loss": 0.6518,
+      "step": 34326
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.880748571428571e-06,
+      "loss": 1.1702,
+      "step": 34327
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.88048e-06,
+      "loss": 1.0367,
+      "step": 34328
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.880211428571429e-06,
+      "loss": 1.2529,
+      "step": 34329
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.879942857142858e-06,
+      "loss": 1.518,
+      "step": 34330
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.879674285714286e-06,
+      "loss": 0.7375,
+      "step": 34331
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.879405714285714e-06,
+      "loss": 1.2064,
+      "step": 34332
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.879137142857143e-06,
+      "loss": 0.9598,
+      "step": 34333
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.878868571428571e-06,
+      "loss": 1.6993,
+      "step": 34334
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8786e-06,
+      "loss": 0.8467,
+      "step": 34335
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.878331428571429e-06,
+      "loss": 1.2349,
+      "step": 34336
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.878062857142857e-06,
+      "loss": 1.314,
+      "step": 34337
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.877794285714286e-06,
+      "loss": 1.2282,
+      "step": 34338
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.877525714285715e-06,
+      "loss": 0.8904,
+      "step": 34339
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.877257142857143e-06,
+      "loss": 1.0629,
+      "step": 34340
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.876988571428572e-06,
+      "loss": 1.3154,
+      "step": 34341
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.87672e-06,
+      "loss": 1.2404,
+      "step": 34342
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.876451428571429e-06,
+      "loss": 0.4258,
+      "step": 34343
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.876182857142857e-06,
+      "loss": 0.9988,
+      "step": 34344
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.875914285714285e-06,
+      "loss": 1.1005,
+      "step": 34345
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.875645714285714e-06,
+      "loss": 1.417,
+      "step": 34346
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8753771428571425e-06,
+      "loss": 1.2417,
+      "step": 34347
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8751085714285716e-06,
+      "loss": 1.5665,
+      "step": 34348
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.874840000000001e-06,
+      "loss": 1.4961,
+      "step": 34349
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.874571428571429e-06,
+      "loss": 1.0536,
+      "step": 34350
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.874302857142858e-06,
+      "loss": 1.0262,
+      "step": 34351
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.874034285714286e-06,
+      "loss": 1.1505,
+      "step": 34352
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.873765714285714e-06,
+      "loss": 1.0083,
+      "step": 34353
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.873497142857143e-06,
+      "loss": 1.1656,
+      "step": 34354
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.873228571428571e-06,
+      "loss": 0.9868,
+      "step": 34355
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.87296e-06,
+      "loss": 1.4978,
+      "step": 34356
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8726914285714285e-06,
+      "loss": 1.0304,
+      "step": 34357
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.872422857142857e-06,
+      "loss": 1.3896,
+      "step": 34358
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.872154285714286e-06,
+      "loss": 1.039,
+      "step": 34359
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.871885714285714e-06,
+      "loss": 1.2413,
+      "step": 34360
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.871617142857143e-06,
+      "loss": 1.1109,
+      "step": 34361
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.871348571428572e-06,
+      "loss": 1.2465,
+      "step": 34362
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.87108e-06,
+      "loss": 0.9674,
+      "step": 34363
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.870811428571429e-06,
+      "loss": 1.0157,
+      "step": 34364
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.870542857142857e-06,
+      "loss": 0.7765,
+      "step": 34365
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.870274285714286e-06,
+      "loss": 1.2856,
+      "step": 34366
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8700057142857145e-06,
+      "loss": 0.9222,
+      "step": 34367
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.869737142857143e-06,
+      "loss": 0.8831,
+      "step": 34368
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.869468571428572e-06,
+      "loss": 1.0292,
+      "step": 34369
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8692e-06,
+      "loss": 1.1873,
+      "step": 34370
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.868931428571429e-06,
+      "loss": 0.9203,
+      "step": 34371
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.868662857142857e-06,
+      "loss": 1.0729,
+      "step": 34372
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.868394285714285e-06,
+      "loss": 1.0562,
+      "step": 34373
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.868125714285714e-06,
+      "loss": 1.1226,
+      "step": 34374
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.867857142857143e-06,
+      "loss": 1.276,
+      "step": 34375
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8675885714285715e-06,
+      "loss": 1.4505,
+      "step": 34376
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8673200000000005e-06,
+      "loss": 1.0082,
+      "step": 34377
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.867051428571429e-06,
+      "loss": 1.301,
+      "step": 34378
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.866782857142858e-06,
+      "loss": 1.2996,
+      "step": 34379
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.866514285714286e-06,
+      "loss": 0.9744,
+      "step": 34380
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.866245714285714e-06,
+      "loss": 0.8196,
+      "step": 34381
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.865977142857143e-06,
+      "loss": 1.2875,
+      "step": 34382
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.865708571428571e-06,
+      "loss": 1.2376,
+      "step": 34383
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.86544e-06,
+      "loss": 1.1276,
+      "step": 34384
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.8651714285714285e-06,
+      "loss": 0.8333,
+      "step": 34385
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.864902857142857e-06,
+      "loss": 1.1518,
+      "step": 34386
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.864634285714286e-06,
+      "loss": 1.0222,
+      "step": 34387
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.864365714285715e-06,
+      "loss": 0.9748,
+      "step": 34388
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.864097142857144e-06,
+      "loss": 0.9073,
+      "step": 34389
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.863828571428572e-06,
+      "loss": 1.2069,
+      "step": 34390
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.86356e-06,
+      "loss": 1.0124,
+      "step": 34391
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.863291428571429e-06,
+      "loss": 1.4577,
+      "step": 34392
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.863022857142857e-06,
+      "loss": 1.0734,
+      "step": 34393
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.862754285714286e-06,
+      "loss": 1.593,
+      "step": 34394
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8624857142857145e-06,
+      "loss": 1.3785,
+      "step": 34395
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.862217142857143e-06,
+      "loss": 1.1128,
+      "step": 34396
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.861948571428572e-06,
+      "loss": 1.2764,
+      "step": 34397
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.86168e-06,
+      "loss": 0.9817,
+      "step": 34398
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.861411428571428e-06,
+      "loss": 1.108,
+      "step": 34399
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.861142857142857e-06,
+      "loss": 0.9075,
+      "step": 34400
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.860874285714286e-06,
+      "loss": 1.2783,
+      "step": 34401
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.860605714285715e-06,
+      "loss": 1.2621,
+      "step": 34402
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.860337142857143e-06,
+      "loss": 0.882,
+      "step": 34403
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8600685714285714e-06,
+      "loss": 0.9629,
+      "step": 34404
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8598000000000005e-06,
+      "loss": 1.5216,
+      "step": 34405
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.859531428571429e-06,
+      "loss": 1.3887,
+      "step": 34406
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.859262857142858e-06,
+      "loss": 1.3268,
+      "step": 34407
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.858994285714286e-06,
+      "loss": 1.4403,
+      "step": 34408
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.858725714285714e-06,
+      "loss": 1.1886,
+      "step": 34409
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.858457142857143e-06,
+      "loss": 0.9202,
+      "step": 34410
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.858188571428571e-06,
+      "loss": 0.9787,
+      "step": 34411
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.85792e-06,
+      "loss": 1.0372,
+      "step": 34412
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.857651428571428e-06,
+      "loss": 0.9749,
+      "step": 34413
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8573828571428574e-06,
+      "loss": 1.381,
+      "step": 34414
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8571142857142865e-06,
+      "loss": 0.9518,
+      "step": 34415
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.856845714285715e-06,
+      "loss": 0.7708,
+      "step": 34416
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.856577142857144e-06,
+      "loss": 1.2285,
+      "step": 34417
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.856308571428572e-06,
+      "loss": 0.6776,
+      "step": 34418
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.85604e-06,
+      "loss": 1.5025,
+      "step": 34419
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.855771428571429e-06,
+      "loss": 0.6246,
+      "step": 34420
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.855502857142857e-06,
+      "loss": 0.9773,
+      "step": 34421
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.855234285714285e-06,
+      "loss": 0.9857,
+      "step": 34422
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.854965714285714e-06,
+      "loss": 1.2178,
+      "step": 34423
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.854697142857143e-06,
+      "loss": 0.936,
+      "step": 34424
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.854428571428572e-06,
+      "loss": 1.2516,
+      "step": 34425
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.85416e-06,
+      "loss": 1.0481,
+      "step": 34426
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.853891428571429e-06,
+      "loss": 1.0032,
+      "step": 34427
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.853622857142858e-06,
+      "loss": 1.1799,
+      "step": 34428
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.853354285714286e-06,
+      "loss": 0.7627,
+      "step": 34429
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.853085714285715e-06,
+      "loss": 0.9702,
+      "step": 34430
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.852817142857143e-06,
+      "loss": 1.0774,
+      "step": 34431
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.852548571428571e-06,
+      "loss": 1.4315,
+      "step": 34432
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.85228e-06,
+      "loss": 1.0675,
+      "step": 34433
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.852011428571429e-06,
+      "loss": 1.4425,
+      "step": 34434
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.851742857142858e-06,
+      "loss": 1.3709,
+      "step": 34435
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.851474285714286e-06,
+      "loss": 1.0534,
+      "step": 34436
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.851205714285714e-06,
+      "loss": 1.2423,
+      "step": 34437
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.850937142857143e-06,
+      "loss": 1.089,
+      "step": 34438
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.850668571428571e-06,
+      "loss": 0.8043,
+      "step": 34439
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8504e-06,
+      "loss": 1.0172,
+      "step": 34440
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.850131428571429e-06,
+      "loss": 1.2216,
+      "step": 34441
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.849862857142857e-06,
+      "loss": 1.2719,
+      "step": 34442
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.849594285714286e-06,
+      "loss": 1.0707,
+      "step": 34443
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.849325714285715e-06,
+      "loss": 0.7258,
+      "step": 34444
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.849057142857143e-06,
+      "loss": 1.2377,
+      "step": 34445
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.848788571428572e-06,
+      "loss": 1.4485,
+      "step": 34446
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.84852e-06,
+      "loss": 1.7828,
+      "step": 34447
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.848251428571429e-06,
+      "loss": 1.1397,
+      "step": 34448
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.847982857142857e-06,
+      "loss": 0.8873,
+      "step": 34449
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.847714285714285e-06,
+      "loss": 0.9225,
+      "step": 34450
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.847445714285714e-06,
+      "loss": 0.9531,
+      "step": 34451
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8471771428571425e-06,
+      "loss": 1.0827,
+      "step": 34452
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8469085714285716e-06,
+      "loss": 1.2104,
+      "step": 34453
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8466400000000006e-06,
+      "loss": 1.2755,
+      "step": 34454
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.846371428571429e-06,
+      "loss": 1.418,
+      "step": 34455
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.846102857142858e-06,
+      "loss": 0.799,
+      "step": 34456
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.845834285714286e-06,
+      "loss": 1.2987,
+      "step": 34457
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.845565714285715e-06,
+      "loss": 1.3579,
+      "step": 34458
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.845297142857143e-06,
+      "loss": 1.061,
+      "step": 34459
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.845028571428571e-06,
+      "loss": 1.0321,
+      "step": 34460
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.84476e-06,
+      "loss": 0.8277,
+      "step": 34461
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8444914285714285e-06,
+      "loss": 0.9382,
+      "step": 34462
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.844222857142857e-06,
+      "loss": 1.0414,
+      "step": 34463
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.843954285714286e-06,
+      "loss": 1.5347,
+      "step": 34464
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.843685714285714e-06,
+      "loss": 1.1458,
+      "step": 34465
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.843417142857143e-06,
+      "loss": 1.2607,
+      "step": 34466
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.843148571428572e-06,
+      "loss": 0.9789,
+      "step": 34467
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.84288e-06,
+      "loss": 1.5978,
+      "step": 34468
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.842611428571429e-06,
+      "loss": 1.155,
+      "step": 34469
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.842342857142857e-06,
+      "loss": 1.5845,
+      "step": 34470
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.842074285714286e-06,
+      "loss": 1.0438,
+      "step": 34471
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8418057142857145e-06,
+      "loss": 1.1149,
+      "step": 34472
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.841537142857143e-06,
+      "loss": 1.023,
+      "step": 34473
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.841268571428572e-06,
+      "loss": 1.0752,
+      "step": 34474
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.841e-06,
+      "loss": 1.3942,
+      "step": 34475
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.840731428571429e-06,
+      "loss": 1.3581,
+      "step": 34476
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.840462857142857e-06,
+      "loss": 1.4611,
+      "step": 34477
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.840194285714285e-06,
+      "loss": 1.6745,
+      "step": 34478
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.839925714285714e-06,
+      "loss": 0.9784,
+      "step": 34479
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.839657142857143e-06,
+      "loss": 1.5976,
+      "step": 34480
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.839388571428572e-06,
+      "loss": 1.4744,
+      "step": 34481
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8391200000000005e-06,
+      "loss": 1.6952,
+      "step": 34482
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.838851428571429e-06,
+      "loss": 1.1339,
+      "step": 34483
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.838582857142858e-06,
+      "loss": 0.985,
+      "step": 34484
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.838314285714286e-06,
+      "loss": 1.0543,
+      "step": 34485
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.838045714285714e-06,
+      "loss": 1.33,
+      "step": 34486
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.837777142857143e-06,
+      "loss": 1.6643,
+      "step": 34487
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.837508571428571e-06,
+      "loss": 1.1049,
+      "step": 34488
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.83724e-06,
+      "loss": 1.1494,
+      "step": 34489
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8369714285714285e-06,
+      "loss": 1.0683,
+      "step": 34490
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.836702857142857e-06,
+      "loss": 0.8252,
+      "step": 34491
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.836434285714286e-06,
+      "loss": 0.8642,
+      "step": 34492
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.836165714285715e-06,
+      "loss": 0.9624,
+      "step": 34493
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.835897142857144e-06,
+      "loss": 0.9746,
+      "step": 34494
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.835628571428572e-06,
+      "loss": 1.278,
+      "step": 34495
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.83536e-06,
+      "loss": 1.004,
+      "step": 34496
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.835091428571429e-06,
+      "loss": 0.7057,
+      "step": 34497
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.834822857142857e-06,
+      "loss": 1.0125,
+      "step": 34498
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.834554285714286e-06,
+      "loss": 1.2172,
+      "step": 34499
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8342857142857145e-06,
+      "loss": 1.5062,
+      "step": 34500
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.834017142857143e-06,
+      "loss": 1.2029,
+      "step": 34501
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.833748571428572e-06,
+      "loss": 1.4051,
+      "step": 34502
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.83348e-06,
+      "loss": 1.1523,
+      "step": 34503
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.833211428571429e-06,
+      "loss": 1.3525,
+      "step": 34504
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.832942857142857e-06,
+      "loss": 0.9188,
+      "step": 34505
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.832674285714286e-06,
+      "loss": 1.4932,
+      "step": 34506
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.832405714285715e-06,
+      "loss": 1.0403,
+      "step": 34507
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.832137142857143e-06,
+      "loss": 0.9799,
+      "step": 34508
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8318685714285714e-06,
+      "loss": 0.8859,
+      "step": 34509
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8316000000000005e-06,
+      "loss": 1.5865,
+      "step": 34510
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.831331428571429e-06,
+      "loss": 1.2251,
+      "step": 34511
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.831062857142858e-06,
+      "loss": 1.0592,
+      "step": 34512
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.830794285714286e-06,
+      "loss": 1.234,
+      "step": 34513
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.830525714285714e-06,
+      "loss": 0.9462,
+      "step": 34514
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.830257142857143e-06,
+      "loss": 1.0496,
+      "step": 34515
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.829988571428571e-06,
+      "loss": 1.107,
+      "step": 34516
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.82972e-06,
+      "loss": 1.1535,
+      "step": 34517
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.829451428571428e-06,
+      "loss": 1.4106,
+      "step": 34518
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8291828571428574e-06,
+      "loss": 0.9902,
+      "step": 34519
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8289142857142865e-06,
+      "loss": 1.1619,
+      "step": 34520
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.828645714285715e-06,
+      "loss": 1.1673,
+      "step": 34521
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.828377142857144e-06,
+      "loss": 0.7912,
+      "step": 34522
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.828108571428572e-06,
+      "loss": 1.1999,
+      "step": 34523
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.82784e-06,
+      "loss": 1.3062,
+      "step": 34524
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.827571428571429e-06,
+      "loss": 1.1818,
+      "step": 34525
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.827302857142857e-06,
+      "loss": 0.7521,
+      "step": 34526
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.827034285714285e-06,
+      "loss": 0.8998,
+      "step": 34527
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.826765714285714e-06,
+      "loss": 1.3237,
+      "step": 34528
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.826497142857143e-06,
+      "loss": 1.2766,
+      "step": 34529
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.826228571428572e-06,
+      "loss": 0.7975,
+      "step": 34530
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.82596e-06,
+      "loss": 1.5983,
+      "step": 34531
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.825691428571429e-06,
+      "loss": 1.3574,
+      "step": 34532
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.825422857142858e-06,
+      "loss": 0.9786,
+      "step": 34533
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.825154285714286e-06,
+      "loss": 1.128,
+      "step": 34534
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.824885714285715e-06,
+      "loss": 0.8966,
+      "step": 34535
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.824617142857143e-06,
+      "loss": 1.4362,
+      "step": 34536
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.824348571428571e-06,
+      "loss": 1.1645,
+      "step": 34537
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.82408e-06,
+      "loss": 1.2327,
+      "step": 34538
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.823811428571429e-06,
+      "loss": 1.1465,
+      "step": 34539
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.823542857142858e-06,
+      "loss": 1.0475,
+      "step": 34540
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.823274285714286e-06,
+      "loss": 0.7569,
+      "step": 34541
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.823005714285714e-06,
+      "loss": 0.9064,
+      "step": 34542
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.822737142857143e-06,
+      "loss": 1.1317,
+      "step": 34543
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.822468571428571e-06,
+      "loss": 1.23,
+      "step": 34544
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8222e-06,
+      "loss": 1.4584,
+      "step": 34545
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.821931428571429e-06,
+      "loss": 0.9743,
+      "step": 34546
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.821662857142857e-06,
+      "loss": 1.2672,
+      "step": 34547
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.821394285714286e-06,
+      "loss": 1.2858,
+      "step": 34548
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.821125714285715e-06,
+      "loss": 1.4196,
+      "step": 34549
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.820857142857143e-06,
+      "loss": 1.0271,
+      "step": 34550
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.820588571428572e-06,
+      "loss": 1.136,
+      "step": 34551
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.82032e-06,
+      "loss": 1.2832,
+      "step": 34552
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.820051428571429e-06,
+      "loss": 0.888,
+      "step": 34553
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.819782857142857e-06,
+      "loss": 1.3085,
+      "step": 34554
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.819514285714285e-06,
+      "loss": 1.1707,
+      "step": 34555
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.819245714285714e-06,
+      "loss": 0.848,
+      "step": 34556
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8189771428571425e-06,
+      "loss": 1.1495,
+      "step": 34557
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8187085714285716e-06,
+      "loss": 0.7731,
+      "step": 34558
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8184400000000006e-06,
+      "loss": 1.393,
+      "step": 34559
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.818171428571429e-06,
+      "loss": 0.8921,
+      "step": 34560
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.817902857142858e-06,
+      "loss": 1.0182,
+      "step": 34561
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.817634285714286e-06,
+      "loss": 1.1704,
+      "step": 34562
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.817365714285715e-06,
+      "loss": 0.9727,
+      "step": 34563
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.817097142857143e-06,
+      "loss": 1.0008,
+      "step": 34564
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.816828571428571e-06,
+      "loss": 1.1564,
+      "step": 34565
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.81656e-06,
+      "loss": 0.9008,
+      "step": 34566
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8162914285714285e-06,
+      "loss": 1.7492,
+      "step": 34567
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8160228571428575e-06,
+      "loss": 0.8962,
+      "step": 34568
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.815754285714286e-06,
+      "loss": 1.1004,
+      "step": 34569
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.815485714285714e-06,
+      "loss": 1.4327,
+      "step": 34570
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.815217142857143e-06,
+      "loss": 1.2264,
+      "step": 34571
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.814948571428572e-06,
+      "loss": 0.5406,
+      "step": 34572
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.81468e-06,
+      "loss": 1.1021,
+      "step": 34573
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.814411428571429e-06,
+      "loss": 1.0222,
+      "step": 34574
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.814142857142857e-06,
+      "loss": 0.8652,
+      "step": 34575
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.813874285714286e-06,
+      "loss": 0.9149,
+      "step": 34576
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8136057142857145e-06,
+      "loss": 0.7266,
+      "step": 34577
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.813337142857143e-06,
+      "loss": 1.3424,
+      "step": 34578
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.813068571428572e-06,
+      "loss": 1.4162,
+      "step": 34579
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8128e-06,
+      "loss": 1.074,
+      "step": 34580
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.812531428571429e-06,
+      "loss": 1.1796,
+      "step": 34581
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.812262857142857e-06,
+      "loss": 0.9511,
+      "step": 34582
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.811994285714285e-06,
+      "loss": 0.7198,
+      "step": 34583
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.811725714285714e-06,
+      "loss": 1.0673,
+      "step": 34584
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.811457142857143e-06,
+      "loss": 1.0874,
+      "step": 34585
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.811188571428572e-06,
+      "loss": 1.5075,
+      "step": 34586
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8109200000000005e-06,
+      "loss": 1.3729,
+      "step": 34587
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.810651428571429e-06,
+      "loss": 0.7398,
+      "step": 34588
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.810382857142858e-06,
+      "loss": 1.1381,
+      "step": 34589
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.810114285714286e-06,
+      "loss": 1.0662,
+      "step": 34590
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.809845714285715e-06,
+      "loss": 1.0379,
+      "step": 34591
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.809577142857143e-06,
+      "loss": 1.3915,
+      "step": 34592
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.809308571428571e-06,
+      "loss": 1.0438,
+      "step": 34593
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.80904e-06,
+      "loss": 0.9883,
+      "step": 34594
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8087714285714285e-06,
+      "loss": 1.4037,
+      "step": 34595
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.808502857142857e-06,
+      "loss": 1.1258,
+      "step": 34596
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.808234285714286e-06,
+      "loss": 1.4028,
+      "step": 34597
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.807965714285715e-06,
+      "loss": 1.4175,
+      "step": 34598
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.807697142857144e-06,
+      "loss": 1.0263,
+      "step": 34599
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.807428571428572e-06,
+      "loss": 1.2502,
+      "step": 34600
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.80716e-06,
+      "loss": 1.3365,
+      "step": 34601
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.806891428571429e-06,
+      "loss": 1.1956,
+      "step": 34602
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.806622857142857e-06,
+      "loss": 0.8417,
+      "step": 34603
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.806354285714286e-06,
+      "loss": 0.6598,
+      "step": 34604
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8060857142857145e-06,
+      "loss": 0.8905,
+      "step": 34605
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.805817142857143e-06,
+      "loss": 1.3476,
+      "step": 34606
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.805548571428572e-06,
+      "loss": 1.2178,
+      "step": 34607
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.80528e-06,
+      "loss": 1.2728,
+      "step": 34608
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.805011428571429e-06,
+      "loss": 0.8269,
+      "step": 34609
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.804742857142857e-06,
+      "loss": 1.2877,
+      "step": 34610
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.804474285714286e-06,
+      "loss": 1.1971,
+      "step": 34611
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.804205714285715e-06,
+      "loss": 1.1477,
+      "step": 34612
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.803937142857143e-06,
+      "loss": 1.2614,
+      "step": 34613
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8036685714285714e-06,
+      "loss": 0.9776,
+      "step": 34614
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8034000000000005e-06,
+      "loss": 1.3194,
+      "step": 34615
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.803131428571429e-06,
+      "loss": 0.9242,
+      "step": 34616
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.802862857142858e-06,
+      "loss": 1.2676,
+      "step": 34617
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.802594285714286e-06,
+      "loss": 1.3692,
+      "step": 34618
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.802325714285714e-06,
+      "loss": 1.3893,
+      "step": 34619
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.802057142857143e-06,
+      "loss": 1.0842,
+      "step": 34620
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.801788571428571e-06,
+      "loss": 0.7859,
+      "step": 34621
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.80152e-06,
+      "loss": 1.3247,
+      "step": 34622
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.801251428571428e-06,
+      "loss": 1.6036,
+      "step": 34623
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.800982857142857e-06,
+      "loss": 1.0248,
+      "step": 34624
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.8007142857142865e-06,
+      "loss": 1.1404,
+      "step": 34625
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.800445714285715e-06,
+      "loss": 0.9227,
+      "step": 34626
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.800177142857144e-06,
+      "loss": 1.6966,
+      "step": 34627
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.799908571428572e-06,
+      "loss": 1.3513,
+      "step": 34628
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.79964e-06,
+      "loss": 1.1869,
+      "step": 34629
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.799371428571429e-06,
+      "loss": 1.3621,
+      "step": 34630
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.799102857142857e-06,
+      "loss": 1.3869,
+      "step": 34631
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.798834285714286e-06,
+      "loss": 1.1689,
+      "step": 34632
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.798565714285714e-06,
+      "loss": 1.4001,
+      "step": 34633
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.798297142857143e-06,
+      "loss": 1.21,
+      "step": 34634
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.798028571428572e-06,
+      "loss": 0.833,
+      "step": 34635
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.79776e-06,
+      "loss": 0.9957,
+      "step": 34636
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.797491428571428e-06,
+      "loss": 1.0282,
+      "step": 34637
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.797222857142858e-06,
+      "loss": 1.1895,
+      "step": 34638
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.796954285714286e-06,
+      "loss": 1.0514,
+      "step": 34639
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.796685714285715e-06,
+      "loss": 0.9631,
+      "step": 34640
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.796417142857143e-06,
+      "loss": 1.1866,
+      "step": 34641
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.796148571428571e-06,
+      "loss": 0.9118,
+      "step": 34642
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.79588e-06,
+      "loss": 0.8911,
+      "step": 34643
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.795611428571429e-06,
+      "loss": 1.2423,
+      "step": 34644
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.795342857142858e-06,
+      "loss": 1.0303,
+      "step": 34645
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.795074285714286e-06,
+      "loss": 1.305,
+      "step": 34646
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.794805714285714e-06,
+      "loss": 0.5862,
+      "step": 34647
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.794537142857143e-06,
+      "loss": 1.3791,
+      "step": 34648
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.794268571428571e-06,
+      "loss": 0.9448,
+      "step": 34649
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.794e-06,
+      "loss": 1.4531,
+      "step": 34650
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.793731428571429e-06,
+      "loss": 1.1921,
+      "step": 34651
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.793462857142857e-06,
+      "loss": 1.0111,
+      "step": 34652
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.793194285714286e-06,
+      "loss": 1.289,
+      "step": 34653
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.792925714285715e-06,
+      "loss": 1.0539,
+      "step": 34654
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.792657142857144e-06,
+      "loss": 1.0074,
+      "step": 34655
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.792388571428572e-06,
+      "loss": 1.0751,
+      "step": 34656
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.79212e-06,
+      "loss": 1.3195,
+      "step": 34657
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.791851428571429e-06,
+      "loss": 0.8682,
+      "step": 34658
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.791582857142857e-06,
+      "loss": 0.6896,
+      "step": 34659
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.791314285714285e-06,
+      "loss": 1.115,
+      "step": 34660
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.791045714285714e-06,
+      "loss": 1.2856,
+      "step": 34661
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7907771428571425e-06,
+      "loss": 1.1206,
+      "step": 34662
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7905085714285716e-06,
+      "loss": 1.1089,
+      "step": 34663
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7902400000000006e-06,
+      "loss": 0.9316,
+      "step": 34664
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.789971428571429e-06,
+      "loss": 1.2093,
+      "step": 34665
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.789702857142858e-06,
+      "loss": 1.1955,
+      "step": 34666
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.789434285714286e-06,
+      "loss": 1.5312,
+      "step": 34667
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.789165714285715e-06,
+      "loss": 0.7756,
+      "step": 34668
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.788897142857143e-06,
+      "loss": 1.6882,
+      "step": 34669
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.788628571428571e-06,
+      "loss": 1.2353,
+      "step": 34670
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.78836e-06,
+      "loss": 0.8705,
+      "step": 34671
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7880914285714285e-06,
+      "loss": 0.8317,
+      "step": 34672
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7878228571428575e-06,
+      "loss": 0.7733,
+      "step": 34673
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.787554285714286e-06,
+      "loss": 0.9417,
+      "step": 34674
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.787285714285714e-06,
+      "loss": 0.7523,
+      "step": 34675
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.787017142857143e-06,
+      "loss": 0.9589,
+      "step": 34676
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.786748571428571e-06,
+      "loss": 1.4475,
+      "step": 34677
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.786480000000001e-06,
+      "loss": 1.1235,
+      "step": 34678
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.786211428571429e-06,
+      "loss": 1.0422,
+      "step": 34679
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.785942857142857e-06,
+      "loss": 1.1733,
+      "step": 34680
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.785674285714286e-06,
+      "loss": 1.4947,
+      "step": 34681
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7854057142857145e-06,
+      "loss": 0.6631,
+      "step": 34682
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.785137142857143e-06,
+      "loss": 0.8064,
+      "step": 34683
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.784868571428572e-06,
+      "loss": 1.0248,
+      "step": 34684
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7846e-06,
+      "loss": 1.4946,
+      "step": 34685
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.784331428571429e-06,
+      "loss": 1.1621,
+      "step": 34686
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.784062857142857e-06,
+      "loss": 1.1135,
+      "step": 34687
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.783794285714285e-06,
+      "loss": 0.936,
+      "step": 34688
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.783525714285714e-06,
+      "loss": 1.0417,
+      "step": 34689
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7832571428571425e-06,
+      "loss": 1.2462,
+      "step": 34690
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.782988571428572e-06,
+      "loss": 0.8101,
+      "step": 34691
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7827200000000005e-06,
+      "loss": 1.0862,
+      "step": 34692
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.782451428571429e-06,
+      "loss": 0.8001,
+      "step": 34693
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.782182857142858e-06,
+      "loss": 1.0264,
+      "step": 34694
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.781914285714286e-06,
+      "loss": 1.103,
+      "step": 34695
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.781645714285715e-06,
+      "loss": 0.6697,
+      "step": 34696
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.781377142857143e-06,
+      "loss": 1.1051,
+      "step": 34697
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.781108571428571e-06,
+      "loss": 0.9163,
+      "step": 34698
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.78084e-06,
+      "loss": 1.2171,
+      "step": 34699
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7805714285714285e-06,
+      "loss": 1.3922,
+      "step": 34700
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.780302857142857e-06,
+      "loss": 1.1089,
+      "step": 34701
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.780034285714286e-06,
+      "loss": 0.8967,
+      "step": 34702
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.779765714285714e-06,
+      "loss": 1.3523,
+      "step": 34703
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.779497142857144e-06,
+      "loss": 0.9176,
+      "step": 34704
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.779228571428572e-06,
+      "loss": 1.2633,
+      "step": 34705
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.77896e-06,
+      "loss": 1.1228,
+      "step": 34706
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.778691428571429e-06,
+      "loss": 1.2763,
+      "step": 34707
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.778422857142857e-06,
+      "loss": 1.1839,
+      "step": 34708
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.778154285714286e-06,
+      "loss": 1.2705,
+      "step": 34709
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7778857142857145e-06,
+      "loss": 1.0429,
+      "step": 34710
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.777617142857143e-06,
+      "loss": 0.7396,
+      "step": 34711
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.777348571428572e-06,
+      "loss": 1.6811,
+      "step": 34712
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.77708e-06,
+      "loss": 0.7321,
+      "step": 34713
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.776811428571429e-06,
+      "loss": 0.7231,
+      "step": 34714
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.776542857142857e-06,
+      "loss": 0.6326,
+      "step": 34715
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.776274285714285e-06,
+      "loss": 0.8825,
+      "step": 34716
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.776005714285715e-06,
+      "loss": 1.2336,
+      "step": 34717
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.775737142857143e-06,
+      "loss": 1.3137,
+      "step": 34718
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.775468571428572e-06,
+      "loss": 1.1439,
+      "step": 34719
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.7752000000000005e-06,
+      "loss": 0.8246,
+      "step": 34720
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.774931428571429e-06,
+      "loss": 1.3442,
+      "step": 34721
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.774662857142858e-06,
+      "loss": 1.2696,
+      "step": 34722
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.774394285714286e-06,
+      "loss": 1.3726,
+      "step": 34723
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.774125714285714e-06,
+      "loss": 1.3259,
+      "step": 34724
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.773857142857143e-06,
+      "loss": 1.2208,
+      "step": 34725
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.773588571428571e-06,
+      "loss": 1.2062,
+      "step": 34726
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.77332e-06,
+      "loss": 1.2253,
+      "step": 34727
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.773051428571428e-06,
+      "loss": 1.241,
+      "step": 34728
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.772782857142857e-06,
+      "loss": 1.2126,
+      "step": 34729
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.772514285714286e-06,
+      "loss": 1.4299,
+      "step": 34730
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.772245714285715e-06,
+      "loss": 1.1068,
+      "step": 34731
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.771977142857144e-06,
+      "loss": 1.4451,
+      "step": 34732
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.771708571428572e-06,
+      "loss": 1.1658,
+      "step": 34733
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.77144e-06,
+      "loss": 0.6257,
+      "step": 34734
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.771171428571429e-06,
+      "loss": 1.185,
+      "step": 34735
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.770902857142857e-06,
+      "loss": 0.6725,
+      "step": 34736
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.770634285714286e-06,
+      "loss": 1.0548,
+      "step": 34737
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.770365714285714e-06,
+      "loss": 1.1028,
+      "step": 34738
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.770097142857143e-06,
+      "loss": 1.2225,
+      "step": 34739
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.769828571428572e-06,
+      "loss": 1.8109,
+      "step": 34740
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.76956e-06,
+      "loss": 1.1147,
+      "step": 34741
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.769291428571429e-06,
+      "loss": 1.2156,
+      "step": 34742
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.769022857142857e-06,
+      "loss": 0.8733,
+      "step": 34743
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.768754285714286e-06,
+      "loss": 1.4213,
+      "step": 34744
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.768485714285715e-06,
+      "loss": 1.3358,
+      "step": 34745
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.768217142857143e-06,
+      "loss": 1.2415,
+      "step": 34746
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.767948571428571e-06,
+      "loss": 0.5459,
+      "step": 34747
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.76768e-06,
+      "loss": 0.6697,
+      "step": 34748
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.767411428571429e-06,
+      "loss": 1.1151,
+      "step": 34749
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.767142857142858e-06,
+      "loss": 1.4896,
+      "step": 34750
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.766874285714286e-06,
+      "loss": 0.8981,
+      "step": 34751
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.766605714285714e-06,
+      "loss": 1.3006,
+      "step": 34752
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.766337142857143e-06,
+      "loss": 1.216,
+      "step": 34753
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.766068571428571e-06,
+      "loss": 0.5187,
+      "step": 34754
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7658e-06,
+      "loss": 1.1156,
+      "step": 34755
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.765531428571428e-06,
+      "loss": 1.046,
+      "step": 34756
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.765262857142857e-06,
+      "loss": 0.9467,
+      "step": 34757
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.764994285714286e-06,
+      "loss": 1.0702,
+      "step": 34758
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.764725714285715e-06,
+      "loss": 0.7914,
+      "step": 34759
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.764457142857144e-06,
+      "loss": 0.9786,
+      "step": 34760
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.764188571428572e-06,
+      "loss": 1.3441,
+      "step": 34761
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.76392e-06,
+      "loss": 1.276,
+      "step": 34762
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.763651428571429e-06,
+      "loss": 1.2936,
+      "step": 34763
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.763382857142857e-06,
+      "loss": 1.516,
+      "step": 34764
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.763114285714285e-06,
+      "loss": 1.2275,
+      "step": 34765
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.762845714285714e-06,
+      "loss": 1.5941,
+      "step": 34766
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7625771428571425e-06,
+      "loss": 1.1001,
+      "step": 34767
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7623085714285715e-06,
+      "loss": 1.2352,
+      "step": 34768
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.76204e-06,
+      "loss": 1.0387,
+      "step": 34769
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.761771428571429e-06,
+      "loss": 1.1109,
+      "step": 34770
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.761502857142858e-06,
+      "loss": 1.2065,
+      "step": 34771
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.761234285714286e-06,
+      "loss": 1.2503,
+      "step": 34772
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.760965714285715e-06,
+      "loss": 0.9926,
+      "step": 34773
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.760697142857143e-06,
+      "loss": 1.1074,
+      "step": 34774
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.760428571428571e-06,
+      "loss": 1.1889,
+      "step": 34775
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.76016e-06,
+      "loss": 0.8525,
+      "step": 34776
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7598914285714285e-06,
+      "loss": 0.8802,
+      "step": 34777
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7596228571428575e-06,
+      "loss": 1.2442,
+      "step": 34778
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.759354285714286e-06,
+      "loss": 1.2359,
+      "step": 34779
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.759085714285714e-06,
+      "loss": 1.3516,
+      "step": 34780
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.758817142857143e-06,
+      "loss": 1.1215,
+      "step": 34781
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.758548571428571e-06,
+      "loss": 0.6687,
+      "step": 34782
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.75828e-06,
+      "loss": 0.961,
+      "step": 34783
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.758011428571429e-06,
+      "loss": 1.4706,
+      "step": 34784
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.757742857142857e-06,
+      "loss": 1.2694,
+      "step": 34785
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.757474285714286e-06,
+      "loss": 1.1785,
+      "step": 34786
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7572057142857145e-06,
+      "loss": 1.5823,
+      "step": 34787
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.756937142857143e-06,
+      "loss": 0.9962,
+      "step": 34788
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.756668571428572e-06,
+      "loss": 1.123,
+      "step": 34789
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7564e-06,
+      "loss": 1.2064,
+      "step": 34790
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.756131428571429e-06,
+      "loss": 1.3802,
+      "step": 34791
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.755862857142857e-06,
+      "loss": 1.1777,
+      "step": 34792
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.755594285714285e-06,
+      "loss": 0.9193,
+      "step": 34793
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.755325714285714e-06,
+      "loss": 1.1942,
+      "step": 34794
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7550571428571425e-06,
+      "loss": 1.1334,
+      "step": 34795
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7547885714285715e-06,
+      "loss": 1.3584,
+      "step": 34796
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7545200000000005e-06,
+      "loss": 0.8398,
+      "step": 34797
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.754251428571429e-06,
+      "loss": 1.1949,
+      "step": 34798
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.753982857142858e-06,
+      "loss": 1.3687,
+      "step": 34799
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.753714285714286e-06,
+      "loss": 1.0196,
+      "step": 34800
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.753445714285715e-06,
+      "loss": 1.0204,
+      "step": 34801
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.753177142857143e-06,
+      "loss": 0.6104,
+      "step": 34802
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.752908571428571e-06,
+      "loss": 1.0064,
+      "step": 34803
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.75264e-06,
+      "loss": 1.4621,
+      "step": 34804
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7523714285714285e-06,
+      "loss": 1.001,
+      "step": 34805
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7521028571428575e-06,
+      "loss": 0.9703,
+      "step": 34806
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.751834285714286e-06,
+      "loss": 1.4577,
+      "step": 34807
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.751565714285714e-06,
+      "loss": 0.8787,
+      "step": 34808
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.751297142857143e-06,
+      "loss": 1.168,
+      "step": 34809
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.751028571428572e-06,
+      "loss": 1.6945,
+      "step": 34810
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.75076e-06,
+      "loss": 1.3527,
+      "step": 34811
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.750491428571429e-06,
+      "loss": 0.7783,
+      "step": 34812
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.750222857142857e-06,
+      "loss": 1.1534,
+      "step": 34813
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.749954285714286e-06,
+      "loss": 0.6801,
+      "step": 34814
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7496857142857145e-06,
+      "loss": 1.0074,
+      "step": 34815
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.749417142857143e-06,
+      "loss": 1.2179,
+      "step": 34816
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.749148571428572e-06,
+      "loss": 1.6726,
+      "step": 34817
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.74888e-06,
+      "loss": 1.2523,
+      "step": 34818
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.748611428571429e-06,
+      "loss": 1.3653,
+      "step": 34819
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.748342857142857e-06,
+      "loss": 1.1251,
+      "step": 34820
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.748074285714285e-06,
+      "loss": 0.8957,
+      "step": 34821
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.747805714285714e-06,
+      "loss": 1.2767,
+      "step": 34822
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.747537142857143e-06,
+      "loss": 1.2057,
+      "step": 34823
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.747268571428572e-06,
+      "loss": 0.9439,
+      "step": 34824
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7470000000000005e-06,
+      "loss": 0.9144,
+      "step": 34825
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.746731428571429e-06,
+      "loss": 1.0466,
+      "step": 34826
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.746462857142858e-06,
+      "loss": 1.425,
+      "step": 34827
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.746194285714286e-06,
+      "loss": 1.0083,
+      "step": 34828
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.745925714285715e-06,
+      "loss": 1.2287,
+      "step": 34829
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.745657142857143e-06,
+      "loss": 1.5235,
+      "step": 34830
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.745388571428571e-06,
+      "loss": 1.4149,
+      "step": 34831
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.74512e-06,
+      "loss": 1.0216,
+      "step": 34832
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.744851428571428e-06,
+      "loss": 1.2999,
+      "step": 34833
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.744582857142857e-06,
+      "loss": 1.2806,
+      "step": 34834
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.744314285714286e-06,
+      "loss": 0.96,
+      "step": 34835
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.744045714285715e-06,
+      "loss": 0.889,
+      "step": 34836
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.743777142857144e-06,
+      "loss": 1.2005,
+      "step": 34837
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.743508571428572e-06,
+      "loss": 1.4307,
+      "step": 34838
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.74324e-06,
+      "loss": 0.9547,
+      "step": 34839
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.742971428571429e-06,
+      "loss": 1.2455,
+      "step": 34840
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.742702857142857e-06,
+      "loss": 1.2788,
+      "step": 34841
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.742434285714286e-06,
+      "loss": 1.1608,
+      "step": 34842
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.742165714285714e-06,
+      "loss": 0.8259,
+      "step": 34843
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.741897142857143e-06,
+      "loss": 0.9621,
+      "step": 34844
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.741628571428572e-06,
+      "loss": 1.4572,
+      "step": 34845
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.74136e-06,
+      "loss": 1.4769,
+      "step": 34846
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.741091428571429e-06,
+      "loss": 1.0581,
+      "step": 34847
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.740822857142857e-06,
+      "loss": 1.2885,
+      "step": 34848
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.740554285714286e-06,
+      "loss": 1.2998,
+      "step": 34849
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.740285714285715e-06,
+      "loss": 0.8864,
+      "step": 34850
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.740017142857143e-06,
+      "loss": 1.0568,
+      "step": 34851
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.739748571428571e-06,
+      "loss": 1.2544,
+      "step": 34852
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.73948e-06,
+      "loss": 0.9702,
+      "step": 34853
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.739211428571429e-06,
+      "loss": 0.7177,
+      "step": 34854
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.738942857142858e-06,
+      "loss": 1.2044,
+      "step": 34855
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.738674285714286e-06,
+      "loss": 0.9601,
+      "step": 34856
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.738405714285714e-06,
+      "loss": 0.996,
+      "step": 34857
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.738137142857143e-06,
+      "loss": 0.8741,
+      "step": 34858
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.737868571428571e-06,
+      "loss": 1.155,
+      "step": 34859
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7376e-06,
+      "loss": 1.1555,
+      "step": 34860
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.737331428571428e-06,
+      "loss": 0.9342,
+      "step": 34861
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.737062857142857e-06,
+      "loss": 1.1901,
+      "step": 34862
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.736794285714286e-06,
+      "loss": 1.3397,
+      "step": 34863
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7365257142857146e-06,
+      "loss": 1.3248,
+      "step": 34864
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.736257142857144e-06,
+      "loss": 1.1251,
+      "step": 34865
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.735988571428572e-06,
+      "loss": 0.9771,
+      "step": 34866
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.73572e-06,
+      "loss": 1.5597,
+      "step": 34867
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.735451428571429e-06,
+      "loss": 1.0922,
+      "step": 34868
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.735182857142857e-06,
+      "loss": 1.2139,
+      "step": 34869
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.734914285714286e-06,
+      "loss": 0.8594,
+      "step": 34870
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.734645714285714e-06,
+      "loss": 0.8546,
+      "step": 34871
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7343771428571425e-06,
+      "loss": 1.1384,
+      "step": 34872
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7341085714285715e-06,
+      "loss": 1.1606,
+      "step": 34873
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.73384e-06,
+      "loss": 1.56,
+      "step": 34874
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.733571428571429e-06,
+      "loss": 1.2772,
+      "step": 34875
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.733302857142858e-06,
+      "loss": 0.9695,
+      "step": 34876
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.733034285714286e-06,
+      "loss": 1.1566,
+      "step": 34877
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.732765714285715e-06,
+      "loss": 1.3242,
+      "step": 34878
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.732497142857143e-06,
+      "loss": 0.9041,
+      "step": 34879
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.732228571428571e-06,
+      "loss": 0.9735,
+      "step": 34880
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.73196e-06,
+      "loss": 0.7637,
+      "step": 34881
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7316914285714285e-06,
+      "loss": 1.1068,
+      "step": 34882
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7314228571428575e-06,
+      "loss": 0.9899,
+      "step": 34883
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.731154285714286e-06,
+      "loss": 1.8998,
+      "step": 34884
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.730885714285714e-06,
+      "loss": 1.2504,
+      "step": 34885
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.730617142857143e-06,
+      "loss": 1.1375,
+      "step": 34886
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.730348571428571e-06,
+      "loss": 1.0355,
+      "step": 34887
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.73008e-06,
+      "loss": 1.1337,
+      "step": 34888
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.729811428571429e-06,
+      "loss": 1.0365,
+      "step": 34889
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.729542857142857e-06,
+      "loss": 1.1677,
+      "step": 34890
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.729274285714286e-06,
+      "loss": 0.8064,
+      "step": 34891
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7290057142857145e-06,
+      "loss": 1.2386,
+      "step": 34892
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7287371428571435e-06,
+      "loss": 0.9107,
+      "step": 34893
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.728468571428572e-06,
+      "loss": 1.423,
+      "step": 34894
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7282e-06,
+      "loss": 1.397,
+      "step": 34895
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.727931428571429e-06,
+      "loss": 1.38,
+      "step": 34896
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.727662857142857e-06,
+      "loss": 1.2214,
+      "step": 34897
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.727394285714285e-06,
+      "loss": 1.0762,
+      "step": 34898
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.727125714285714e-06,
+      "loss": 1.2948,
+      "step": 34899
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7268571428571425e-06,
+      "loss": 1.0851,
+      "step": 34900
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7265885714285715e-06,
+      "loss": 1.2592,
+      "step": 34901
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7263200000000005e-06,
+      "loss": 1.008,
+      "step": 34902
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.726051428571429e-06,
+      "loss": 1.0807,
+      "step": 34903
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.725782857142858e-06,
+      "loss": 1.0047,
+      "step": 34904
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.725514285714286e-06,
+      "loss": 1.5043,
+      "step": 34905
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.725245714285715e-06,
+      "loss": 1.4424,
+      "step": 34906
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.724977142857143e-06,
+      "loss": 1.1948,
+      "step": 34907
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.724708571428571e-06,
+      "loss": 1.1862,
+      "step": 34908
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.72444e-06,
+      "loss": 1.1941,
+      "step": 34909
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7241714285714285e-06,
+      "loss": 1.1342,
+      "step": 34910
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7239028571428575e-06,
+      "loss": 0.9492,
+      "step": 34911
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.723634285714286e-06,
+      "loss": 0.9724,
+      "step": 34912
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.723365714285714e-06,
+      "loss": 1.2368,
+      "step": 34913
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.723097142857143e-06,
+      "loss": 1.1483,
+      "step": 34914
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.722828571428572e-06,
+      "loss": 1.3146,
+      "step": 34915
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.722560000000001e-06,
+      "loss": 0.9416,
+      "step": 34916
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.722291428571429e-06,
+      "loss": 0.9648,
+      "step": 34917
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.722022857142857e-06,
+      "loss": 0.9957,
+      "step": 34918
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.721754285714286e-06,
+      "loss": 1.2595,
+      "step": 34919
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7214857142857145e-06,
+      "loss": 1.223,
+      "step": 34920
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.721217142857143e-06,
+      "loss": 1.4835,
+      "step": 34921
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.720948571428572e-06,
+      "loss": 0.8603,
+      "step": 34922
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.72068e-06,
+      "loss": 1.3104,
+      "step": 34923
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.720411428571429e-06,
+      "loss": 0.959,
+      "step": 34924
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.720142857142857e-06,
+      "loss": 0.7662,
+      "step": 34925
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.719874285714285e-06,
+      "loss": 1.5056,
+      "step": 34926
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.719605714285714e-06,
+      "loss": 1.3891,
+      "step": 34927
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.719337142857143e-06,
+      "loss": 0.7155,
+      "step": 34928
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.719068571428572e-06,
+      "loss": 0.883,
+      "step": 34929
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7188000000000005e-06,
+      "loss": 0.8595,
+      "step": 34930
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.718531428571429e-06,
+      "loss": 0.8024,
+      "step": 34931
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.718262857142858e-06,
+      "loss": 1.0067,
+      "step": 34932
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.717994285714286e-06,
+      "loss": 1.294,
+      "step": 34933
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.717725714285715e-06,
+      "loss": 1.2813,
+      "step": 34934
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.717457142857143e-06,
+      "loss": 1.1412,
+      "step": 34935
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.717188571428571e-06,
+      "loss": 1.0283,
+      "step": 34936
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.71692e-06,
+      "loss": 1.1877,
+      "step": 34937
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.716651428571428e-06,
+      "loss": 1.0191,
+      "step": 34938
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.716382857142857e-06,
+      "loss": 1.0984,
+      "step": 34939
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.716114285714286e-06,
+      "loss": 0.8636,
+      "step": 34940
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.715845714285715e-06,
+      "loss": 1.1531,
+      "step": 34941
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.715577142857144e-06,
+      "loss": 1.0818,
+      "step": 34942
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.715308571428572e-06,
+      "loss": 1.3046,
+      "step": 34943
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.71504e-06,
+      "loss": 1.4315,
+      "step": 34944
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.714771428571429e-06,
+      "loss": 0.8225,
+      "step": 34945
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.714502857142857e-06,
+      "loss": 0.652,
+      "step": 34946
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.714234285714286e-06,
+      "loss": 0.7203,
+      "step": 34947
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.713965714285714e-06,
+      "loss": 1.0251,
+      "step": 34948
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.713697142857143e-06,
+      "loss": 0.9068,
+      "step": 34949
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.713428571428572e-06,
+      "loss": 1.5269,
+      "step": 34950
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.71316e-06,
+      "loss": 1.1166,
+      "step": 34951
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.712891428571429e-06,
+      "loss": 1.4705,
+      "step": 34952
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.712622857142857e-06,
+      "loss": 0.997,
+      "step": 34953
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.712354285714286e-06,
+      "loss": 1.3219,
+      "step": 34954
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.712085714285715e-06,
+      "loss": 1.3867,
+      "step": 34955
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.711817142857143e-06,
+      "loss": 1.298,
+      "step": 34956
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.711548571428572e-06,
+      "loss": 0.7496,
+      "step": 34957
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.71128e-06,
+      "loss": 0.7853,
+      "step": 34958
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.711011428571429e-06,
+      "loss": 0.9086,
+      "step": 34959
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.710742857142858e-06,
+      "loss": 1.3479,
+      "step": 34960
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.710474285714286e-06,
+      "loss": 1.0613,
+      "step": 34961
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.710205714285714e-06,
+      "loss": 1.2359,
+      "step": 34962
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.709937142857143e-06,
+      "loss": 1.7488,
+      "step": 34963
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.709668571428571e-06,
+      "loss": 1.1348,
+      "step": 34964
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7094e-06,
+      "loss": 1.6561,
+      "step": 34965
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.709131428571428e-06,
+      "loss": 1.0181,
+      "step": 34966
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.708862857142857e-06,
+      "loss": 1.1599,
+      "step": 34967
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.708594285714286e-06,
+      "loss": 1.1515,
+      "step": 34968
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7083257142857146e-06,
+      "loss": 1.2094,
+      "step": 34969
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.708057142857144e-06,
+      "loss": 1.2717,
+      "step": 34970
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.707788571428572e-06,
+      "loss": 1.2805,
+      "step": 34971
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.70752e-06,
+      "loss": 1.0503,
+      "step": 34972
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.707251428571429e-06,
+      "loss": 1.3923,
+      "step": 34973
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.706982857142857e-06,
+      "loss": 1.3666,
+      "step": 34974
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.706714285714286e-06,
+      "loss": 1.4044,
+      "step": 34975
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.706445714285714e-06,
+      "loss": 1.2425,
+      "step": 34976
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7061771428571425e-06,
+      "loss": 0.8565,
+      "step": 34977
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7059085714285715e-06,
+      "loss": 1.1775,
+      "step": 34978
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.70564e-06,
+      "loss": 1.4334,
+      "step": 34979
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.705371428571429e-06,
+      "loss": 1.4479,
+      "step": 34980
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.705102857142858e-06,
+      "loss": 0.9875,
+      "step": 34981
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.704834285714286e-06,
+      "loss": 0.808,
+      "step": 34982
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.704565714285715e-06,
+      "loss": 1.1794,
+      "step": 34983
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.704297142857143e-06,
+      "loss": 1.0621,
+      "step": 34984
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.704028571428571e-06,
+      "loss": 1.0071,
+      "step": 34985
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.70376e-06,
+      "loss": 0.9769,
+      "step": 34986
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7034914285714285e-06,
+      "loss": 1.2365,
+      "step": 34987
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7032228571428575e-06,
+      "loss": 0.7545,
+      "step": 34988
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.702954285714286e-06,
+      "loss": 1.291,
+      "step": 34989
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.702685714285714e-06,
+      "loss": 1.0776,
+      "step": 34990
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.702417142857143e-06,
+      "loss": 1.3532,
+      "step": 34991
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.702148571428571e-06,
+      "loss": 1.2644,
+      "step": 34992
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.70188e-06,
+      "loss": 1.3401,
+      "step": 34993
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.701611428571429e-06,
+      "loss": 1.0746,
+      "step": 34994
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.701342857142857e-06,
+      "loss": 1.0956,
+      "step": 34995
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.701074285714286e-06,
+      "loss": 1.0895,
+      "step": 34996
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7008057142857145e-06,
+      "loss": 0.9257,
+      "step": 34997
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7005371428571435e-06,
+      "loss": 1.4213,
+      "step": 34998
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.700268571428572e-06,
+      "loss": 1.2718,
+      "step": 34999
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.7e-06,
+      "loss": 1.3625,
+      "step": 35000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.699731428571429e-06,
+      "loss": 1.1737,
+      "step": 35001
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.699462857142857e-06,
+      "loss": 0.8857,
+      "step": 35002
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.699194285714286e-06,
+      "loss": 1.1997,
+      "step": 35003
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.698925714285714e-06,
+      "loss": 0.8796,
+      "step": 35004
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6986571428571425e-06,
+      "loss": 1.0255,
+      "step": 35005
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6983885714285715e-06,
+      "loss": 1.0354,
+      "step": 35006
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6981200000000005e-06,
+      "loss": 1.1769,
+      "step": 35007
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.697851428571429e-06,
+      "loss": 0.9637,
+      "step": 35008
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.697582857142858e-06,
+      "loss": 0.9218,
+      "step": 35009
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.697314285714286e-06,
+      "loss": 1.1655,
+      "step": 35010
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.697045714285715e-06,
+      "loss": 0.6588,
+      "step": 35011
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.696777142857143e-06,
+      "loss": 1.1293,
+      "step": 35012
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.696508571428571e-06,
+      "loss": 1.6246,
+      "step": 35013
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.69624e-06,
+      "loss": 0.7789,
+      "step": 35014
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6959714285714285e-06,
+      "loss": 1.4463,
+      "step": 35015
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6957028571428575e-06,
+      "loss": 1.3797,
+      "step": 35016
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.695434285714286e-06,
+      "loss": 0.7574,
+      "step": 35017
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.695165714285714e-06,
+      "loss": 1.3435,
+      "step": 35018
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.694897142857143e-06,
+      "loss": 1.3788,
+      "step": 35019
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.694628571428572e-06,
+      "loss": 0.7702,
+      "step": 35020
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.694360000000001e-06,
+      "loss": 1.1984,
+      "step": 35021
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.694091428571429e-06,
+      "loss": 1.1151,
+      "step": 35022
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.693822857142857e-06,
+      "loss": 0.9264,
+      "step": 35023
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.693554285714286e-06,
+      "loss": 1.1207,
+      "step": 35024
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6932857142857145e-06,
+      "loss": 0.9395,
+      "step": 35025
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.693017142857143e-06,
+      "loss": 1.1141,
+      "step": 35026
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.692748571428572e-06,
+      "loss": 0.9773,
+      "step": 35027
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.69248e-06,
+      "loss": 1.3361,
+      "step": 35028
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.692211428571429e-06,
+      "loss": 1.0088,
+      "step": 35029
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.691942857142857e-06,
+      "loss": 1.2359,
+      "step": 35030
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.691674285714285e-06,
+      "loss": 0.7881,
+      "step": 35031
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.691405714285714e-06,
+      "loss": 1.3573,
+      "step": 35032
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.691137142857143e-06,
+      "loss": 0.9318,
+      "step": 35033
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.690868571428572e-06,
+      "loss": 1.0841,
+      "step": 35034
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6906000000000005e-06,
+      "loss": 1.2118,
+      "step": 35035
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.690331428571429e-06,
+      "loss": 0.9375,
+      "step": 35036
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.690062857142858e-06,
+      "loss": 1.3335,
+      "step": 35037
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.689794285714286e-06,
+      "loss": 1.4963,
+      "step": 35038
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.689525714285715e-06,
+      "loss": 1.3371,
+      "step": 35039
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.689257142857143e-06,
+      "loss": 1.1272,
+      "step": 35040
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.688988571428571e-06,
+      "loss": 1.0261,
+      "step": 35041
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.68872e-06,
+      "loss": 0.9863,
+      "step": 35042
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.688451428571428e-06,
+      "loss": 0.9762,
+      "step": 35043
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6881828571428574e-06,
+      "loss": 1.472,
+      "step": 35044
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.687914285714286e-06,
+      "loss": 1.0014,
+      "step": 35045
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.687645714285715e-06,
+      "loss": 1.0189,
+      "step": 35046
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.687377142857144e-06,
+      "loss": 1.4729,
+      "step": 35047
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.687108571428572e-06,
+      "loss": 1.8855,
+      "step": 35048
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.68684e-06,
+      "loss": 0.9539,
+      "step": 35049
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.686571428571429e-06,
+      "loss": 1.1308,
+      "step": 35050
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.686302857142857e-06,
+      "loss": 1.1032,
+      "step": 35051
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.686034285714286e-06,
+      "loss": 1.2748,
+      "step": 35052
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.685765714285714e-06,
+      "loss": 0.8765,
+      "step": 35053
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.685497142857143e-06,
+      "loss": 1.292,
+      "step": 35054
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.685228571428572e-06,
+      "loss": 1.318,
+      "step": 35055
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.68496e-06,
+      "loss": 1.2551,
+      "step": 35056
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.684691428571429e-06,
+      "loss": 1.3925,
+      "step": 35057
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.684422857142857e-06,
+      "loss": 0.8153,
+      "step": 35058
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.684154285714286e-06,
+      "loss": 1.0619,
+      "step": 35059
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.683885714285715e-06,
+      "loss": 1.1167,
+      "step": 35060
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.683617142857143e-06,
+      "loss": 1.3323,
+      "step": 35061
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.683348571428572e-06,
+      "loss": 1.2697,
+      "step": 35062
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.68308e-06,
+      "loss": 1.312,
+      "step": 35063
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.682811428571429e-06,
+      "loss": 0.6865,
+      "step": 35064
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.682542857142858e-06,
+      "loss": 1.277,
+      "step": 35065
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.682274285714286e-06,
+      "loss": 1.1039,
+      "step": 35066
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.682005714285715e-06,
+      "loss": 0.9612,
+      "step": 35067
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.681737142857143e-06,
+      "loss": 0.728,
+      "step": 35068
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.681468571428571e-06,
+      "loss": 1.0821,
+      "step": 35069
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6812e-06,
+      "loss": 1.2591,
+      "step": 35070
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.680931428571428e-06,
+      "loss": 0.8878,
+      "step": 35071
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6806628571428565e-06,
+      "loss": 1.2996,
+      "step": 35072
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.680394285714286e-06,
+      "loss": 1.1727,
+      "step": 35073
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6801257142857146e-06,
+      "loss": 1.1766,
+      "step": 35074
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.679857142857144e-06,
+      "loss": 1.4092,
+      "step": 35075
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.679588571428572e-06,
+      "loss": 1.3716,
+      "step": 35076
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.67932e-06,
+      "loss": 1.3688,
+      "step": 35077
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.679051428571429e-06,
+      "loss": 1.4559,
+      "step": 35078
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.678782857142857e-06,
+      "loss": 0.8498,
+      "step": 35079
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.678514285714286e-06,
+      "loss": 0.9513,
+      "step": 35080
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.678245714285714e-06,
+      "loss": 1.4752,
+      "step": 35081
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6779771428571425e-06,
+      "loss": 0.9545,
+      "step": 35082
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.6777085714285715e-06,
+      "loss": 1.1959,
+      "step": 35083
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.67744e-06,
+      "loss": 0.94,
+      "step": 35084
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.677171428571429e-06,
+      "loss": 1.1779,
+      "step": 35085
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.676902857142858e-06,
+      "loss": 1.2532,
+      "step": 35086
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.676634285714286e-06,
+      "loss": 1.6055,
+      "step": 35087
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.676365714285715e-06,
+      "loss": 1.0294,
+      "step": 35088
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.676097142857143e-06,
+      "loss": 1.2357,
+      "step": 35089
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.675828571428572e-06,
+      "loss": 1.0808,
+      "step": 35090
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.67556e-06,
+      "loss": 0.8957,
+      "step": 35091
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6752914285714285e-06,
+      "loss": 1.1794,
+      "step": 35092
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6750228571428575e-06,
+      "loss": 1.0682,
+      "step": 35093
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.674754285714286e-06,
+      "loss": 1.198,
+      "step": 35094
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.674485714285714e-06,
+      "loss": 0.9507,
+      "step": 35095
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.674217142857143e-06,
+      "loss": 1.0223,
+      "step": 35096
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.673948571428571e-06,
+      "loss": 1.2926,
+      "step": 35097
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.67368e-06,
+      "loss": 1.0352,
+      "step": 35098
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.673411428571429e-06,
+      "loss": 1.009,
+      "step": 35099
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.673142857142857e-06,
+      "loss": 1.2009,
+      "step": 35100
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.672874285714286e-06,
+      "loss": 0.9309,
+      "step": 35101
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6726057142857145e-06,
+      "loss": 1.1752,
+      "step": 35102
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6723371428571435e-06,
+      "loss": 0.9834,
+      "step": 35103
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.672068571428572e-06,
+      "loss": 1.1051,
+      "step": 35104
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6718e-06,
+      "loss": 1.1048,
+      "step": 35105
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.671531428571429e-06,
+      "loss": 1.08,
+      "step": 35106
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.671262857142857e-06,
+      "loss": 1.1634,
+      "step": 35107
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.670994285714286e-06,
+      "loss": 1.2099,
+      "step": 35108
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.670725714285714e-06,
+      "loss": 1.1989,
+      "step": 35109
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6704571428571425e-06,
+      "loss": 1.3901,
+      "step": 35110
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6701885714285715e-06,
+      "loss": 0.9344,
+      "step": 35111
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6699200000000005e-06,
+      "loss": 1.4282,
+      "step": 35112
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.669651428571429e-06,
+      "loss": 1.008,
+      "step": 35113
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.669382857142858e-06,
+      "loss": 1.1412,
+      "step": 35114
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.669114285714286e-06,
+      "loss": 0.9741,
+      "step": 35115
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.668845714285715e-06,
+      "loss": 0.8486,
+      "step": 35116
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.668577142857143e-06,
+      "loss": 1.5625,
+      "step": 35117
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.668308571428571e-06,
+      "loss": 1.0276,
+      "step": 35118
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.66804e-06,
+      "loss": 1.015,
+      "step": 35119
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6677714285714285e-06,
+      "loss": 1.2023,
+      "step": 35120
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6675028571428575e-06,
+      "loss": 1.0851,
+      "step": 35121
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.667234285714286e-06,
+      "loss": 1.2304,
+      "step": 35122
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.666965714285714e-06,
+      "loss": 1.2141,
+      "step": 35123
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.666697142857143e-06,
+      "loss": 0.8441,
+      "step": 35124
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.666428571428571e-06,
+      "loss": 0.9389,
+      "step": 35125
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.666160000000001e-06,
+      "loss": 1.0631,
+      "step": 35126
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.665891428571429e-06,
+      "loss": 0.9473,
+      "step": 35127
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.665622857142857e-06,
+      "loss": 1.0364,
+      "step": 35128
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.665354285714286e-06,
+      "loss": 1.3609,
+      "step": 35129
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6650857142857145e-06,
+      "loss": 1.3259,
+      "step": 35130
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6648171428571435e-06,
+      "loss": 1.4878,
+      "step": 35131
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.664548571428572e-06,
+      "loss": 1.2424,
+      "step": 35132
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.66428e-06,
+      "loss": 0.946,
+      "step": 35133
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.664011428571429e-06,
+      "loss": 1.2313,
+      "step": 35134
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.663742857142857e-06,
+      "loss": 1.0002,
+      "step": 35135
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.663474285714285e-06,
+      "loss": 1.1104,
+      "step": 35136
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.663205714285714e-06,
+      "loss": 0.9131,
+      "step": 35137
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.662937142857142e-06,
+      "loss": 1.4531,
+      "step": 35138
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.662668571428572e-06,
+      "loss": 1.2508,
+      "step": 35139
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6624000000000005e-06,
+      "loss": 1.1158,
+      "step": 35140
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.662131428571429e-06,
+      "loss": 0.798,
+      "step": 35141
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.661862857142858e-06,
+      "loss": 1.3305,
+      "step": 35142
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.661594285714286e-06,
+      "loss": 0.9893,
+      "step": 35143
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.661325714285715e-06,
+      "loss": 1.8009,
+      "step": 35144
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.661057142857143e-06,
+      "loss": 1.0457,
+      "step": 35145
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.660788571428571e-06,
+      "loss": 1.0126,
+      "step": 35146
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.66052e-06,
+      "loss": 1.1131,
+      "step": 35147
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.660251428571428e-06,
+      "loss": 1.1436,
+      "step": 35148
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6599828571428574e-06,
+      "loss": 1.1904,
+      "step": 35149
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.659714285714286e-06,
+      "loss": 1.0676,
+      "step": 35150
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.659445714285714e-06,
+      "loss": 0.8723,
+      "step": 35151
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.659177142857144e-06,
+      "loss": 1.1364,
+      "step": 35152
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.658908571428572e-06,
+      "loss": 0.5263,
+      "step": 35153
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.658640000000001e-06,
+      "loss": 1.1116,
+      "step": 35154
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.658371428571429e-06,
+      "loss": 1.5018,
+      "step": 35155
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.658102857142857e-06,
+      "loss": 0.9129,
+      "step": 35156
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.657834285714286e-06,
+      "loss": 1.2459,
+      "step": 35157
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.657565714285714e-06,
+      "loss": 0.8559,
+      "step": 35158
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.657297142857143e-06,
+      "loss": 1.3249,
+      "step": 35159
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.657028571428572e-06,
+      "loss": 1.6708,
+      "step": 35160
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.65676e-06,
+      "loss": 1.1556,
+      "step": 35161
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.656491428571429e-06,
+      "loss": 1.4277,
+      "step": 35162
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.656222857142857e-06,
+      "loss": 1.3146,
+      "step": 35163
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.655954285714285e-06,
+      "loss": 1.0626,
+      "step": 35164
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.655685714285715e-06,
+      "loss": 1.1785,
+      "step": 35165
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.655417142857143e-06,
+      "loss": 1.3751,
+      "step": 35166
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.655148571428572e-06,
+      "loss": 1.1007,
+      "step": 35167
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.65488e-06,
+      "loss": 1.4778,
+      "step": 35168
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6546114285714286e-06,
+      "loss": 1.1884,
+      "step": 35169
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.654342857142858e-06,
+      "loss": 0.8155,
+      "step": 35170
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.654074285714286e-06,
+      "loss": 1.4043,
+      "step": 35171
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.653805714285715e-06,
+      "loss": 1.0543,
+      "step": 35172
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.653537142857143e-06,
+      "loss": 1.2631,
+      "step": 35173
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.653268571428571e-06,
+      "loss": 1.2883,
+      "step": 35174
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.653e-06,
+      "loss": 1.0776,
+      "step": 35175
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.652731428571428e-06,
+      "loss": 1.0459,
+      "step": 35176
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.652462857142857e-06,
+      "loss": 1.0298,
+      "step": 35177
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6521942857142855e-06,
+      "loss": 1.1766,
+      "step": 35178
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6519257142857146e-06,
+      "loss": 0.9355,
+      "step": 35179
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.651657142857144e-06,
+      "loss": 1.2038,
+      "step": 35180
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.651388571428572e-06,
+      "loss": 1.2185,
+      "step": 35181
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.65112e-06,
+      "loss": 0.8275,
+      "step": 35182
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.650851428571429e-06,
+      "loss": 1.4352,
+      "step": 35183
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.650582857142857e-06,
+      "loss": 1.3499,
+      "step": 35184
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.650314285714286e-06,
+      "loss": 1.2485,
+      "step": 35185
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.650045714285714e-06,
+      "loss": 1.6057,
+      "step": 35186
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6497771428571425e-06,
+      "loss": 0.9409,
+      "step": 35187
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6495085714285715e-06,
+      "loss": 0.9634,
+      "step": 35188
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.64924e-06,
+      "loss": 0.855,
+      "step": 35189
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.648971428571429e-06,
+      "loss": 1.1718,
+      "step": 35190
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.648702857142857e-06,
+      "loss": 1.2401,
+      "step": 35191
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.648434285714286e-06,
+      "loss": 0.8918,
+      "step": 35192
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.648165714285715e-06,
+      "loss": 1.1729,
+      "step": 35193
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.647897142857143e-06,
+      "loss": 1.3012,
+      "step": 35194
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.647628571428572e-06,
+      "loss": 0.8988,
+      "step": 35195
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.64736e-06,
+      "loss": 1.5119,
+      "step": 35196
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6470914285714285e-06,
+      "loss": 1.2441,
+      "step": 35197
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6468228571428575e-06,
+      "loss": 1.3988,
+      "step": 35198
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.646554285714286e-06,
+      "loss": 1.4957,
+      "step": 35199
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.646285714285714e-06,
+      "loss": 1.3593,
+      "step": 35200
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.646017142857143e-06,
+      "loss": 1.3442,
+      "step": 35201
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.645748571428571e-06,
+      "loss": 1.1077,
+      "step": 35202
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.64548e-06,
+      "loss": 0.8837,
+      "step": 35203
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.645211428571428e-06,
+      "loss": 1.2652,
+      "step": 35204
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.644942857142857e-06,
+      "loss": 0.989,
+      "step": 35205
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.644674285714286e-06,
+      "loss": 1.2737,
+      "step": 35206
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6444057142857145e-06,
+      "loss": 1.1028,
+      "step": 35207
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6441371428571435e-06,
+      "loss": 1.0513,
+      "step": 35208
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.643868571428572e-06,
+      "loss": 0.8691,
+      "step": 35209
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6436e-06,
+      "loss": 1.7285,
+      "step": 35210
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.643331428571429e-06,
+      "loss": 0.6874,
+      "step": 35211
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.643062857142857e-06,
+      "loss": 1.3505,
+      "step": 35212
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.642794285714286e-06,
+      "loss": 0.8043,
+      "step": 35213
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.642525714285714e-06,
+      "loss": 1.1462,
+      "step": 35214
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6422571428571425e-06,
+      "loss": 1.6568,
+      "step": 35215
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6419885714285715e-06,
+      "loss": 1.0035,
+      "step": 35216
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.64172e-06,
+      "loss": 1.1143,
+      "step": 35217
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6414514285714295e-06,
+      "loss": 1.2054,
+      "step": 35218
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.641182857142858e-06,
+      "loss": 1.3488,
+      "step": 35219
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.640914285714286e-06,
+      "loss": 1.1578,
+      "step": 35220
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.640645714285715e-06,
+      "loss": 1.1343,
+      "step": 35221
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.640377142857143e-06,
+      "loss": 1.0707,
+      "step": 35222
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.640108571428571e-06,
+      "loss": 0.6362,
+      "step": 35223
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.63984e-06,
+      "loss": 1.1772,
+      "step": 35224
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6395714285714285e-06,
+      "loss": 1.3456,
+      "step": 35225
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6393028571428575e-06,
+      "loss": 1.0561,
+      "step": 35226
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.639034285714286e-06,
+      "loss": 1.1726,
+      "step": 35227
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.638765714285714e-06,
+      "loss": 1.1028,
+      "step": 35228
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.638497142857143e-06,
+      "loss": 0.9828,
+      "step": 35229
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.638228571428571e-06,
+      "loss": 0.9191,
+      "step": 35230
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.63796e-06,
+      "loss": 1.3128,
+      "step": 35231
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.637691428571429e-06,
+      "loss": 1.1755,
+      "step": 35232
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.637422857142857e-06,
+      "loss": 0.8441,
+      "step": 35233
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.637154285714286e-06,
+      "loss": 1.5122,
+      "step": 35234
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6368857142857145e-06,
+      "loss": 1.2353,
+      "step": 35235
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6366171428571435e-06,
+      "loss": 0.8572,
+      "step": 35236
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.636348571428572e-06,
+      "loss": 1.1314,
+      "step": 35237
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.63608e-06,
+      "loss": 1.2649,
+      "step": 35238
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.635811428571429e-06,
+      "loss": 1.5585,
+      "step": 35239
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.635542857142857e-06,
+      "loss": 1.3397,
+      "step": 35240
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.635274285714286e-06,
+      "loss": 1.2219,
+      "step": 35241
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.635005714285714e-06,
+      "loss": 1.2381,
+      "step": 35242
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.634737142857142e-06,
+      "loss": 0.8185,
+      "step": 35243
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6344685714285714e-06,
+      "loss": 1.6183,
+      "step": 35244
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6342000000000005e-06,
+      "loss": 1.0042,
+      "step": 35245
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.633931428571429e-06,
+      "loss": 1.1174,
+      "step": 35246
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.633662857142858e-06,
+      "loss": 0.9453,
+      "step": 35247
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.633394285714286e-06,
+      "loss": 1.1597,
+      "step": 35248
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.633125714285715e-06,
+      "loss": 1.0988,
+      "step": 35249
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.632857142857143e-06,
+      "loss": 1.2998,
+      "step": 35250
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.632588571428571e-06,
+      "loss": 1.6076,
+      "step": 35251
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.63232e-06,
+      "loss": 0.9341,
+      "step": 35252
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.632051428571428e-06,
+      "loss": 0.9398,
+      "step": 35253
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.6317828571428574e-06,
+      "loss": 1.1055,
+      "step": 35254
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.631514285714286e-06,
+      "loss": 1.3597,
+      "step": 35255
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.631245714285714e-06,
+      "loss": 0.831,
+      "step": 35256
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.630977142857143e-06,
+      "loss": 1.069,
+      "step": 35257
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.630708571428572e-06,
+      "loss": 1.0612,
+      "step": 35258
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.630440000000001e-06,
+      "loss": 1.2739,
+      "step": 35259
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.630171428571429e-06,
+      "loss": 1.3303,
+      "step": 35260
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.629902857142857e-06,
+      "loss": 1.5106,
+      "step": 35261
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.629634285714286e-06,
+      "loss": 1.0946,
+      "step": 35262
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.629365714285714e-06,
+      "loss": 0.9921,
+      "step": 35263
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.629097142857143e-06,
+      "loss": 1.4971,
+      "step": 35264
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.628828571428572e-06,
+      "loss": 1.3607,
+      "step": 35265
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.62856e-06,
+      "loss": 0.7866,
+      "step": 35266
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.628291428571429e-06,
+      "loss": 1.434,
+      "step": 35267
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.628022857142857e-06,
+      "loss": 1.4547,
+      "step": 35268
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.627754285714285e-06,
+      "loss": 1.0416,
+      "step": 35269
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.627485714285714e-06,
+      "loss": 0.8613,
+      "step": 35270
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.627217142857143e-06,
+      "loss": 1.0351,
+      "step": 35271
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.626948571428572e-06,
+      "loss": 1.2695,
+      "step": 35272
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.62668e-06,
+      "loss": 0.8588,
+      "step": 35273
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6264114285714286e-06,
+      "loss": 0.9208,
+      "step": 35274
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.626142857142858e-06,
+      "loss": 0.8577,
+      "step": 35275
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.625874285714286e-06,
+      "loss": 1.1008,
+      "step": 35276
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.625605714285715e-06,
+      "loss": 1.2575,
+      "step": 35277
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.625337142857143e-06,
+      "loss": 1.1104,
+      "step": 35278
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.625068571428571e-06,
+      "loss": 1.0994,
+      "step": 35279
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6248e-06,
+      "loss": 1.1293,
+      "step": 35280
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.624531428571428e-06,
+      "loss": 1.6262,
+      "step": 35281
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.624262857142857e-06,
+      "loss": 1.1435,
+      "step": 35282
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6239942857142855e-06,
+      "loss": 1.0176,
+      "step": 35283
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6237257142857146e-06,
+      "loss": 1.3962,
+      "step": 35284
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.623457142857144e-06,
+      "loss": 1.2368,
+      "step": 35285
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.623188571428572e-06,
+      "loss": 0.8747,
+      "step": 35286
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.62292e-06,
+      "loss": 1.0704,
+      "step": 35287
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.622651428571429e-06,
+      "loss": 1.2047,
+      "step": 35288
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.622382857142857e-06,
+      "loss": 1.1258,
+      "step": 35289
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.622114285714286e-06,
+      "loss": 1.6551,
+      "step": 35290
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.621845714285714e-06,
+      "loss": 1.0773,
+      "step": 35291
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6215771428571425e-06,
+      "loss": 1.1927,
+      "step": 35292
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6213085714285715e-06,
+      "loss": 0.7932,
+      "step": 35293
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.62104e-06,
+      "loss": 1.3008,
+      "step": 35294
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.620771428571429e-06,
+      "loss": 0.9771,
+      "step": 35295
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.620502857142857e-06,
+      "loss": 1.6249,
+      "step": 35296
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.620234285714286e-06,
+      "loss": 1.0952,
+      "step": 35297
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.619965714285715e-06,
+      "loss": 1.0794,
+      "step": 35298
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.619697142857143e-06,
+      "loss": 1.1422,
+      "step": 35299
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.619428571428572e-06,
+      "loss": 1.1099,
+      "step": 35300
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.61916e-06,
+      "loss": 1.4132,
+      "step": 35301
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6188914285714285e-06,
+      "loss": 1.0507,
+      "step": 35302
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6186228571428575e-06,
+      "loss": 1.3818,
+      "step": 35303
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.618354285714286e-06,
+      "loss": 1.0209,
+      "step": 35304
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.618085714285715e-06,
+      "loss": 1.3206,
+      "step": 35305
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.617817142857143e-06,
+      "loss": 1.3232,
+      "step": 35306
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.617548571428571e-06,
+      "loss": 1.0917,
+      "step": 35307
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.61728e-06,
+      "loss": 0.9309,
+      "step": 35308
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.617011428571428e-06,
+      "loss": 0.9883,
+      "step": 35309
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.616742857142857e-06,
+      "loss": 1.0779,
+      "step": 35310
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.616474285714286e-06,
+      "loss": 1.302,
+      "step": 35311
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6162057142857145e-06,
+      "loss": 1.0672,
+      "step": 35312
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6159371428571435e-06,
+      "loss": 1.6444,
+      "step": 35313
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.615668571428572e-06,
+      "loss": 1.2884,
+      "step": 35314
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6154e-06,
+      "loss": 1.1369,
+      "step": 35315
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.615131428571429e-06,
+      "loss": 0.9196,
+      "step": 35316
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.614862857142857e-06,
+      "loss": 1.2328,
+      "step": 35317
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.614594285714286e-06,
+      "loss": 1.1266,
+      "step": 35318
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.614325714285714e-06,
+      "loss": 1.2372,
+      "step": 35319
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6140571428571425e-06,
+      "loss": 1.3153,
+      "step": 35320
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6137885714285715e-06,
+      "loss": 0.8698,
+      "step": 35321
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.61352e-06,
+      "loss": 0.9251,
+      "step": 35322
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.613251428571429e-06,
+      "loss": 0.8974,
+      "step": 35323
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.612982857142858e-06,
+      "loss": 1.3424,
+      "step": 35324
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.612714285714286e-06,
+      "loss": 1.1787,
+      "step": 35325
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.612445714285715e-06,
+      "loss": 1.2599,
+      "step": 35326
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.612177142857143e-06,
+      "loss": 0.9476,
+      "step": 35327
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.611908571428572e-06,
+      "loss": 1.1029,
+      "step": 35328
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.61164e-06,
+      "loss": 0.7414,
+      "step": 35329
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6113714285714285e-06,
+      "loss": 1.1916,
+      "step": 35330
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6111028571428575e-06,
+      "loss": 0.8955,
+      "step": 35331
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.610834285714286e-06,
+      "loss": 0.9429,
+      "step": 35332
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.610565714285714e-06,
+      "loss": 1.4127,
+      "step": 35333
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.610297142857143e-06,
+      "loss": 1.3186,
+      "step": 35334
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.610028571428571e-06,
+      "loss": 1.3301,
+      "step": 35335
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.60976e-06,
+      "loss": 1.242,
+      "step": 35336
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.609491428571429e-06,
+      "loss": 1.0829,
+      "step": 35337
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.609222857142857e-06,
+      "loss": 1.3397,
+      "step": 35338
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.608954285714286e-06,
+      "loss": 0.9232,
+      "step": 35339
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6086857142857145e-06,
+      "loss": 1.0961,
+      "step": 35340
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6084171428571435e-06,
+      "loss": 1.0813,
+      "step": 35341
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.608148571428572e-06,
+      "loss": 1.2423,
+      "step": 35342
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.60788e-06,
+      "loss": 1.2911,
+      "step": 35343
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.607611428571429e-06,
+      "loss": 0.9216,
+      "step": 35344
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.607342857142857e-06,
+      "loss": 1.2288,
+      "step": 35345
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.607074285714286e-06,
+      "loss": 1.1176,
+      "step": 35346
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.606805714285714e-06,
+      "loss": 1.2299,
+      "step": 35347
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.606537142857142e-06,
+      "loss": 0.5861,
+      "step": 35348
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6062685714285714e-06,
+      "loss": 0.9546,
+      "step": 35349
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6060000000000005e-06,
+      "loss": 1.351,
+      "step": 35350
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.605731428571429e-06,
+      "loss": 0.9593,
+      "step": 35351
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.605462857142858e-06,
+      "loss": 0.7341,
+      "step": 35352
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.605194285714286e-06,
+      "loss": 1.1183,
+      "step": 35353
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.604925714285715e-06,
+      "loss": 1.2654,
+      "step": 35354
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.604657142857143e-06,
+      "loss": 1.1928,
+      "step": 35355
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.604388571428571e-06,
+      "loss": 1.1839,
+      "step": 35356
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.60412e-06,
+      "loss": 0.886,
+      "step": 35357
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.603851428571428e-06,
+      "loss": 1.097,
+      "step": 35358
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6035828571428574e-06,
+      "loss": 1.133,
+      "step": 35359
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.603314285714286e-06,
+      "loss": 1.2161,
+      "step": 35360
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.603045714285714e-06,
+      "loss": 1.2507,
+      "step": 35361
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.602777142857143e-06,
+      "loss": 1.4097,
+      "step": 35362
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.602508571428572e-06,
+      "loss": 0.9072,
+      "step": 35363
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.602240000000001e-06,
+      "loss": 1.2314,
+      "step": 35364
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.601971428571429e-06,
+      "loss": 1.1328,
+      "step": 35365
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.601702857142857e-06,
+      "loss": 1.396,
+      "step": 35366
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.601434285714286e-06,
+      "loss": 0.8932,
+      "step": 35367
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.601165714285714e-06,
+      "loss": 0.6832,
+      "step": 35368
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.6008971428571434e-06,
+      "loss": 1.0519,
+      "step": 35369
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.600628571428572e-06,
+      "loss": 0.8678,
+      "step": 35370
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.60036e-06,
+      "loss": 1.0599,
+      "step": 35371
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.600091428571429e-06,
+      "loss": 0.7993,
+      "step": 35372
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.599822857142857e-06,
+      "loss": 0.9827,
+      "step": 35373
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.599554285714285e-06,
+      "loss": 1.02,
+      "step": 35374
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.599285714285714e-06,
+      "loss": 1.5132,
+      "step": 35375
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.599017142857143e-06,
+      "loss": 1.1273,
+      "step": 35376
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.598748571428572e-06,
+      "loss": 1.3101,
+      "step": 35377
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.59848e-06,
+      "loss": 0.945,
+      "step": 35378
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5982114285714286e-06,
+      "loss": 1.3941,
+      "step": 35379
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.597942857142858e-06,
+      "loss": 1.1281,
+      "step": 35380
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.597674285714286e-06,
+      "loss": 1.1391,
+      "step": 35381
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.597405714285715e-06,
+      "loss": 1.708,
+      "step": 35382
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.597137142857143e-06,
+      "loss": 0.9034,
+      "step": 35383
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.596868571428571e-06,
+      "loss": 1.2698,
+      "step": 35384
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5966e-06,
+      "loss": 1.2144,
+      "step": 35385
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.596331428571428e-06,
+      "loss": 1.2823,
+      "step": 35386
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.596062857142857e-06,
+      "loss": 1.2854,
+      "step": 35387
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5957942857142855e-06,
+      "loss": 1.2802,
+      "step": 35388
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5955257142857146e-06,
+      "loss": 1.0915,
+      "step": 35389
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.595257142857144e-06,
+      "loss": 1.0108,
+      "step": 35390
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.594988571428572e-06,
+      "loss": 1.4391,
+      "step": 35391
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.594720000000001e-06,
+      "loss": 0.8885,
+      "step": 35392
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.594451428571429e-06,
+      "loss": 0.9389,
+      "step": 35393
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.594182857142857e-06,
+      "loss": 0.8066,
+      "step": 35394
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.593914285714286e-06,
+      "loss": 1.227,
+      "step": 35395
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.593645714285714e-06,
+      "loss": 1.2305,
+      "step": 35396
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5933771428571425e-06,
+      "loss": 0.9965,
+      "step": 35397
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5931085714285715e-06,
+      "loss": 1.278,
+      "step": 35398
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.59284e-06,
+      "loss": 1.3475,
+      "step": 35399
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.592571428571429e-06,
+      "loss": 1.3723,
+      "step": 35400
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.592302857142857e-06,
+      "loss": 0.9749,
+      "step": 35401
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.592034285714286e-06,
+      "loss": 1.2064,
+      "step": 35402
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.591765714285715e-06,
+      "loss": 1.1769,
+      "step": 35403
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.591497142857143e-06,
+      "loss": 1.277,
+      "step": 35404
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.591228571428572e-06,
+      "loss": 1.4947,
+      "step": 35405
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.59096e-06,
+      "loss": 1.1606,
+      "step": 35406
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5906914285714285e-06,
+      "loss": 1.1215,
+      "step": 35407
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5904228571428575e-06,
+      "loss": 0.8499,
+      "step": 35408
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.590154285714286e-06,
+      "loss": 1.1649,
+      "step": 35409
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.589885714285715e-06,
+      "loss": 1.1892,
+      "step": 35410
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.589617142857143e-06,
+      "loss": 0.9504,
+      "step": 35411
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.589348571428571e-06,
+      "loss": 1.1568,
+      "step": 35412
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.58908e-06,
+      "loss": 1.071,
+      "step": 35413
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.588811428571428e-06,
+      "loss": 1.157,
+      "step": 35414
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.588542857142857e-06,
+      "loss": 0.8787,
+      "step": 35415
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.588274285714286e-06,
+      "loss": 1.2457,
+      "step": 35416
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5880057142857145e-06,
+      "loss": 1.1077,
+      "step": 35417
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5877371428571435e-06,
+      "loss": 0.989,
+      "step": 35418
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.587468571428572e-06,
+      "loss": 0.8549,
+      "step": 35419
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5872e-06,
+      "loss": 1.0803,
+      "step": 35420
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.586931428571429e-06,
+      "loss": 0.8256,
+      "step": 35421
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.586662857142857e-06,
+      "loss": 0.8492,
+      "step": 35422
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.586394285714286e-06,
+      "loss": 0.9922,
+      "step": 35423
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.586125714285714e-06,
+      "loss": 0.9107,
+      "step": 35424
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5858571428571425e-06,
+      "loss": 1.1128,
+      "step": 35425
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5855885714285715e-06,
+      "loss": 1.0964,
+      "step": 35426
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.58532e-06,
+      "loss": 1.3741,
+      "step": 35427
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.585051428571429e-06,
+      "loss": 1.0847,
+      "step": 35428
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.584782857142858e-06,
+      "loss": 0.9303,
+      "step": 35429
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.584514285714286e-06,
+      "loss": 1.0509,
+      "step": 35430
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.584245714285715e-06,
+      "loss": 1.4945,
+      "step": 35431
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.583977142857143e-06,
+      "loss": 1.091,
+      "step": 35432
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.583708571428572e-06,
+      "loss": 1.0221,
+      "step": 35433
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.58344e-06,
+      "loss": 1.1079,
+      "step": 35434
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5831714285714285e-06,
+      "loss": 1.1984,
+      "step": 35435
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5829028571428575e-06,
+      "loss": 1.3854,
+      "step": 35436
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.582634285714286e-06,
+      "loss": 1.0349,
+      "step": 35437
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.582365714285714e-06,
+      "loss": 0.9309,
+      "step": 35438
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.582097142857143e-06,
+      "loss": 1.2434,
+      "step": 35439
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.581828571428571e-06,
+      "loss": 1.3738,
+      "step": 35440
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.58156e-06,
+      "loss": 0.8706,
+      "step": 35441
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.581291428571429e-06,
+      "loss": 1.0107,
+      "step": 35442
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.581022857142857e-06,
+      "loss": 1.2495,
+      "step": 35443
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.580754285714286e-06,
+      "loss": 0.8726,
+      "step": 35444
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5804857142857145e-06,
+      "loss": 1.4226,
+      "step": 35445
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5802171428571435e-06,
+      "loss": 1.1882,
+      "step": 35446
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.579948571428572e-06,
+      "loss": 1.1075,
+      "step": 35447
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.57968e-06,
+      "loss": 1.2575,
+      "step": 35448
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.579411428571429e-06,
+      "loss": 1.0878,
+      "step": 35449
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.579142857142857e-06,
+      "loss": 1.2642,
+      "step": 35450
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.578874285714286e-06,
+      "loss": 1.2379,
+      "step": 35451
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.578605714285714e-06,
+      "loss": 1.2899,
+      "step": 35452
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.578337142857142e-06,
+      "loss": 1.4581,
+      "step": 35453
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5780685714285714e-06,
+      "loss": 1.1393,
+      "step": 35454
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5778000000000005e-06,
+      "loss": 1.1588,
+      "step": 35455
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5775314285714295e-06,
+      "loss": 0.945,
+      "step": 35456
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.577262857142858e-06,
+      "loss": 1.279,
+      "step": 35457
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.576994285714286e-06,
+      "loss": 0.8169,
+      "step": 35458
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.576725714285715e-06,
+      "loss": 0.8151,
+      "step": 35459
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.576457142857143e-06,
+      "loss": 0.8515,
+      "step": 35460
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.576188571428571e-06,
+      "loss": 1.0721,
+      "step": 35461
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.57592e-06,
+      "loss": 1.3132,
+      "step": 35462
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.575651428571428e-06,
+      "loss": 1.0862,
+      "step": 35463
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5753828571428574e-06,
+      "loss": 1.2222,
+      "step": 35464
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.575114285714286e-06,
+      "loss": 0.9064,
+      "step": 35465
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.574845714285714e-06,
+      "loss": 0.8309,
+      "step": 35466
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.574577142857143e-06,
+      "loss": 1.0356,
+      "step": 35467
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.574308571428572e-06,
+      "loss": 1.2975,
+      "step": 35468
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.574040000000001e-06,
+      "loss": 1.0773,
+      "step": 35469
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.573771428571429e-06,
+      "loss": 1.1361,
+      "step": 35470
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.573502857142857e-06,
+      "loss": 0.9017,
+      "step": 35471
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.573234285714286e-06,
+      "loss": 1.0931,
+      "step": 35472
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.572965714285714e-06,
+      "loss": 1.1306,
+      "step": 35473
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.572697142857143e-06,
+      "loss": 1.2806,
+      "step": 35474
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.572428571428572e-06,
+      "loss": 1.3503,
+      "step": 35475
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.57216e-06,
+      "loss": 1.317,
+      "step": 35476
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.571891428571429e-06,
+      "loss": 1.3665,
+      "step": 35477
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.571622857142857e-06,
+      "loss": 1.3933,
+      "step": 35478
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.571354285714286e-06,
+      "loss": 1.1932,
+      "step": 35479
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.571085714285714e-06,
+      "loss": 1.4211,
+      "step": 35480
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.570817142857143e-06,
+      "loss": 1.2723,
+      "step": 35481
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.570548571428572e-06,
+      "loss": 0.9274,
+      "step": 35482
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.57028e-06,
+      "loss": 1.0413,
+      "step": 35483
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5700114285714286e-06,
+      "loss": 1.2237,
+      "step": 35484
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.569742857142858e-06,
+      "loss": 1.1106,
+      "step": 35485
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.569474285714286e-06,
+      "loss": 0.9832,
+      "step": 35486
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.569205714285715e-06,
+      "loss": 1.3697,
+      "step": 35487
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.568937142857143e-06,
+      "loss": 0.8396,
+      "step": 35488
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.568668571428571e-06,
+      "loss": 1.2509,
+      "step": 35489
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5684e-06,
+      "loss": 1.0123,
+      "step": 35490
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.568131428571428e-06,
+      "loss": 0.9165,
+      "step": 35491
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.567862857142857e-06,
+      "loss": 1.0155,
+      "step": 35492
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5675942857142855e-06,
+      "loss": 1.019,
+      "step": 35493
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5673257142857146e-06,
+      "loss": 1.0834,
+      "step": 35494
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.567057142857144e-06,
+      "loss": 1.1103,
+      "step": 35495
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.566788571428572e-06,
+      "loss": 1.2505,
+      "step": 35496
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.566520000000001e-06,
+      "loss": 1.0894,
+      "step": 35497
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.566251428571429e-06,
+      "loss": 1.2039,
+      "step": 35498
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.565982857142857e-06,
+      "loss": 0.8476,
+      "step": 35499
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.565714285714286e-06,
+      "loss": 1.3026,
+      "step": 35500
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.565445714285714e-06,
+      "loss": 1.2613,
+      "step": 35501
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.565177142857143e-06,
+      "loss": 0.948,
+      "step": 35502
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5649085714285715e-06,
+      "loss": 1.1176,
+      "step": 35503
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.56464e-06,
+      "loss": 0.8748,
+      "step": 35504
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.564371428571429e-06,
+      "loss": 0.7224,
+      "step": 35505
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.564102857142857e-06,
+      "loss": 1.0519,
+      "step": 35506
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.563834285714286e-06,
+      "loss": 1.0265,
+      "step": 35507
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.563565714285715e-06,
+      "loss": 1.2136,
+      "step": 35508
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.563297142857143e-06,
+      "loss": 1.4167,
+      "step": 35509
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.563028571428572e-06,
+      "loss": 1.2633,
+      "step": 35510
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.56276e-06,
+      "loss": 1.1052,
+      "step": 35511
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5624914285714285e-06,
+      "loss": 1.0876,
+      "step": 35512
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5622228571428575e-06,
+      "loss": 1.162,
+      "step": 35513
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.561954285714286e-06,
+      "loss": 0.9639,
+      "step": 35514
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.561685714285715e-06,
+      "loss": 1.1967,
+      "step": 35515
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.561417142857143e-06,
+      "loss": 1.2853,
+      "step": 35516
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.561148571428571e-06,
+      "loss": 1.223,
+      "step": 35517
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.56088e-06,
+      "loss": 0.7424,
+      "step": 35518
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.560611428571428e-06,
+      "loss": 1.2139,
+      "step": 35519
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.560342857142857e-06,
+      "loss": 1.192,
+      "step": 35520
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.560074285714286e-06,
+      "loss": 1.5393,
+      "step": 35521
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5598057142857145e-06,
+      "loss": 1.0143,
+      "step": 35522
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5595371428571435e-06,
+      "loss": 1.1726,
+      "step": 35523
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.559268571428572e-06,
+      "loss": 1.1186,
+      "step": 35524
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.559e-06,
+      "loss": 1.5452,
+      "step": 35525
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.558731428571429e-06,
+      "loss": 1.6029,
+      "step": 35526
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.558462857142857e-06,
+      "loss": 1.5823,
+      "step": 35527
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.558194285714286e-06,
+      "loss": 1.5239,
+      "step": 35528
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.557925714285714e-06,
+      "loss": 0.9723,
+      "step": 35529
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5576571428571425e-06,
+      "loss": 1.3032,
+      "step": 35530
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5573885714285715e-06,
+      "loss": 0.7284,
+      "step": 35531
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.55712e-06,
+      "loss": 0.9821,
+      "step": 35532
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.556851428571429e-06,
+      "loss": 1.1959,
+      "step": 35533
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.556582857142858e-06,
+      "loss": 1.0188,
+      "step": 35534
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.556314285714286e-06,
+      "loss": 1.4023,
+      "step": 35535
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.556045714285715e-06,
+      "loss": 1.5577,
+      "step": 35536
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.555777142857143e-06,
+      "loss": 1.269,
+      "step": 35537
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.555508571428572e-06,
+      "loss": 1.5749,
+      "step": 35538
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.55524e-06,
+      "loss": 1.2006,
+      "step": 35539
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5549714285714285e-06,
+      "loss": 0.9575,
+      "step": 35540
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5547028571428575e-06,
+      "loss": 1.0952,
+      "step": 35541
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.554434285714286e-06,
+      "loss": 1.2925,
+      "step": 35542
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.554165714285715e-06,
+      "loss": 1.2112,
+      "step": 35543
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.553897142857143e-06,
+      "loss": 1.1942,
+      "step": 35544
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.553628571428571e-06,
+      "loss": 0.9914,
+      "step": 35545
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.55336e-06,
+      "loss": 1.2,
+      "step": 35546
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.553091428571429e-06,
+      "loss": 1.3332,
+      "step": 35547
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.552822857142857e-06,
+      "loss": 1.4876,
+      "step": 35548
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.552554285714286e-06,
+      "loss": 1.2523,
+      "step": 35549
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5522857142857145e-06,
+      "loss": 1.2603,
+      "step": 35550
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5520171428571435e-06,
+      "loss": 1.2389,
+      "step": 35551
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.551748571428572e-06,
+      "loss": 1.1514,
+      "step": 35552
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.55148e-06,
+      "loss": 1.1949,
+      "step": 35553
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.551211428571429e-06,
+      "loss": 1.3803,
+      "step": 35554
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.550942857142857e-06,
+      "loss": 1.1165,
+      "step": 35555
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.550674285714286e-06,
+      "loss": 0.9202,
+      "step": 35556
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.550405714285714e-06,
+      "loss": 1.149,
+      "step": 35557
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.550137142857142e-06,
+      "loss": 1.6028,
+      "step": 35558
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5498685714285714e-06,
+      "loss": 1.4551,
+      "step": 35559
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5496000000000005e-06,
+      "loss": 1.2085,
+      "step": 35560
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5493314285714295e-06,
+      "loss": 0.9259,
+      "step": 35561
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.549062857142858e-06,
+      "loss": 1.5914,
+      "step": 35562
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.548794285714286e-06,
+      "loss": 0.8187,
+      "step": 35563
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.548525714285715e-06,
+      "loss": 1.2433,
+      "step": 35564
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.548257142857143e-06,
+      "loss": 0.8178,
+      "step": 35565
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.547988571428572e-06,
+      "loss": 1.1192,
+      "step": 35566
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.54772e-06,
+      "loss": 1.2991,
+      "step": 35567
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.547451428571428e-06,
+      "loss": 1.2939,
+      "step": 35568
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5471828571428574e-06,
+      "loss": 1.059,
+      "step": 35569
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.546914285714286e-06,
+      "loss": 1.482,
+      "step": 35570
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.546645714285714e-06,
+      "loss": 1.2065,
+      "step": 35571
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.546377142857143e-06,
+      "loss": 1.2702,
+      "step": 35572
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.546108571428572e-06,
+      "loss": 0.8318,
+      "step": 35573
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.545840000000001e-06,
+      "loss": 0.7805,
+      "step": 35574
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.545571428571429e-06,
+      "loss": 1.2628,
+      "step": 35575
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.545302857142857e-06,
+      "loss": 0.8553,
+      "step": 35576
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.545034285714286e-06,
+      "loss": 1.0538,
+      "step": 35577
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.544765714285714e-06,
+      "loss": 1.1883,
+      "step": 35578
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.544497142857143e-06,
+      "loss": 1.2959,
+      "step": 35579
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.544228571428572e-06,
+      "loss": 0.7872,
+      "step": 35580
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.54396e-06,
+      "loss": 0.8,
+      "step": 35581
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.543691428571429e-06,
+      "loss": 0.95,
+      "step": 35582
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.543422857142857e-06,
+      "loss": 1.3941,
+      "step": 35583
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.543154285714286e-06,
+      "loss": 1.5425,
+      "step": 35584
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.542885714285714e-06,
+      "loss": 1.5582,
+      "step": 35585
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.542617142857142e-06,
+      "loss": 1.3207,
+      "step": 35586
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.542348571428572e-06,
+      "loss": 0.7321,
+      "step": 35587
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.54208e-06,
+      "loss": 0.9051,
+      "step": 35588
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.541811428571429e-06,
+      "loss": 0.9756,
+      "step": 35589
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.541542857142858e-06,
+      "loss": 0.9637,
+      "step": 35590
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.541274285714286e-06,
+      "loss": 1.1487,
+      "step": 35591
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.541005714285715e-06,
+      "loss": 1.5539,
+      "step": 35592
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.540737142857143e-06,
+      "loss": 1.2414,
+      "step": 35593
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.540468571428571e-06,
+      "loss": 0.8679,
+      "step": 35594
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5402e-06,
+      "loss": 0.9075,
+      "step": 35595
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.539931428571428e-06,
+      "loss": 1.0683,
+      "step": 35596
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.539662857142857e-06,
+      "loss": 0.7071,
+      "step": 35597
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5393942857142855e-06,
+      "loss": 1.1201,
+      "step": 35598
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.539125714285714e-06,
+      "loss": 1.0243,
+      "step": 35599
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.538857142857144e-06,
+      "loss": 1.1897,
+      "step": 35600
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.538588571428572e-06,
+      "loss": 1.0493,
+      "step": 35601
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.538320000000001e-06,
+      "loss": 0.8955,
+      "step": 35602
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.538051428571429e-06,
+      "loss": 1.2994,
+      "step": 35603
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.537782857142857e-06,
+      "loss": 1.0553,
+      "step": 35604
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.537514285714286e-06,
+      "loss": 1.3599,
+      "step": 35605
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.537245714285714e-06,
+      "loss": 1.2369,
+      "step": 35606
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.536977142857143e-06,
+      "loss": 1.1957,
+      "step": 35607
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5367085714285715e-06,
+      "loss": 1.4702,
+      "step": 35608
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.53644e-06,
+      "loss": 1.6126,
+      "step": 35609
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.536171428571429e-06,
+      "loss": 1.22,
+      "step": 35610
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.535902857142857e-06,
+      "loss": 1.2632,
+      "step": 35611
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.535634285714285e-06,
+      "loss": 1.0997,
+      "step": 35612
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.535365714285715e-06,
+      "loss": 0.7682,
+      "step": 35613
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.535097142857143e-06,
+      "loss": 0.727,
+      "step": 35614
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.534828571428572e-06,
+      "loss": 1.2047,
+      "step": 35615
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.53456e-06,
+      "loss": 0.764,
+      "step": 35616
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5342914285714285e-06,
+      "loss": 0.6572,
+      "step": 35617
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5340228571428575e-06,
+      "loss": 1.1276,
+      "step": 35618
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.533754285714286e-06,
+      "loss": 1.0926,
+      "step": 35619
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.533485714285715e-06,
+      "loss": 0.8303,
+      "step": 35620
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.533217142857143e-06,
+      "loss": 1.0202,
+      "step": 35621
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.532948571428571e-06,
+      "loss": 1.3031,
+      "step": 35622
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.53268e-06,
+      "loss": 1.0264,
+      "step": 35623
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.532411428571428e-06,
+      "loss": 1.6824,
+      "step": 35624
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.532142857142857e-06,
+      "loss": 0.8344,
+      "step": 35625
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.531874285714286e-06,
+      "loss": 1.0666,
+      "step": 35626
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5316057142857145e-06,
+      "loss": 0.7118,
+      "step": 35627
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5313371428571435e-06,
+      "loss": 1.2902,
+      "step": 35628
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.531068571428572e-06,
+      "loss": 1.3406,
+      "step": 35629
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.530800000000001e-06,
+      "loss": 0.7176,
+      "step": 35630
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.530531428571429e-06,
+      "loss": 1.3717,
+      "step": 35631
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.530262857142857e-06,
+      "loss": 1.0279,
+      "step": 35632
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.529994285714286e-06,
+      "loss": 1.1321,
+      "step": 35633
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.529725714285714e-06,
+      "loss": 1.3525,
+      "step": 35634
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5294571428571425e-06,
+      "loss": 1.3389,
+      "step": 35635
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5291885714285715e-06,
+      "loss": 1.637,
+      "step": 35636
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.52892e-06,
+      "loss": 1.1135,
+      "step": 35637
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.528651428571429e-06,
+      "loss": 0.8381,
+      "step": 35638
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.528382857142857e-06,
+      "loss": 1.2566,
+      "step": 35639
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.528114285714286e-06,
+      "loss": 1.2121,
+      "step": 35640
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.527845714285715e-06,
+      "loss": 1.1524,
+      "step": 35641
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.527577142857143e-06,
+      "loss": 0.9352,
+      "step": 35642
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.527308571428572e-06,
+      "loss": 1.0023,
+      "step": 35643
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.52704e-06,
+      "loss": 0.8532,
+      "step": 35644
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5267714285714285e-06,
+      "loss": 0.8458,
+      "step": 35645
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5265028571428575e-06,
+      "loss": 0.7561,
+      "step": 35646
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.526234285714286e-06,
+      "loss": 0.9392,
+      "step": 35647
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.525965714285715e-06,
+      "loss": 1.1053,
+      "step": 35648
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.525697142857143e-06,
+      "loss": 0.8339,
+      "step": 35649
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.525428571428571e-06,
+      "loss": 1.3763,
+      "step": 35650
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.52516e-06,
+      "loss": 1.1472,
+      "step": 35651
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.524891428571428e-06,
+      "loss": 1.2178,
+      "step": 35652
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.524622857142858e-06,
+      "loss": 0.6848,
+      "step": 35653
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.524354285714286e-06,
+      "loss": 1.2732,
+      "step": 35654
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5240857142857145e-06,
+      "loss": 0.9201,
+      "step": 35655
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5238171428571435e-06,
+      "loss": 1.3241,
+      "step": 35656
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.523548571428572e-06,
+      "loss": 1.5113,
+      "step": 35657
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.52328e-06,
+      "loss": 1.1277,
+      "step": 35658
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.523011428571429e-06,
+      "loss": 1.3806,
+      "step": 35659
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.522742857142857e-06,
+      "loss": 1.2759,
+      "step": 35660
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.522474285714286e-06,
+      "loss": 1.1406,
+      "step": 35661
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.522205714285714e-06,
+      "loss": 1.2589,
+      "step": 35662
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.521937142857142e-06,
+      "loss": 1.1328,
+      "step": 35663
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5216685714285714e-06,
+      "loss": 1.0801,
+      "step": 35664
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5214e-06,
+      "loss": 1.0662,
+      "step": 35665
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5211314285714295e-06,
+      "loss": 0.8493,
+      "step": 35666
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.520862857142858e-06,
+      "loss": 1.2423,
+      "step": 35667
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.520594285714286e-06,
+      "loss": 1.1849,
+      "step": 35668
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.520325714285715e-06,
+      "loss": 1.1268,
+      "step": 35669
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.520057142857143e-06,
+      "loss": 1.2029,
+      "step": 35670
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.519788571428572e-06,
+      "loss": 1.4612,
+      "step": 35671
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.51952e-06,
+      "loss": 1.2329,
+      "step": 35672
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.519251428571428e-06,
+      "loss": 1.6064,
+      "step": 35673
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5189828571428574e-06,
+      "loss": 1.2328,
+      "step": 35674
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.518714285714286e-06,
+      "loss": 0.9226,
+      "step": 35675
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.518445714285715e-06,
+      "loss": 1.758,
+      "step": 35676
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.518177142857143e-06,
+      "loss": 1.4176,
+      "step": 35677
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.517908571428571e-06,
+      "loss": 0.848,
+      "step": 35678
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.517640000000001e-06,
+      "loss": 0.9857,
+      "step": 35679
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.517371428571429e-06,
+      "loss": 1.3756,
+      "step": 35680
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.517102857142857e-06,
+      "loss": 1.3067,
+      "step": 35681
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.516834285714286e-06,
+      "loss": 1.0249,
+      "step": 35682
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.516565714285714e-06,
+      "loss": 0.6753,
+      "step": 35683
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.516297142857143e-06,
+      "loss": 1.3838,
+      "step": 35684
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.516028571428572e-06,
+      "loss": 1.5803,
+      "step": 35685
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.51576e-06,
+      "loss": 1.172,
+      "step": 35686
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.515491428571429e-06,
+      "loss": 1.2463,
+      "step": 35687
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.515222857142857e-06,
+      "loss": 1.2377,
+      "step": 35688
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.514954285714286e-06,
+      "loss": 0.9988,
+      "step": 35689
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.514685714285714e-06,
+      "loss": 1.3194,
+      "step": 35690
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.514417142857142e-06,
+      "loss": 1.05,
+      "step": 35691
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.514148571428571e-06,
+      "loss": 1.4156,
+      "step": 35692
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.51388e-06,
+      "loss": 1.4804,
+      "step": 35693
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.513611428571429e-06,
+      "loss": 1.4091,
+      "step": 35694
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.513342857142858e-06,
+      "loss": 1.1364,
+      "step": 35695
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.513074285714286e-06,
+      "loss": 1.0019,
+      "step": 35696
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.512805714285715e-06,
+      "loss": 1.3463,
+      "step": 35697
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.512537142857143e-06,
+      "loss": 0.8789,
+      "step": 35698
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.512268571428571e-06,
+      "loss": 0.6523,
+      "step": 35699
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.512e-06,
+      "loss": 1.7437,
+      "step": 35700
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.511731428571428e-06,
+      "loss": 1.3378,
+      "step": 35701
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.511462857142857e-06,
+      "loss": 0.8605,
+      "step": 35702
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5111942857142855e-06,
+      "loss": 1.3421,
+      "step": 35703
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.510925714285714e-06,
+      "loss": 1.6597,
+      "step": 35704
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.510657142857143e-06,
+      "loss": 0.9497,
+      "step": 35705
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.510388571428572e-06,
+      "loss": 1.1035,
+      "step": 35706
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.510120000000001e-06,
+      "loss": 1.1724,
+      "step": 35707
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.509851428571429e-06,
+      "loss": 1.4862,
+      "step": 35708
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.509582857142857e-06,
+      "loss": 1.3389,
+      "step": 35709
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.509314285714286e-06,
+      "loss": 1.0478,
+      "step": 35710
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.509045714285714e-06,
+      "loss": 1.1608,
+      "step": 35711
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.508777142857143e-06,
+      "loss": 0.9365,
+      "step": 35712
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5085085714285715e-06,
+      "loss": 1.2142,
+      "step": 35713
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.50824e-06,
+      "loss": 1.1599,
+      "step": 35714
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.507971428571429e-06,
+      "loss": 1.3816,
+      "step": 35715
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.507702857142857e-06,
+      "loss": 1.2359,
+      "step": 35716
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.507434285714286e-06,
+      "loss": 1.2538,
+      "step": 35717
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.507165714285714e-06,
+      "loss": 1.536,
+      "step": 35718
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.506897142857143e-06,
+      "loss": 0.832,
+      "step": 35719
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.506628571428572e-06,
+      "loss": 0.7386,
+      "step": 35720
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.50636e-06,
+      "loss": 1.2227,
+      "step": 35721
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5060914285714285e-06,
+      "loss": 1.2621,
+      "step": 35722
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5058228571428575e-06,
+      "loss": 1.0862,
+      "step": 35723
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.505554285714286e-06,
+      "loss": 0.8228,
+      "step": 35724
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.505285714285715e-06,
+      "loss": 0.963,
+      "step": 35725
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.505017142857143e-06,
+      "loss": 1.3267,
+      "step": 35726
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.504748571428571e-06,
+      "loss": 1.3905,
+      "step": 35727
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.50448e-06,
+      "loss": 0.8323,
+      "step": 35728
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.504211428571428e-06,
+      "loss": 1.1076,
+      "step": 35729
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.503942857142857e-06,
+      "loss": 1.5405,
+      "step": 35730
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5036742857142855e-06,
+      "loss": 1.155,
+      "step": 35731
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5034057142857145e-06,
+      "loss": 0.5978,
+      "step": 35732
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5031371428571435e-06,
+      "loss": 1.1146,
+      "step": 35733
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.502868571428572e-06,
+      "loss": 1.0055,
+      "step": 35734
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.502600000000001e-06,
+      "loss": 1.0109,
+      "step": 35735
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.502331428571429e-06,
+      "loss": 0.9581,
+      "step": 35736
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.502062857142857e-06,
+      "loss": 1.3218,
+      "step": 35737
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.501794285714286e-06,
+      "loss": 1.3285,
+      "step": 35738
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.501525714285714e-06,
+      "loss": 0.9073,
+      "step": 35739
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.501257142857143e-06,
+      "loss": 1.2535,
+      "step": 35740
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5009885714285715e-06,
+      "loss": 0.7178,
+      "step": 35741
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.50072e-06,
+      "loss": 1.131,
+      "step": 35742
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.500451428571429e-06,
+      "loss": 1.1985,
+      "step": 35743
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.500182857142857e-06,
+      "loss": 1.3777,
+      "step": 35744
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.499914285714286e-06,
+      "loss": 1.1955,
+      "step": 35745
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.499645714285715e-06,
+      "loss": 1.2364,
+      "step": 35746
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.499377142857143e-06,
+      "loss": 0.8572,
+      "step": 35747
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.499108571428572e-06,
+      "loss": 0.6798,
+      "step": 35748
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.49884e-06,
+      "loss": 1.0037,
+      "step": 35749
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4985714285714285e-06,
+      "loss": 1.1421,
+      "step": 35750
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4983028571428575e-06,
+      "loss": 1.2704,
+      "step": 35751
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.498034285714286e-06,
+      "loss": 1.225,
+      "step": 35752
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.497765714285715e-06,
+      "loss": 1.17,
+      "step": 35753
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.497497142857143e-06,
+      "loss": 1.3687,
+      "step": 35754
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.497228571428571e-06,
+      "loss": 0.8375,
+      "step": 35755
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.49696e-06,
+      "loss": 1.3561,
+      "step": 35756
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.496691428571428e-06,
+      "loss": 1.6669,
+      "step": 35757
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.496422857142857e-06,
+      "loss": 0.8767,
+      "step": 35758
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.496154285714286e-06,
+      "loss": 1.2588,
+      "step": 35759
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4958857142857145e-06,
+      "loss": 1.1342,
+      "step": 35760
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4956171428571435e-06,
+      "loss": 0.9676,
+      "step": 35761
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.495348571428572e-06,
+      "loss": 1.3146,
+      "step": 35762
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.49508e-06,
+      "loss": 1.1624,
+      "step": 35763
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.494811428571429e-06,
+      "loss": 1.3247,
+      "step": 35764
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.494542857142857e-06,
+      "loss": 0.9203,
+      "step": 35765
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.494274285714286e-06,
+      "loss": 0.9038,
+      "step": 35766
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.494005714285714e-06,
+      "loss": 1.1765,
+      "step": 35767
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.493737142857142e-06,
+      "loss": 1.0267,
+      "step": 35768
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4934685714285714e-06,
+      "loss": 1.0576,
+      "step": 35769
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4932e-06,
+      "loss": 0.7177,
+      "step": 35770
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.492931428571429e-06,
+      "loss": 1.2316,
+      "step": 35771
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.492662857142858e-06,
+      "loss": 1.3299,
+      "step": 35772
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.492394285714286e-06,
+      "loss": 1.3587,
+      "step": 35773
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.492125714285715e-06,
+      "loss": 1.0349,
+      "step": 35774
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.491857142857143e-06,
+      "loss": 1.4252,
+      "step": 35775
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.491588571428572e-06,
+      "loss": 1.0557,
+      "step": 35776
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.49132e-06,
+      "loss": 1.2614,
+      "step": 35777
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.491051428571428e-06,
+      "loss": 1.0655,
+      "step": 35778
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4907828571428574e-06,
+      "loss": 1.1901,
+      "step": 35779
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.490514285714286e-06,
+      "loss": 1.3411,
+      "step": 35780
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.490245714285715e-06,
+      "loss": 1.6784,
+      "step": 35781
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.489977142857143e-06,
+      "loss": 1.258,
+      "step": 35782
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.489708571428571e-06,
+      "loss": 1.2279,
+      "step": 35783
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.48944e-06,
+      "loss": 1.2076,
+      "step": 35784
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.489171428571429e-06,
+      "loss": 0.9549,
+      "step": 35785
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.488902857142857e-06,
+      "loss": 0.786,
+      "step": 35786
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.488634285714286e-06,
+      "loss": 0.9911,
+      "step": 35787
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.488365714285714e-06,
+      "loss": 0.99,
+      "step": 35788
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.488097142857143e-06,
+      "loss": 1.0745,
+      "step": 35789
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.487828571428572e-06,
+      "loss": 1.1208,
+      "step": 35790
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.48756e-06,
+      "loss": 1.148,
+      "step": 35791
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.487291428571429e-06,
+      "loss": 1.0276,
+      "step": 35792
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.487022857142857e-06,
+      "loss": 0.9913,
+      "step": 35793
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.486754285714286e-06,
+      "loss": 1.125,
+      "step": 35794
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.486485714285714e-06,
+      "loss": 1.2129,
+      "step": 35795
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.486217142857142e-06,
+      "loss": 1.1894,
+      "step": 35796
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.485948571428571e-06,
+      "loss": 1.0432,
+      "step": 35797
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.48568e-06,
+      "loss": 1.1738,
+      "step": 35798
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.485411428571429e-06,
+      "loss": 1.2401,
+      "step": 35799
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.485142857142858e-06,
+      "loss": 1.1439,
+      "step": 35800
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.484874285714286e-06,
+      "loss": 0.8643,
+      "step": 35801
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.484605714285715e-06,
+      "loss": 1.4621,
+      "step": 35802
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.484337142857143e-06,
+      "loss": 0.9738,
+      "step": 35803
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.484068571428572e-06,
+      "loss": 0.8611,
+      "step": 35804
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4838e-06,
+      "loss": 1.389,
+      "step": 35805
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.483531428571428e-06,
+      "loss": 1.421,
+      "step": 35806
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.483262857142857e-06,
+      "loss": 1.0038,
+      "step": 35807
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4829942857142855e-06,
+      "loss": 0.9982,
+      "step": 35808
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.482725714285714e-06,
+      "loss": 0.8329,
+      "step": 35809
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.482457142857143e-06,
+      "loss": 1.4562,
+      "step": 35810
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.482188571428572e-06,
+      "loss": 1.2249,
+      "step": 35811
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.481920000000001e-06,
+      "loss": 0.8626,
+      "step": 35812
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.481651428571429e-06,
+      "loss": 1.2324,
+      "step": 35813
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.481382857142857e-06,
+      "loss": 0.9493,
+      "step": 35814
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.481114285714286e-06,
+      "loss": 1.2882,
+      "step": 35815
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.480845714285714e-06,
+      "loss": 0.9611,
+      "step": 35816
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.480577142857143e-06,
+      "loss": 1.3051,
+      "step": 35817
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4803085714285715e-06,
+      "loss": 1.1463,
+      "step": 35818
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.48004e-06,
+      "loss": 0.893,
+      "step": 35819
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.479771428571429e-06,
+      "loss": 0.906,
+      "step": 35820
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.479502857142857e-06,
+      "loss": 1.4365,
+      "step": 35821
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.479234285714286e-06,
+      "loss": 1.2193,
+      "step": 35822
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.478965714285714e-06,
+      "loss": 0.8836,
+      "step": 35823
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.478697142857143e-06,
+      "loss": 1.0589,
+      "step": 35824
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.478428571428572e-06,
+      "loss": 0.8155,
+      "step": 35825
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.47816e-06,
+      "loss": 1.1897,
+      "step": 35826
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.477891428571429e-06,
+      "loss": 1.2546,
+      "step": 35827
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4776228571428575e-06,
+      "loss": 1.1544,
+      "step": 35828
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.477354285714286e-06,
+      "loss": 1.2548,
+      "step": 35829
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.477085714285715e-06,
+      "loss": 1.5673,
+      "step": 35830
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.476817142857143e-06,
+      "loss": 1.0997,
+      "step": 35831
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.476548571428571e-06,
+      "loss": 1.0558,
+      "step": 35832
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.47628e-06,
+      "loss": 0.679,
+      "step": 35833
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.476011428571428e-06,
+      "loss": 1.2677,
+      "step": 35834
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.475742857142857e-06,
+      "loss": 0.9246,
+      "step": 35835
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4754742857142855e-06,
+      "loss": 0.897,
+      "step": 35836
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4752057142857145e-06,
+      "loss": 1.0429,
+      "step": 35837
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4749371428571435e-06,
+      "loss": 1.1607,
+      "step": 35838
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.474668571428572e-06,
+      "loss": 0.9766,
+      "step": 35839
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.474400000000001e-06,
+      "loss": 0.9412,
+      "step": 35840
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.474131428571429e-06,
+      "loss": 1.2741,
+      "step": 35841
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.473862857142857e-06,
+      "loss": 1.0325,
+      "step": 35842
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.473594285714286e-06,
+      "loss": 0.9576,
+      "step": 35843
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.473325714285714e-06,
+      "loss": 0.6468,
+      "step": 35844
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.473057142857143e-06,
+      "loss": 1.0666,
+      "step": 35845
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4727885714285715e-06,
+      "loss": 1.1682,
+      "step": 35846
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.47252e-06,
+      "loss": 1.1352,
+      "step": 35847
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.472251428571429e-06,
+      "loss": 1.0307,
+      "step": 35848
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.471982857142857e-06,
+      "loss": 1.27,
+      "step": 35849
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.471714285714286e-06,
+      "loss": 1.0163,
+      "step": 35850
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.471445714285715e-06,
+      "loss": 1.0385,
+      "step": 35851
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.471177142857143e-06,
+      "loss": 1.3374,
+      "step": 35852
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.470908571428572e-06,
+      "loss": 1.1622,
+      "step": 35853
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.47064e-06,
+      "loss": 1.0362,
+      "step": 35854
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4703714285714285e-06,
+      "loss": 1.0842,
+      "step": 35855
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4701028571428575e-06,
+      "loss": 1.1811,
+      "step": 35856
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.469834285714286e-06,
+      "loss": 1.3343,
+      "step": 35857
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.469565714285715e-06,
+      "loss": 1.5219,
+      "step": 35858
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.469297142857143e-06,
+      "loss": 1.4352,
+      "step": 35859
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.469028571428571e-06,
+      "loss": 0.9392,
+      "step": 35860
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.46876e-06,
+      "loss": 1.3345,
+      "step": 35861
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.468491428571428e-06,
+      "loss": 1.1429,
+      "step": 35862
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.468222857142857e-06,
+      "loss": 0.9392,
+      "step": 35863
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.467954285714286e-06,
+      "loss": 1.2109,
+      "step": 35864
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4676857142857145e-06,
+      "loss": 1.0713,
+      "step": 35865
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4674171428571435e-06,
+      "loss": 1.0505,
+      "step": 35866
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.467148571428572e-06,
+      "loss": 1.0339,
+      "step": 35867
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.466880000000001e-06,
+      "loss": 1.2401,
+      "step": 35868
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.466611428571429e-06,
+      "loss": 1.1735,
+      "step": 35869
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.466342857142857e-06,
+      "loss": 1.3223,
+      "step": 35870
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.466074285714286e-06,
+      "loss": 1.7405,
+      "step": 35871
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.465805714285714e-06,
+      "loss": 1.2382,
+      "step": 35872
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.465537142857142e-06,
+      "loss": 0.8734,
+      "step": 35873
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4652685714285714e-06,
+      "loss": 1.5131,
+      "step": 35874
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.465e-06,
+      "loss": 1.2267,
+      "step": 35875
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.464731428571429e-06,
+      "loss": 1.2978,
+      "step": 35876
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.464462857142858e-06,
+      "loss": 1.0679,
+      "step": 35877
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.464194285714286e-06,
+      "loss": 0.9621,
+      "step": 35878
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.463925714285715e-06,
+      "loss": 1.0205,
+      "step": 35879
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.463657142857143e-06,
+      "loss": 1.0174,
+      "step": 35880
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.463388571428572e-06,
+      "loss": 1.1006,
+      "step": 35881
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.46312e-06,
+      "loss": 1.3288,
+      "step": 35882
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.462851428571428e-06,
+      "loss": 0.9199,
+      "step": 35883
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.462582857142857e-06,
+      "loss": 1.1168,
+      "step": 35884
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.462314285714286e-06,
+      "loss": 1.0532,
+      "step": 35885
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.462045714285715e-06,
+      "loss": 1.339,
+      "step": 35886
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.461777142857143e-06,
+      "loss": 1.1234,
+      "step": 35887
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.461508571428571e-06,
+      "loss": 0.9717,
+      "step": 35888
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.46124e-06,
+      "loss": 0.9639,
+      "step": 35889
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.460971428571429e-06,
+      "loss": 0.8199,
+      "step": 35890
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.460702857142858e-06,
+      "loss": 0.8025,
+      "step": 35891
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.460434285714286e-06,
+      "loss": 0.543,
+      "step": 35892
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.460165714285714e-06,
+      "loss": 1.3218,
+      "step": 35893
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.459897142857143e-06,
+      "loss": 1.3051,
+      "step": 35894
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.459628571428572e-06,
+      "loss": 0.9574,
+      "step": 35895
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.45936e-06,
+      "loss": 1.3477,
+      "step": 35896
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.459091428571429e-06,
+      "loss": 0.91,
+      "step": 35897
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.458822857142857e-06,
+      "loss": 1.3509,
+      "step": 35898
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.458554285714286e-06,
+      "loss": 1.0527,
+      "step": 35899
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.458285714285714e-06,
+      "loss": 0.9977,
+      "step": 35900
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.458017142857142e-06,
+      "loss": 1.0588,
+      "step": 35901
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.457748571428571e-06,
+      "loss": 1.2035,
+      "step": 35902
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.45748e-06,
+      "loss": 1.0462,
+      "step": 35903
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.457211428571429e-06,
+      "loss": 1.2496,
+      "step": 35904
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.456942857142858e-06,
+      "loss": 0.9558,
+      "step": 35905
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.456674285714286e-06,
+      "loss": 1.3213,
+      "step": 35906
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.456405714285715e-06,
+      "loss": 0.979,
+      "step": 35907
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.456137142857143e-06,
+      "loss": 0.9179,
+      "step": 35908
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.455868571428572e-06,
+      "loss": 0.8701,
+      "step": 35909
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4556e-06,
+      "loss": 1.3427,
+      "step": 35910
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.455331428571428e-06,
+      "loss": 1.4558,
+      "step": 35911
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.455062857142857e-06,
+      "loss": 0.7499,
+      "step": 35912
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4547942857142855e-06,
+      "loss": 1.1295,
+      "step": 35913
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4545257142857146e-06,
+      "loss": 1.271,
+      "step": 35914
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.454257142857143e-06,
+      "loss": 1.1236,
+      "step": 35915
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.453988571428572e-06,
+      "loss": 0.993,
+      "step": 35916
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.453720000000001e-06,
+      "loss": 1.1197,
+      "step": 35917
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.453451428571429e-06,
+      "loss": 1.2778,
+      "step": 35918
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.453182857142857e-06,
+      "loss": 1.2897,
+      "step": 35919
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.452914285714286e-06,
+      "loss": 1.062,
+      "step": 35920
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.452645714285714e-06,
+      "loss": 1.1915,
+      "step": 35921
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.452377142857143e-06,
+      "loss": 0.7794,
+      "step": 35922
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4521085714285715e-06,
+      "loss": 1.2348,
+      "step": 35923
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.45184e-06,
+      "loss": 1.2294,
+      "step": 35924
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.451571428571429e-06,
+      "loss": 1.4948,
+      "step": 35925
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.451302857142857e-06,
+      "loss": 1.1765,
+      "step": 35926
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.451034285714286e-06,
+      "loss": 0.9905,
+      "step": 35927
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.450765714285714e-06,
+      "loss": 1.4734,
+      "step": 35928
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.450497142857143e-06,
+      "loss": 1.1725,
+      "step": 35929
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.450228571428572e-06,
+      "loss": 0.8988,
+      "step": 35930
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.44996e-06,
+      "loss": 1.3163,
+      "step": 35931
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.449691428571429e-06,
+      "loss": 0.79,
+      "step": 35932
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4494228571428575e-06,
+      "loss": 0.9396,
+      "step": 35933
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.449154285714286e-06,
+      "loss": 1.0616,
+      "step": 35934
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.448885714285715e-06,
+      "loss": 1.49,
+      "step": 35935
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.448617142857143e-06,
+      "loss": 0.979,
+      "step": 35936
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.448348571428571e-06,
+      "loss": 1.1109,
+      "step": 35937
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.44808e-06,
+      "loss": 1.4535,
+      "step": 35938
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.447811428571428e-06,
+      "loss": 0.9414,
+      "step": 35939
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.447542857142857e-06,
+      "loss": 1.0857,
+      "step": 35940
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4472742857142855e-06,
+      "loss": 1.5047,
+      "step": 35941
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4470057142857145e-06,
+      "loss": 0.9236,
+      "step": 35942
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4467371428571435e-06,
+      "loss": 0.8743,
+      "step": 35943
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.446468571428572e-06,
+      "loss": 1.0465,
+      "step": 35944
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.446200000000001e-06,
+      "loss": 1.1447,
+      "step": 35945
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.445931428571429e-06,
+      "loss": 1.3055,
+      "step": 35946
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.445662857142857e-06,
+      "loss": 1.5897,
+      "step": 35947
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.445394285714286e-06,
+      "loss": 1.127,
+      "step": 35948
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.445125714285714e-06,
+      "loss": 0.8705,
+      "step": 35949
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.444857142857143e-06,
+      "loss": 1.5272,
+      "step": 35950
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4445885714285715e-06,
+      "loss": 0.7153,
+      "step": 35951
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.44432e-06,
+      "loss": 1.2172,
+      "step": 35952
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.444051428571429e-06,
+      "loss": 1.2502,
+      "step": 35953
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.443782857142857e-06,
+      "loss": 0.9818,
+      "step": 35954
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.443514285714286e-06,
+      "loss": 1.0305,
+      "step": 35955
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.443245714285715e-06,
+      "loss": 0.9466,
+      "step": 35956
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.442977142857143e-06,
+      "loss": 1.2526,
+      "step": 35957
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.442708571428572e-06,
+      "loss": 1.1592,
+      "step": 35958
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.44244e-06,
+      "loss": 1.283,
+      "step": 35959
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4421714285714285e-06,
+      "loss": 1.4211,
+      "step": 35960
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4419028571428575e-06,
+      "loss": 0.8469,
+      "step": 35961
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.441634285714286e-06,
+      "loss": 1.088,
+      "step": 35962
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.441365714285715e-06,
+      "loss": 0.7895,
+      "step": 35963
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.441097142857143e-06,
+      "loss": 1.5328,
+      "step": 35964
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.440828571428571e-06,
+      "loss": 1.1008,
+      "step": 35965
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.44056e-06,
+      "loss": 1.2256,
+      "step": 35966
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.440291428571428e-06,
+      "loss": 1.0338,
+      "step": 35967
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.440022857142857e-06,
+      "loss": 1.3046,
+      "step": 35968
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.439754285714286e-06,
+      "loss": 0.7172,
+      "step": 35969
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4394857142857145e-06,
+      "loss": 0.9912,
+      "step": 35970
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4392171428571435e-06,
+      "loss": 0.967,
+      "step": 35971
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.438948571428572e-06,
+      "loss": 1.1136,
+      "step": 35972
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.438680000000001e-06,
+      "loss": 1.5942,
+      "step": 35973
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.438411428571429e-06,
+      "loss": 1.0586,
+      "step": 35974
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.438142857142857e-06,
+      "loss": 1.0352,
+      "step": 35975
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.437874285714286e-06,
+      "loss": 0.9462,
+      "step": 35976
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.437605714285714e-06,
+      "loss": 1.3955,
+      "step": 35977
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.437337142857143e-06,
+      "loss": 0.9772,
+      "step": 35978
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4370685714285714e-06,
+      "loss": 0.5939,
+      "step": 35979
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4368e-06,
+      "loss": 1.2489,
+      "step": 35980
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.436531428571429e-06,
+      "loss": 1.2648,
+      "step": 35981
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.436262857142858e-06,
+      "loss": 1.4933,
+      "step": 35982
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.435994285714286e-06,
+      "loss": 1.0932,
+      "step": 35983
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.435725714285715e-06,
+      "loss": 1.1594,
+      "step": 35984
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.435457142857143e-06,
+      "loss": 1.3813,
+      "step": 35985
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.435188571428572e-06,
+      "loss": 1.6397,
+      "step": 35986
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.43492e-06,
+      "loss": 1.4426,
+      "step": 35987
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.434651428571428e-06,
+      "loss": 1.2254,
+      "step": 35988
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.434382857142857e-06,
+      "loss": 1.0818,
+      "step": 35989
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.434114285714286e-06,
+      "loss": 0.8077,
+      "step": 35990
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.433845714285715e-06,
+      "loss": 1.2236,
+      "step": 35991
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.433577142857143e-06,
+      "loss": 1.506,
+      "step": 35992
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.433308571428571e-06,
+      "loss": 1.1096,
+      "step": 35993
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.43304e-06,
+      "loss": 1.1283,
+      "step": 35994
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.432771428571429e-06,
+      "loss": 1.2638,
+      "step": 35995
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.432502857142858e-06,
+      "loss": 0.7783,
+      "step": 35996
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.432234285714286e-06,
+      "loss": 1.1135,
+      "step": 35997
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.431965714285714e-06,
+      "loss": 1.3442,
+      "step": 35998
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.431697142857143e-06,
+      "loss": 1.327,
+      "step": 35999
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.431428571428572e-06,
+      "loss": 1.4023,
+      "step": 36000
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.431160000000001e-06,
+      "loss": 1.1791,
+      "step": 36001
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.430891428571429e-06,
+      "loss": 0.7478,
+      "step": 36002
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.430622857142857e-06,
+      "loss": 1.3296,
+      "step": 36003
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.430354285714286e-06,
+      "loss": 1.3135,
+      "step": 36004
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.430085714285714e-06,
+      "loss": 1.2741,
+      "step": 36005
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.429817142857142e-06,
+      "loss": 0.8129,
+      "step": 36006
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.429548571428571e-06,
+      "loss": 1.1922,
+      "step": 36007
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.42928e-06,
+      "loss": 1.3604,
+      "step": 36008
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.429011428571429e-06,
+      "loss": 0.9273,
+      "step": 36009
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.428742857142858e-06,
+      "loss": 1.2627,
+      "step": 36010
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.428474285714286e-06,
+      "loss": 1.0435,
+      "step": 36011
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.428205714285715e-06,
+      "loss": 1.2814,
+      "step": 36012
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.427937142857143e-06,
+      "loss": 1.1453,
+      "step": 36013
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.427668571428572e-06,
+      "loss": 1.0996,
+      "step": 36014
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4274e-06,
+      "loss": 1.3884,
+      "step": 36015
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.427131428571428e-06,
+      "loss": 1.2666,
+      "step": 36016
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.426862857142857e-06,
+      "loss": 0.999,
+      "step": 36017
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4265942857142855e-06,
+      "loss": 0.9657,
+      "step": 36018
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4263257142857146e-06,
+      "loss": 1.2206,
+      "step": 36019
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.426057142857143e-06,
+      "loss": 1.0701,
+      "step": 36020
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.425788571428572e-06,
+      "loss": 1.0227,
+      "step": 36021
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.425520000000001e-06,
+      "loss": 1.1222,
+      "step": 36022
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.425251428571429e-06,
+      "loss": 0.8937,
+      "step": 36023
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.424982857142857e-06,
+      "loss": 1.2442,
+      "step": 36024
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.424714285714286e-06,
+      "loss": 1.1562,
+      "step": 36025
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.424445714285714e-06,
+      "loss": 0.8623,
+      "step": 36026
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.424177142857143e-06,
+      "loss": 1.4091,
+      "step": 36027
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4239085714285715e-06,
+      "loss": 1.1715,
+      "step": 36028
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.42364e-06,
+      "loss": 1.3064,
+      "step": 36029
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.423371428571429e-06,
+      "loss": 1.1198,
+      "step": 36030
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.423102857142857e-06,
+      "loss": 0.9516,
+      "step": 36031
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.422834285714286e-06,
+      "loss": 1.4352,
+      "step": 36032
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.422565714285714e-06,
+      "loss": 0.9897,
+      "step": 36033
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.422297142857142e-06,
+      "loss": 1.7238,
+      "step": 36034
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.422028571428572e-06,
+      "loss": 0.9709,
+      "step": 36035
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.42176e-06,
+      "loss": 1.4281,
+      "step": 36036
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.421491428571429e-06,
+      "loss": 0.9332,
+      "step": 36037
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4212228571428575e-06,
+      "loss": 1.1737,
+      "step": 36038
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.420954285714286e-06,
+      "loss": 1.2541,
+      "step": 36039
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.420685714285715e-06,
+      "loss": 1.1178,
+      "step": 36040
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.420417142857143e-06,
+      "loss": 0.9869,
+      "step": 36041
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.420148571428572e-06,
+      "loss": 1.4021,
+      "step": 36042
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.41988e-06,
+      "loss": 1.0044,
+      "step": 36043
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.419611428571428e-06,
+      "loss": 1.5715,
+      "step": 36044
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.419342857142857e-06,
+      "loss": 1.1492,
+      "step": 36045
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4190742857142855e-06,
+      "loss": 1.535,
+      "step": 36046
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.418805714285714e-06,
+      "loss": 1.0038,
+      "step": 36047
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4185371428571435e-06,
+      "loss": 0.8732,
+      "step": 36048
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.418268571428572e-06,
+      "loss": 0.8997,
+      "step": 36049
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.418000000000001e-06,
+      "loss": 0.9333,
+      "step": 36050
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.417731428571429e-06,
+      "loss": 1.1645,
+      "step": 36051
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.417462857142857e-06,
+      "loss": 1.1989,
+      "step": 36052
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.417194285714286e-06,
+      "loss": 0.5994,
+      "step": 36053
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.416925714285714e-06,
+      "loss": 0.8354,
+      "step": 36054
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.416657142857143e-06,
+      "loss": 1.3223,
+      "step": 36055
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4163885714285715e-06,
+      "loss": 1.1962,
+      "step": 36056
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.41612e-06,
+      "loss": 0.9687,
+      "step": 36057
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.415851428571429e-06,
+      "loss": 1.0739,
+      "step": 36058
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.415582857142857e-06,
+      "loss": 1.1394,
+      "step": 36059
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.415314285714286e-06,
+      "loss": 1.258,
+      "step": 36060
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.415045714285715e-06,
+      "loss": 1.051,
+      "step": 36061
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.414777142857143e-06,
+      "loss": 1.371,
+      "step": 36062
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.414508571428572e-06,
+      "loss": 0.9614,
+      "step": 36063
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.41424e-06,
+      "loss": 0.6209,
+      "step": 36064
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.413971428571429e-06,
+      "loss": 0.8851,
+      "step": 36065
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4137028571428575e-06,
+      "loss": 1.5125,
+      "step": 36066
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.413434285714286e-06,
+      "loss": 0.7982,
+      "step": 36067
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.413165714285715e-06,
+      "loss": 0.9741,
+      "step": 36068
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.412897142857143e-06,
+      "loss": 1.1489,
+      "step": 36069
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.412628571428571e-06,
+      "loss": 0.9944,
+      "step": 36070
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.41236e-06,
+      "loss": 1.0644,
+      "step": 36071
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.412091428571428e-06,
+      "loss": 0.7501,
+      "step": 36072
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.411822857142857e-06,
+      "loss": 1.3708,
+      "step": 36073
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.411554285714286e-06,
+      "loss": 0.8537,
+      "step": 36074
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4112857142857145e-06,
+      "loss": 1.1452,
+      "step": 36075
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4110171428571435e-06,
+      "loss": 1.4214,
+      "step": 36076
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.410748571428572e-06,
+      "loss": 1.3673,
+      "step": 36077
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.410480000000001e-06,
+      "loss": 1.1529,
+      "step": 36078
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.410211428571429e-06,
+      "loss": 1.2179,
+      "step": 36079
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.409942857142857e-06,
+      "loss": 1.6421,
+      "step": 36080
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.409674285714286e-06,
+      "loss": 0.9651,
+      "step": 36081
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.409405714285714e-06,
+      "loss": 1.5972,
+      "step": 36082
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.409137142857143e-06,
+      "loss": 0.9001,
+      "step": 36083
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4088685714285714e-06,
+      "loss": 1.0354,
+      "step": 36084
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4086e-06,
+      "loss": 0.8604,
+      "step": 36085
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.408331428571429e-06,
+      "loss": 1.1018,
+      "step": 36086
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.408062857142857e-06,
+      "loss": 1.3757,
+      "step": 36087
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.407794285714287e-06,
+      "loss": 0.6292,
+      "step": 36088
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.407525714285715e-06,
+      "loss": 1.0586,
+      "step": 36089
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.407257142857143e-06,
+      "loss": 1.2453,
+      "step": 36090
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.406988571428572e-06,
+      "loss": 1.0412,
+      "step": 36091
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.40672e-06,
+      "loss": 1.4656,
+      "step": 36092
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.406451428571428e-06,
+      "loss": 0.7288,
+      "step": 36093
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.406182857142857e-06,
+      "loss": 1.3372,
+      "step": 36094
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.405914285714286e-06,
+      "loss": 1.2441,
+      "step": 36095
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.405645714285715e-06,
+      "loss": 1.2594,
+      "step": 36096
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.405377142857143e-06,
+      "loss": 1.0455,
+      "step": 36097
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.405108571428571e-06,
+      "loss": 0.8574,
+      "step": 36098
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.40484e-06,
+      "loss": 1.1827,
+      "step": 36099
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.404571428571428e-06,
+      "loss": 1.0535,
+      "step": 36100
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.404302857142858e-06,
+      "loss": 0.9586,
+      "step": 36101
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.404034285714286e-06,
+      "loss": 1.1496,
+      "step": 36102
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.403765714285714e-06,
+      "loss": 1.3157,
+      "step": 36103
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.403497142857143e-06,
+      "loss": 1.5983,
+      "step": 36104
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.403228571428572e-06,
+      "loss": 1.4566,
+      "step": 36105
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.402960000000001e-06,
+      "loss": 1.1181,
+      "step": 36106
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.402691428571429e-06,
+      "loss": 1.3292,
+      "step": 36107
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.402422857142857e-06,
+      "loss": 1.1425,
+      "step": 36108
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.402154285714286e-06,
+      "loss": 1.1236,
+      "step": 36109
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.401885714285714e-06,
+      "loss": 1.4048,
+      "step": 36110
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.401617142857142e-06,
+      "loss": 1.106,
+      "step": 36111
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.401348571428571e-06,
+      "loss": 0.8043,
+      "step": 36112
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4010799999999995e-06,
+      "loss": 1.1317,
+      "step": 36113
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.400811428571429e-06,
+      "loss": 0.8968,
+      "step": 36114
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.400542857142858e-06,
+      "loss": 1.3293,
+      "step": 36115
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.400274285714286e-06,
+      "loss": 0.9369,
+      "step": 36116
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.400005714285715e-06,
+      "loss": 1.2127,
+      "step": 36117
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.399737142857143e-06,
+      "loss": 1.3803,
+      "step": 36118
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.399468571428572e-06,
+      "loss": 1.0451,
+      "step": 36119
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.3992e-06,
+      "loss": 1.0751,
+      "step": 36120
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.398931428571428e-06,
+      "loss": 1.027,
+      "step": 36121
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.398662857142857e-06,
+      "loss": 1.418,
+      "step": 36122
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.3983942857142855e-06,
+      "loss": 1.2836,
+      "step": 36123
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.3981257142857146e-06,
+      "loss": 1.4694,
+      "step": 36124
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.397857142857143e-06,
+      "loss": 1.2595,
+      "step": 36125
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.397588571428571e-06,
+      "loss": 1.3602,
+      "step": 36126
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.397320000000001e-06,
+      "loss": 1.1964,
+      "step": 36127
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.397051428571429e-06,
+      "loss": 0.7688,
+      "step": 36128
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.396782857142858e-06,
+      "loss": 1.2875,
+      "step": 36129
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.396514285714286e-06,
+      "loss": 0.9459,
+      "step": 36130
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.396245714285714e-06,
+      "loss": 1.2892,
+      "step": 36131
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.395977142857143e-06,
+      "loss": 0.9741,
+      "step": 36132
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.3957085714285715e-06,
+      "loss": 0.9446,
+      "step": 36133
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.39544e-06,
+      "loss": 1.0552,
+      "step": 36134
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.395171428571429e-06,
+      "loss": 1.2455,
+      "step": 36135
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.394902857142857e-06,
+      "loss": 0.8041,
+      "step": 36136
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.394634285714286e-06,
+      "loss": 1.1865,
+      "step": 36137
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.394365714285714e-06,
+      "loss": 1.6106,
+      "step": 36138
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.394097142857142e-06,
+      "loss": 1.0214,
+      "step": 36139
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.393828571428571e-06,
+      "loss": 1.3303,
+      "step": 36140
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.39356e-06,
+      "loss": 0.8492,
+      "step": 36141
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.393291428571429e-06,
+      "loss": 1.1812,
+      "step": 36142
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3930228571428575e-06,
+      "loss": 0.9777,
+      "step": 36143
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.392754285714286e-06,
+      "loss": 0.6832,
+      "step": 36144
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.392485714285715e-06,
+      "loss": 1.143,
+      "step": 36145
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.392217142857143e-06,
+      "loss": 1.13,
+      "step": 36146
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.391948571428572e-06,
+      "loss": 1.3188,
+      "step": 36147
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.39168e-06,
+      "loss": 1.3576,
+      "step": 36148
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.391411428571428e-06,
+      "loss": 1.2114,
+      "step": 36149
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.391142857142857e-06,
+      "loss": 1.2246,
+      "step": 36150
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3908742857142855e-06,
+      "loss": 1.2933,
+      "step": 36151
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3906057142857145e-06,
+      "loss": 0.8203,
+      "step": 36152
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.390337142857143e-06,
+      "loss": 1.7456,
+      "step": 36153
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.390068571428572e-06,
+      "loss": 1.1014,
+      "step": 36154
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.389800000000001e-06,
+      "loss": 1.1768,
+      "step": 36155
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.389531428571429e-06,
+      "loss": 1.2652,
+      "step": 36156
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.389262857142857e-06,
+      "loss": 0.8565,
+      "step": 36157
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.388994285714286e-06,
+      "loss": 1.2929,
+      "step": 36158
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.388725714285714e-06,
+      "loss": 0.6653,
+      "step": 36159
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.388457142857143e-06,
+      "loss": 1.3193,
+      "step": 36160
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3881885714285715e-06,
+      "loss": 1.2358,
+      "step": 36161
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.38792e-06,
+      "loss": 0.792,
+      "step": 36162
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.387651428571429e-06,
+      "loss": 1.0957,
+      "step": 36163
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.387382857142857e-06,
+      "loss": 1.7331,
+      "step": 36164
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.387114285714286e-06,
+      "loss": 1.4175,
+      "step": 36165
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.386845714285714e-06,
+      "loss": 1.3003,
+      "step": 36166
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.386577142857143e-06,
+      "loss": 1.0972,
+      "step": 36167
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.386308571428572e-06,
+      "loss": 1.1146,
+      "step": 36168
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.38604e-06,
+      "loss": 1.4161,
+      "step": 36169
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.385771428571429e-06,
+      "loss": 1.2093,
+      "step": 36170
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3855028571428575e-06,
+      "loss": 0.6804,
+      "step": 36171
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.385234285714286e-06,
+      "loss": 1.466,
+      "step": 36172
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.384965714285715e-06,
+      "loss": 0.9087,
+      "step": 36173
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.384697142857143e-06,
+      "loss": 1.3908,
+      "step": 36174
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.384428571428572e-06,
+      "loss": 0.8004,
+      "step": 36175
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.38416e-06,
+      "loss": 1.1634,
+      "step": 36176
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.383891428571428e-06,
+      "loss": 1.1068,
+      "step": 36177
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.383622857142857e-06,
+      "loss": 1.1617,
+      "step": 36178
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3833542857142854e-06,
+      "loss": 1.0816,
+      "step": 36179
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3830857142857144e-06,
+      "loss": 1.041,
+      "step": 36180
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3828171428571435e-06,
+      "loss": 1.3439,
+      "step": 36181
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.382548571428572e-06,
+      "loss": 0.9632,
+      "step": 36182
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.382280000000001e-06,
+      "loss": 1.1141,
+      "step": 36183
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.382011428571429e-06,
+      "loss": 1.3611,
+      "step": 36184
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.381742857142857e-06,
+      "loss": 0.8442,
+      "step": 36185
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.381474285714286e-06,
+      "loss": 0.7919,
+      "step": 36186
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.381205714285714e-06,
+      "loss": 1.2003,
+      "step": 36187
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.380937142857143e-06,
+      "loss": 1.4283,
+      "step": 36188
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.380668571428571e-06,
+      "loss": 1.341,
+      "step": 36189
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3804e-06,
+      "loss": 0.7447,
+      "step": 36190
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.380131428571429e-06,
+      "loss": 1.0256,
+      "step": 36191
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.379862857142857e-06,
+      "loss": 1.1484,
+      "step": 36192
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.379594285714286e-06,
+      "loss": 0.9932,
+      "step": 36193
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.379325714285715e-06,
+      "loss": 1.6438,
+      "step": 36194
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.379057142857143e-06,
+      "loss": 0.8743,
+      "step": 36195
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.378788571428572e-06,
+      "loss": 1.0564,
+      "step": 36196
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.37852e-06,
+      "loss": 0.943,
+      "step": 36197
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.378251428571428e-06,
+      "loss": 1.1192,
+      "step": 36198
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.377982857142857e-06,
+      "loss": 0.7005,
+      "step": 36199
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.377714285714286e-06,
+      "loss": 1.3497,
+      "step": 36200
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.377445714285715e-06,
+      "loss": 0.9643,
+      "step": 36201
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.377177142857143e-06,
+      "loss": 0.9712,
+      "step": 36202
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.376908571428571e-06,
+      "loss": 0.9359,
+      "step": 36203
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.37664e-06,
+      "loss": 1.3607,
+      "step": 36204
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.376371428571428e-06,
+      "loss": 1.1264,
+      "step": 36205
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.376102857142857e-06,
+      "loss": 1.4684,
+      "step": 36206
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.375834285714286e-06,
+      "loss": 1.2873,
+      "step": 36207
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.375565714285714e-06,
+      "loss": 1.328,
+      "step": 36208
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.375297142857143e-06,
+      "loss": 0.9798,
+      "step": 36209
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.375028571428572e-06,
+      "loss": 1.2675,
+      "step": 36210
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.374760000000001e-06,
+      "loss": 1.1944,
+      "step": 36211
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.374491428571429e-06,
+      "loss": 1.143,
+      "step": 36212
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.374222857142857e-06,
+      "loss": 1.0755,
+      "step": 36213
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.373954285714286e-06,
+      "loss": 0.8552,
+      "step": 36214
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.373685714285714e-06,
+      "loss": 0.8933,
+      "step": 36215
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.373417142857143e-06,
+      "loss": 1.4265,
+      "step": 36216
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.373148571428571e-06,
+      "loss": 1.2388,
+      "step": 36217
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3728799999999995e-06,
+      "loss": 0.9403,
+      "step": 36218
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3726114285714286e-06,
+      "loss": 0.9692,
+      "step": 36219
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.372342857142858e-06,
+      "loss": 1.369,
+      "step": 36220
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.372074285714286e-06,
+      "loss": 0.9282,
+      "step": 36221
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.371805714285715e-06,
+      "loss": 1.0605,
+      "step": 36222
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.371537142857143e-06,
+      "loss": 0.8359,
+      "step": 36223
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.371268571428572e-06,
+      "loss": 0.9746,
+      "step": 36224
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.371e-06,
+      "loss": 1.5229,
+      "step": 36225
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.370731428571428e-06,
+      "loss": 0.9488,
+      "step": 36226
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.370462857142857e-06,
+      "loss": 1.7655,
+      "step": 36227
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3701942857142855e-06,
+      "loss": 1.3901,
+      "step": 36228
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3699257142857146e-06,
+      "loss": 1.4978,
+      "step": 36229
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.369657142857143e-06,
+      "loss": 0.8672,
+      "step": 36230
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.369388571428571e-06,
+      "loss": 1.2482,
+      "step": 36231
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.36912e-06,
+      "loss": 1.0694,
+      "step": 36232
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.368851428571429e-06,
+      "loss": 1.4557,
+      "step": 36233
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.368582857142858e-06,
+      "loss": 1.1108,
+      "step": 36234
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.368314285714286e-06,
+      "loss": 1.2043,
+      "step": 36235
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.368045714285714e-06,
+      "loss": 1.0798,
+      "step": 36236
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.367777142857143e-06,
+      "loss": 0.6883,
+      "step": 36237
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3675085714285715e-06,
+      "loss": 1.0186,
+      "step": 36238
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3672400000000006e-06,
+      "loss": 1.0081,
+      "step": 36239
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.366971428571429e-06,
+      "loss": 0.8861,
+      "step": 36240
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.366702857142857e-06,
+      "loss": 0.7135,
+      "step": 36241
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.366434285714286e-06,
+      "loss": 0.9954,
+      "step": 36242
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.366165714285714e-06,
+      "loss": 1.2768,
+      "step": 36243
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.365897142857142e-06,
+      "loss": 1.332,
+      "step": 36244
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.365628571428571e-06,
+      "loss": 1.2462,
+      "step": 36245
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.36536e-06,
+      "loss": 1.2258,
+      "step": 36246
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.365091428571429e-06,
+      "loss": 1.1699,
+      "step": 36247
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3648228571428575e-06,
+      "loss": 0.8603,
+      "step": 36248
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.364554285714286e-06,
+      "loss": 1.0977,
+      "step": 36249
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.364285714285715e-06,
+      "loss": 1.0262,
+      "step": 36250
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.364017142857143e-06,
+      "loss": 1.0767,
+      "step": 36251
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.363748571428572e-06,
+      "loss": 0.8728,
+      "step": 36252
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.36348e-06,
+      "loss": 0.8529,
+      "step": 36253
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.363211428571428e-06,
+      "loss": 0.5638,
+      "step": 36254
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.362942857142857e-06,
+      "loss": 1.1238,
+      "step": 36255
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3626742857142855e-06,
+      "loss": 1.277,
+      "step": 36256
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3624057142857145e-06,
+      "loss": 0.8313,
+      "step": 36257
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.362137142857143e-06,
+      "loss": 1.2677,
+      "step": 36258
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.361868571428572e-06,
+      "loss": 1.105,
+      "step": 36259
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.361600000000001e-06,
+      "loss": 1.1842,
+      "step": 36260
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.361331428571429e-06,
+      "loss": 1.1371,
+      "step": 36261
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.361062857142857e-06,
+      "loss": 0.9747,
+      "step": 36262
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.360794285714286e-06,
+      "loss": 1.0262,
+      "step": 36263
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.360525714285714e-06,
+      "loss": 0.8992,
+      "step": 36264
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.360257142857143e-06,
+      "loss": 1.5414,
+      "step": 36265
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3599885714285715e-06,
+      "loss": 1.5406,
+      "step": 36266
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.35972e-06,
+      "loss": 0.9451,
+      "step": 36267
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.359451428571429e-06,
+      "loss": 1.2534,
+      "step": 36268
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.359182857142857e-06,
+      "loss": 1.0498,
+      "step": 36269
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.358914285714286e-06,
+      "loss": 1.1475,
+      "step": 36270
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.358645714285714e-06,
+      "loss": 1.1437,
+      "step": 36271
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.358377142857143e-06,
+      "loss": 1.4241,
+      "step": 36272
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.358108571428572e-06,
+      "loss": 1.0126,
+      "step": 36273
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.35784e-06,
+      "loss": 0.6674,
+      "step": 36274
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.357571428571429e-06,
+      "loss": 1.0156,
+      "step": 36275
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3573028571428575e-06,
+      "loss": 0.9964,
+      "step": 36276
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.357034285714286e-06,
+      "loss": 1.3224,
+      "step": 36277
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.356765714285715e-06,
+      "loss": 1.1806,
+      "step": 36278
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.356497142857143e-06,
+      "loss": 1.3289,
+      "step": 36279
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.356228571428572e-06,
+      "loss": 1.1902,
+      "step": 36280
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.35596e-06,
+      "loss": 1.7851,
+      "step": 36281
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.355691428571428e-06,
+      "loss": 0.9113,
+      "step": 36282
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.355422857142857e-06,
+      "loss": 1.2039,
+      "step": 36283
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3551542857142854e-06,
+      "loss": 1.3091,
+      "step": 36284
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3548857142857144e-06,
+      "loss": 1.2084,
+      "step": 36285
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3546171428571435e-06,
+      "loss": 1.1719,
+      "step": 36286
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.354348571428572e-06,
+      "loss": 1.1199,
+      "step": 36287
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.354080000000001e-06,
+      "loss": 1.2211,
+      "step": 36288
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.353811428571429e-06,
+      "loss": 1.4237,
+      "step": 36289
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.353542857142857e-06,
+      "loss": 1.3553,
+      "step": 36290
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.353274285714286e-06,
+      "loss": 1.272,
+      "step": 36291
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.353005714285714e-06,
+      "loss": 1.1308,
+      "step": 36292
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.352737142857143e-06,
+      "loss": 1.1509,
+      "step": 36293
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.352468571428571e-06,
+      "loss": 1.4335,
+      "step": 36294
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3522e-06,
+      "loss": 1.0506,
+      "step": 36295
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.351931428571429e-06,
+      "loss": 0.5961,
+      "step": 36296
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.351662857142857e-06,
+      "loss": 1.0577,
+      "step": 36297
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.351394285714286e-06,
+      "loss": 1.2597,
+      "step": 36298
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.351125714285715e-06,
+      "loss": 1.0346,
+      "step": 36299
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.350857142857143e-06,
+      "loss": 1.157,
+      "step": 36300
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.350588571428572e-06,
+      "loss": 0.8491,
+      "step": 36301
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.35032e-06,
+      "loss": 1.0928,
+      "step": 36302
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.350051428571429e-06,
+      "loss": 1.2229,
+      "step": 36303
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.349782857142857e-06,
+      "loss": 1.4512,
+      "step": 36304
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.349514285714286e-06,
+      "loss": 1.0512,
+      "step": 36305
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.349245714285715e-06,
+      "loss": 1.5416,
+      "step": 36306
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.348977142857143e-06,
+      "loss": 0.8392,
+      "step": 36307
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.348708571428571e-06,
+      "loss": 1.1527,
+      "step": 36308
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.34844e-06,
+      "loss": 1.0717,
+      "step": 36309
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.348171428571428e-06,
+      "loss": 1.4687,
+      "step": 36310
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.347902857142857e-06,
+      "loss": 1.1846,
+      "step": 36311
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.347634285714286e-06,
+      "loss": 1.3355,
+      "step": 36312
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.347365714285714e-06,
+      "loss": 0.8261,
+      "step": 36313
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.347097142857143e-06,
+      "loss": 1.1052,
+      "step": 36314
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.346828571428572e-06,
+      "loss": 1.3214,
+      "step": 36315
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.346560000000001e-06,
+      "loss": 1.2496,
+      "step": 36316
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.346291428571429e-06,
+      "loss": 1.4162,
+      "step": 36317
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.346022857142857e-06,
+      "loss": 0.9641,
+      "step": 36318
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.345754285714286e-06,
+      "loss": 1.1518,
+      "step": 36319
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.345485714285714e-06,
+      "loss": 1.6004,
+      "step": 36320
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.345217142857143e-06,
+      "loss": 0.7506,
+      "step": 36321
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.344948571428571e-06,
+      "loss": 1.2098,
+      "step": 36322
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3446799999999995e-06,
+      "loss": 1.0279,
+      "step": 36323
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3444114285714286e-06,
+      "loss": 1.4111,
+      "step": 36324
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.344142857142858e-06,
+      "loss": 1.0688,
+      "step": 36325
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.343874285714287e-06,
+      "loss": 1.465,
+      "step": 36326
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.343605714285715e-06,
+      "loss": 0.7453,
+      "step": 36327
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.343337142857143e-06,
+      "loss": 0.9297,
+      "step": 36328
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.343068571428572e-06,
+      "loss": 1.0527,
+      "step": 36329
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3428e-06,
+      "loss": 0.918,
+      "step": 36330
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.342531428571428e-06,
+      "loss": 0.9891,
+      "step": 36331
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.342262857142857e-06,
+      "loss": 0.9155,
+      "step": 36332
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3419942857142855e-06,
+      "loss": 1.0211,
+      "step": 36333
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3417257142857146e-06,
+      "loss": 0.9819,
+      "step": 36334
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.341457142857143e-06,
+      "loss": 0.8109,
+      "step": 36335
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.341188571428571e-06,
+      "loss": 1.0845,
+      "step": 36336
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.34092e-06,
+      "loss": 1.2269,
+      "step": 36337
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.340651428571429e-06,
+      "loss": 1.2311,
+      "step": 36338
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.340382857142858e-06,
+      "loss": 1.0766,
+      "step": 36339
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.340114285714286e-06,
+      "loss": 0.7053,
+      "step": 36340
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.339845714285714e-06,
+      "loss": 0.9306,
+      "step": 36341
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.339577142857143e-06,
+      "loss": 1.5262,
+      "step": 36342
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3393085714285715e-06,
+      "loss": 1.3543,
+      "step": 36343
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3390400000000006e-06,
+      "loss": 1.0955,
+      "step": 36344
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.338771428571429e-06,
+      "loss": 0.9865,
+      "step": 36345
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.338502857142857e-06,
+      "loss": 1.3261,
+      "step": 36346
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.338234285714286e-06,
+      "loss": 1.4433,
+      "step": 36347
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.337965714285714e-06,
+      "loss": 1.0145,
+      "step": 36348
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.337697142857142e-06,
+      "loss": 1.1595,
+      "step": 36349
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.337428571428571e-06,
+      "loss": 1.2608,
+      "step": 36350
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.33716e-06,
+      "loss": 1.019,
+      "step": 36351
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.336891428571429e-06,
+      "loss": 1.0742,
+      "step": 36352
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3366228571428575e-06,
+      "loss": 1.1297,
+      "step": 36353
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.336354285714286e-06,
+      "loss": 1.7599,
+      "step": 36354
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.336085714285715e-06,
+      "loss": 0.9791,
+      "step": 36355
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.335817142857143e-06,
+      "loss": 0.8333,
+      "step": 36356
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.335548571428572e-06,
+      "loss": 0.9917,
+      "step": 36357
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.33528e-06,
+      "loss": 1.5939,
+      "step": 36358
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.335011428571428e-06,
+      "loss": 0.7994,
+      "step": 36359
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.334742857142857e-06,
+      "loss": 1.5423,
+      "step": 36360
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3344742857142855e-06,
+      "loss": 1.5271,
+      "step": 36361
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3342057142857145e-06,
+      "loss": 1.3308,
+      "step": 36362
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.333937142857143e-06,
+      "loss": 0.7981,
+      "step": 36363
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.333668571428572e-06,
+      "loss": 0.6952,
+      "step": 36364
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.333400000000001e-06,
+      "loss": 1.1187,
+      "step": 36365
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.333131428571429e-06,
+      "loss": 1.0817,
+      "step": 36366
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.332862857142858e-06,
+      "loss": 1.3219,
+      "step": 36367
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.332594285714286e-06,
+      "loss": 1.2362,
+      "step": 36368
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.332325714285714e-06,
+      "loss": 0.8145,
+      "step": 36369
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.332057142857143e-06,
+      "loss": 1.0584,
+      "step": 36370
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3317885714285715e-06,
+      "loss": 0.8447,
+      "step": 36371
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.33152e-06,
+      "loss": 1.1384,
+      "step": 36372
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.331251428571429e-06,
+      "loss": 1.1675,
+      "step": 36373
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.330982857142857e-06,
+      "loss": 1.0431,
+      "step": 36374
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.330714285714286e-06,
+      "loss": 1.2175,
+      "step": 36375
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.330445714285714e-06,
+      "loss": 0.9537,
+      "step": 36376
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.330177142857143e-06,
+      "loss": 1.0858,
+      "step": 36377
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.329908571428572e-06,
+      "loss": 0.8983,
+      "step": 36378
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.32964e-06,
+      "loss": 0.9279,
+      "step": 36379
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.329371428571429e-06,
+      "loss": 1.4484,
+      "step": 36380
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3291028571428575e-06,
+      "loss": 1.1824,
+      "step": 36381
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.328834285714286e-06,
+      "loss": 1.0178,
+      "step": 36382
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.328565714285715e-06,
+      "loss": 1.249,
+      "step": 36383
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.328297142857143e-06,
+      "loss": 1.1755,
+      "step": 36384
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.328028571428572e-06,
+      "loss": 1.1602,
+      "step": 36385
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.32776e-06,
+      "loss": 0.8347,
+      "step": 36386
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.327491428571428e-06,
+      "loss": 1.0314,
+      "step": 36387
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.327222857142857e-06,
+      "loss": 0.8762,
+      "step": 36388
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3269542857142854e-06,
+      "loss": 0.7168,
+      "step": 36389
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3266857142857144e-06,
+      "loss": 0.6963,
+      "step": 36390
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3264171428571435e-06,
+      "loss": 0.8498,
+      "step": 36391
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.326148571428572e-06,
+      "loss": 0.8821,
+      "step": 36392
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.325880000000001e-06,
+      "loss": 0.9416,
+      "step": 36393
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.325611428571429e-06,
+      "loss": 1.3365,
+      "step": 36394
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.325342857142857e-06,
+      "loss": 1.2492,
+      "step": 36395
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.325074285714286e-06,
+      "loss": 0.8919,
+      "step": 36396
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.324805714285714e-06,
+      "loss": 1.1374,
+      "step": 36397
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.324537142857143e-06,
+      "loss": 0.8459,
+      "step": 36398
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.324268571428571e-06,
+      "loss": 1.1546,
+      "step": 36399
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.324e-06,
+      "loss": 1.3363,
+      "step": 36400
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.323731428571429e-06,
+      "loss": 0.8098,
+      "step": 36401
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.323462857142857e-06,
+      "loss": 0.9569,
+      "step": 36402
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.323194285714286e-06,
+      "loss": 1.0498,
+      "step": 36403
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.322925714285715e-06,
+      "loss": 1.2398,
+      "step": 36404
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.322657142857143e-06,
+      "loss": 1.2231,
+      "step": 36405
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.322388571428572e-06,
+      "loss": 0.9289,
+      "step": 36406
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.32212e-06,
+      "loss": 1.044,
+      "step": 36407
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.321851428571429e-06,
+      "loss": 1.0814,
+      "step": 36408
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.321582857142857e-06,
+      "loss": 1.2239,
+      "step": 36409
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.321314285714286e-06,
+      "loss": 1.2068,
+      "step": 36410
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.321045714285715e-06,
+      "loss": 1.6988,
+      "step": 36411
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.320777142857143e-06,
+      "loss": 1.1756,
+      "step": 36412
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.320508571428572e-06,
+      "loss": 0.8931,
+      "step": 36413
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.32024e-06,
+      "loss": 1.2243,
+      "step": 36414
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.319971428571428e-06,
+      "loss": 1.0448,
+      "step": 36415
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.319702857142857e-06,
+      "loss": 1.3241,
+      "step": 36416
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.319434285714286e-06,
+      "loss": 1.0694,
+      "step": 36417
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.319165714285714e-06,
+      "loss": 1.0304,
+      "step": 36418
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.318897142857143e-06,
+      "loss": 1.5268,
+      "step": 36419
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.318628571428572e-06,
+      "loss": 1.2707,
+      "step": 36420
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.318360000000001e-06,
+      "loss": 1.0376,
+      "step": 36421
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.318091428571429e-06,
+      "loss": 1.2801,
+      "step": 36422
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.317822857142857e-06,
+      "loss": 1.2089,
+      "step": 36423
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.317554285714286e-06,
+      "loss": 1.113,
+      "step": 36424
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.317285714285714e-06,
+      "loss": 1.1107,
+      "step": 36425
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.317017142857143e-06,
+      "loss": 1.3755,
+      "step": 36426
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.316748571428571e-06,
+      "loss": 1.031,
+      "step": 36427
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3164799999999995e-06,
+      "loss": 1.0945,
+      "step": 36428
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3162114285714286e-06,
+      "loss": 1.1113,
+      "step": 36429
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.315942857142858e-06,
+      "loss": 1.5871,
+      "step": 36430
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.315674285714287e-06,
+      "loss": 0.8434,
+      "step": 36431
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.315405714285715e-06,
+      "loss": 1.5309,
+      "step": 36432
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.315137142857143e-06,
+      "loss": 0.6749,
+      "step": 36433
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.314868571428572e-06,
+      "loss": 1.1627,
+      "step": 36434
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3146e-06,
+      "loss": 1.2307,
+      "step": 36435
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.314331428571428e-06,
+      "loss": 0.8584,
+      "step": 36436
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.314062857142857e-06,
+      "loss": 0.8899,
+      "step": 36437
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3137942857142855e-06,
+      "loss": 1.4572,
+      "step": 36438
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3135257142857146e-06,
+      "loss": 1.3625,
+      "step": 36439
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.313257142857143e-06,
+      "loss": 1.3408,
+      "step": 36440
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.312988571428571e-06,
+      "loss": 1.2402,
+      "step": 36441
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.31272e-06,
+      "loss": 0.6326,
+      "step": 36442
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.312451428571429e-06,
+      "loss": 1.4191,
+      "step": 36443
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.312182857142858e-06,
+      "loss": 1.1859,
+      "step": 36444
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.311914285714286e-06,
+      "loss": 0.8803,
+      "step": 36445
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.311645714285714e-06,
+      "loss": 1.0767,
+      "step": 36446
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.311377142857143e-06,
+      "loss": 1.4034,
+      "step": 36447
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3111085714285715e-06,
+      "loss": 0.7197,
+      "step": 36448
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3108400000000006e-06,
+      "loss": 1.4636,
+      "step": 36449
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.310571428571429e-06,
+      "loss": 0.6309,
+      "step": 36450
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.310302857142857e-06,
+      "loss": 0.8614,
+      "step": 36451
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.310034285714286e-06,
+      "loss": 1.1057,
+      "step": 36452
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.309765714285714e-06,
+      "loss": 1.1513,
+      "step": 36453
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.309497142857143e-06,
+      "loss": 0.9606,
+      "step": 36454
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.309228571428571e-06,
+      "loss": 1.0429,
+      "step": 36455
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.30896e-06,
+      "loss": 0.9209,
+      "step": 36456
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.308691428571429e-06,
+      "loss": 1.0418,
+      "step": 36457
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3084228571428575e-06,
+      "loss": 0.7679,
+      "step": 36458
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.308154285714286e-06,
+      "loss": 1.3556,
+      "step": 36459
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.307885714285715e-06,
+      "loss": 1.1677,
+      "step": 36460
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.307617142857143e-06,
+      "loss": 1.1106,
+      "step": 36461
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.307348571428572e-06,
+      "loss": 1.5205,
+      "step": 36462
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.30708e-06,
+      "loss": 1.3286,
+      "step": 36463
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.306811428571428e-06,
+      "loss": 1.0074,
+      "step": 36464
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.306542857142857e-06,
+      "loss": 1.1398,
+      "step": 36465
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3062742857142855e-06,
+      "loss": 0.9969,
+      "step": 36466
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3060057142857145e-06,
+      "loss": 0.8952,
+      "step": 36467
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.305737142857143e-06,
+      "loss": 0.7575,
+      "step": 36468
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.305468571428572e-06,
+      "loss": 0.9046,
+      "step": 36469
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.305200000000001e-06,
+      "loss": 1.3343,
+      "step": 36470
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.304931428571429e-06,
+      "loss": 1.2979,
+      "step": 36471
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.304662857142858e-06,
+      "loss": 0.7342,
+      "step": 36472
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.304394285714286e-06,
+      "loss": 1.1914,
+      "step": 36473
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.304125714285714e-06,
+      "loss": 1.3052,
+      "step": 36474
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.303857142857143e-06,
+      "loss": 0.9526,
+      "step": 36475
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3035885714285715e-06,
+      "loss": 1.2203,
+      "step": 36476
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3033200000000005e-06,
+      "loss": 1.3191,
+      "step": 36477
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.303051428571429e-06,
+      "loss": 0.7101,
+      "step": 36478
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.302782857142857e-06,
+      "loss": 1.1875,
+      "step": 36479
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.302514285714286e-06,
+      "loss": 0.9044,
+      "step": 36480
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.302245714285714e-06,
+      "loss": 0.9946,
+      "step": 36481
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.301977142857142e-06,
+      "loss": 0.9675,
+      "step": 36482
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.301708571428572e-06,
+      "loss": 1.4058,
+      "step": 36483
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.30144e-06,
+      "loss": 1.0595,
+      "step": 36484
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.301171428571429e-06,
+      "loss": 0.7642,
+      "step": 36485
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3009028571428575e-06,
+      "loss": 0.7366,
+      "step": 36486
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.300634285714286e-06,
+      "loss": 0.8082,
+      "step": 36487
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.300365714285715e-06,
+      "loss": 0.8334,
+      "step": 36488
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.300097142857143e-06,
+      "loss": 0.9167,
+      "step": 36489
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.299828571428572e-06,
+      "loss": 0.9245,
+      "step": 36490
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.29956e-06,
+      "loss": 0.7298,
+      "step": 36491
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.299291428571428e-06,
+      "loss": 0.665,
+      "step": 36492
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.299022857142857e-06,
+      "loss": 0.8335,
+      "step": 36493
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.298754285714285e-06,
+      "loss": 1.6485,
+      "step": 36494
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2984857142857144e-06,
+      "loss": 0.8592,
+      "step": 36495
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2982171428571435e-06,
+      "loss": 1.1343,
+      "step": 36496
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.297948571428572e-06,
+      "loss": 0.8446,
+      "step": 36497
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.297680000000001e-06,
+      "loss": 1.308,
+      "step": 36498
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.297411428571429e-06,
+      "loss": 1.7399,
+      "step": 36499
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.297142857142858e-06,
+      "loss": 0.8153,
+      "step": 36500
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.296874285714286e-06,
+      "loss": 1.1653,
+      "step": 36501
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.296605714285714e-06,
+      "loss": 1.3141,
+      "step": 36502
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.296337142857143e-06,
+      "loss": 0.9632,
+      "step": 36503
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.296068571428571e-06,
+      "loss": 0.9098,
+      "step": 36504
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2958e-06,
+      "loss": 1.1461,
+      "step": 36505
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.295531428571429e-06,
+      "loss": 1.2734,
+      "step": 36506
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.295262857142857e-06,
+      "loss": 1.2039,
+      "step": 36507
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.294994285714286e-06,
+      "loss": 1.4502,
+      "step": 36508
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.294725714285715e-06,
+      "loss": 1.2489,
+      "step": 36509
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.294457142857143e-06,
+      "loss": 1.1616,
+      "step": 36510
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.294188571428572e-06,
+      "loss": 1.1713,
+      "step": 36511
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.29392e-06,
+      "loss": 0.9278,
+      "step": 36512
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.293651428571429e-06,
+      "loss": 1.2739,
+      "step": 36513
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.293382857142857e-06,
+      "loss": 1.1312,
+      "step": 36514
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.293114285714286e-06,
+      "loss": 0.8346,
+      "step": 36515
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.292845714285715e-06,
+      "loss": 1.0266,
+      "step": 36516
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.292577142857143e-06,
+      "loss": 0.982,
+      "step": 36517
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.292308571428572e-06,
+      "loss": 1.2026,
+      "step": 36518
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.29204e-06,
+      "loss": 0.9952,
+      "step": 36519
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.291771428571428e-06,
+      "loss": 1.2888,
+      "step": 36520
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.291502857142857e-06,
+      "loss": 1.0992,
+      "step": 36521
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.291234285714286e-06,
+      "loss": 1.3275,
+      "step": 36522
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.290965714285714e-06,
+      "loss": 0.9919,
+      "step": 36523
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.290697142857143e-06,
+      "loss": 1.1489,
+      "step": 36524
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.290428571428572e-06,
+      "loss": 1.3815,
+      "step": 36525
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.290160000000001e-06,
+      "loss": 1.1917,
+      "step": 36526
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.289891428571429e-06,
+      "loss": 1.2978,
+      "step": 36527
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.289622857142857e-06,
+      "loss": 1.1847,
+      "step": 36528
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.289354285714286e-06,
+      "loss": 0.9681,
+      "step": 36529
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.289085714285714e-06,
+      "loss": 1.0819,
+      "step": 36530
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.288817142857143e-06,
+      "loss": 1.1443,
+      "step": 36531
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.288548571428571e-06,
+      "loss": 1.2743,
+      "step": 36532
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2882799999999995e-06,
+      "loss": 1.2603,
+      "step": 36533
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2880114285714286e-06,
+      "loss": 1.1069,
+      "step": 36534
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.287742857142857e-06,
+      "loss": 1.2412,
+      "step": 36535
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.287474285714287e-06,
+      "loss": 1.3226,
+      "step": 36536
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.287205714285715e-06,
+      "loss": 1.475,
+      "step": 36537
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.286937142857143e-06,
+      "loss": 1.3826,
+      "step": 36538
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.286668571428572e-06,
+      "loss": 1.2474,
+      "step": 36539
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2864e-06,
+      "loss": 0.8027,
+      "step": 36540
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.286131428571429e-06,
+      "loss": 0.9977,
+      "step": 36541
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.285862857142857e-06,
+      "loss": 1.17,
+      "step": 36542
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2855942857142855e-06,
+      "loss": 1.0742,
+      "step": 36543
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2853257142857146e-06,
+      "loss": 1.1641,
+      "step": 36544
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.285057142857143e-06,
+      "loss": 0.9687,
+      "step": 36545
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.284788571428571e-06,
+      "loss": 1.3097,
+      "step": 36546
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.28452e-06,
+      "loss": 1.2726,
+      "step": 36547
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.284251428571428e-06,
+      "loss": 1.0892,
+      "step": 36548
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.283982857142858e-06,
+      "loss": 0.9701,
+      "step": 36549
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.283714285714286e-06,
+      "loss": 1.3301,
+      "step": 36550
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.283445714285714e-06,
+      "loss": 1.0679,
+      "step": 36551
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.283177142857143e-06,
+      "loss": 0.9861,
+      "step": 36552
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2829085714285715e-06,
+      "loss": 0.7946,
+      "step": 36553
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2826400000000006e-06,
+      "loss": 1.3397,
+      "step": 36554
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.282371428571429e-06,
+      "loss": 0.9856,
+      "step": 36555
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.282102857142857e-06,
+      "loss": 1.2117,
+      "step": 36556
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.281834285714286e-06,
+      "loss": 1.3945,
+      "step": 36557
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.281565714285714e-06,
+      "loss": 1.4097,
+      "step": 36558
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.281297142857143e-06,
+      "loss": 1.0321,
+      "step": 36559
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.281028571428571e-06,
+      "loss": 1.5123,
+      "step": 36560
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2807599999999995e-06,
+      "loss": 1.0914,
+      "step": 36561
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.280491428571429e-06,
+      "loss": 1.2627,
+      "step": 36562
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2802228571428575e-06,
+      "loss": 1.2525,
+      "step": 36563
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2799542857142866e-06,
+      "loss": 1.562,
+      "step": 36564
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.279685714285715e-06,
+      "loss": 1.3651,
+      "step": 36565
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.279417142857143e-06,
+      "loss": 1.2226,
+      "step": 36566
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.279148571428572e-06,
+      "loss": 1.2116,
+      "step": 36567
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.27888e-06,
+      "loss": 1.7269,
+      "step": 36568
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.278611428571428e-06,
+      "loss": 1.0699,
+      "step": 36569
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.278342857142857e-06,
+      "loss": 1.0952,
+      "step": 36570
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2780742857142855e-06,
+      "loss": 0.8751,
+      "step": 36571
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2778057142857145e-06,
+      "loss": 0.8803,
+      "step": 36572
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.277537142857143e-06,
+      "loss": 1.2747,
+      "step": 36573
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.277268571428571e-06,
+      "loss": 1.4778,
+      "step": 36574
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.277000000000001e-06,
+      "loss": 1.0764,
+      "step": 36575
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.276731428571429e-06,
+      "loss": 0.9049,
+      "step": 36576
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.276462857142858e-06,
+      "loss": 1.257,
+      "step": 36577
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.276194285714286e-06,
+      "loss": 1.2267,
+      "step": 36578
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.275925714285714e-06,
+      "loss": 0.8014,
+      "step": 36579
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.275657142857143e-06,
+      "loss": 1.3432,
+      "step": 36580
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2753885714285715e-06,
+      "loss": 1.4624,
+      "step": 36581
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2751200000000005e-06,
+      "loss": 0.9343,
+      "step": 36582
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.274851428571429e-06,
+      "loss": 1.134,
+      "step": 36583
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.274582857142857e-06,
+      "loss": 1.3804,
+      "step": 36584
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.274314285714286e-06,
+      "loss": 1.2524,
+      "step": 36585
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.274045714285714e-06,
+      "loss": 0.9096,
+      "step": 36586
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.273777142857143e-06,
+      "loss": 1.5336,
+      "step": 36587
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.273508571428571e-06,
+      "loss": 1.0521,
+      "step": 36588
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.27324e-06,
+      "loss": 1.5659,
+      "step": 36589
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.272971428571429e-06,
+      "loss": 1.1374,
+      "step": 36590
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2727028571428575e-06,
+      "loss": 1.1819,
+      "step": 36591
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.272434285714286e-06,
+      "loss": 1.3242,
+      "step": 36592
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.272165714285715e-06,
+      "loss": 1.0211,
+      "step": 36593
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.271897142857143e-06,
+      "loss": 1.1498,
+      "step": 36594
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.271628571428572e-06,
+      "loss": 0.7523,
+      "step": 36595
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.27136e-06,
+      "loss": 1.3719,
+      "step": 36596
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.271091428571428e-06,
+      "loss": 1.4077,
+      "step": 36597
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.270822857142857e-06,
+      "loss": 0.8776,
+      "step": 36598
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.270554285714285e-06,
+      "loss": 1.3705,
+      "step": 36599
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2702857142857144e-06,
+      "loss": 1.29,
+      "step": 36600
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.270017142857143e-06,
+      "loss": 1.1731,
+      "step": 36601
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.269748571428572e-06,
+      "loss": 0.8721,
+      "step": 36602
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.269480000000001e-06,
+      "loss": 1.2822,
+      "step": 36603
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.269211428571429e-06,
+      "loss": 0.9858,
+      "step": 36604
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.268942857142858e-06,
+      "loss": 1.0784,
+      "step": 36605
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.268674285714286e-06,
+      "loss": 1.4167,
+      "step": 36606
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.268405714285714e-06,
+      "loss": 1.0697,
+      "step": 36607
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.268137142857143e-06,
+      "loss": 1.109,
+      "step": 36608
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.267868571428571e-06,
+      "loss": 0.9842,
+      "step": 36609
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2676e-06,
+      "loss": 1.2964,
+      "step": 36610
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.267331428571429e-06,
+      "loss": 1.4957,
+      "step": 36611
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.267062857142857e-06,
+      "loss": 1.2947,
+      "step": 36612
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.266794285714286e-06,
+      "loss": 1.3335,
+      "step": 36613
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.266525714285714e-06,
+      "loss": 1.1536,
+      "step": 36614
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.266257142857143e-06,
+      "loss": 1.1101,
+      "step": 36615
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.265988571428572e-06,
+      "loss": 0.825,
+      "step": 36616
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.26572e-06,
+      "loss": 1.5132,
+      "step": 36617
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.265451428571429e-06,
+      "loss": 1.0045,
+      "step": 36618
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.265182857142857e-06,
+      "loss": 0.9829,
+      "step": 36619
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.264914285714286e-06,
+      "loss": 1.3384,
+      "step": 36620
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.264645714285715e-06,
+      "loss": 1.0977,
+      "step": 36621
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.264377142857143e-06,
+      "loss": 1.1945,
+      "step": 36622
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.264108571428572e-06,
+      "loss": 0.9237,
+      "step": 36623
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.26384e-06,
+      "loss": 0.8281,
+      "step": 36624
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.263571428571428e-06,
+      "loss": 0.9162,
+      "step": 36625
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.263302857142857e-06,
+      "loss": 0.6816,
+      "step": 36626
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.263034285714285e-06,
+      "loss": 1.1978,
+      "step": 36627
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.262765714285715e-06,
+      "loss": 1.1998,
+      "step": 36628
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.262497142857143e-06,
+      "loss": 1.0498,
+      "step": 36629
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.262228571428572e-06,
+      "loss": 1.3276,
+      "step": 36630
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.261960000000001e-06,
+      "loss": 1.1059,
+      "step": 36631
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.261691428571429e-06,
+      "loss": 1.2434,
+      "step": 36632
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.261422857142857e-06,
+      "loss": 1.0246,
+      "step": 36633
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.261154285714286e-06,
+      "loss": 1.079,
+      "step": 36634
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.260885714285714e-06,
+      "loss": 1.4324,
+      "step": 36635
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.260617142857143e-06,
+      "loss": 1.3215,
+      "step": 36636
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.260348571428571e-06,
+      "loss": 1.0783,
+      "step": 36637
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2600799999999995e-06,
+      "loss": 0.5677,
+      "step": 36638
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2598114285714286e-06,
+      "loss": 1.31,
+      "step": 36639
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.259542857142857e-06,
+      "loss": 0.8254,
+      "step": 36640
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.259274285714286e-06,
+      "loss": 1.1865,
+      "step": 36641
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.259005714285715e-06,
+      "loss": 1.0511,
+      "step": 36642
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.258737142857143e-06,
+      "loss": 1.268,
+      "step": 36643
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.258468571428572e-06,
+      "loss": 0.7372,
+      "step": 36644
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2582e-06,
+      "loss": 1.064,
+      "step": 36645
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.257931428571429e-06,
+      "loss": 1.2888,
+      "step": 36646
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.257662857142857e-06,
+      "loss": 0.984,
+      "step": 36647
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2573942857142855e-06,
+      "loss": 0.8133,
+      "step": 36648
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2571257142857146e-06,
+      "loss": 0.9635,
+      "step": 36649
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.256857142857143e-06,
+      "loss": 1.0679,
+      "step": 36650
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.256588571428572e-06,
+      "loss": 1.1257,
+      "step": 36651
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.25632e-06,
+      "loss": 0.823,
+      "step": 36652
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.256051428571428e-06,
+      "loss": 1.2917,
+      "step": 36653
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.255782857142857e-06,
+      "loss": 1.0179,
+      "step": 36654
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.255514285714286e-06,
+      "loss": 1.3173,
+      "step": 36655
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.255245714285714e-06,
+      "loss": 0.9594,
+      "step": 36656
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.254977142857143e-06,
+      "loss": 1.4095,
+      "step": 36657
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2547085714285715e-06,
+      "loss": 1.2098,
+      "step": 36658
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2544400000000006e-06,
+      "loss": 0.9954,
+      "step": 36659
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.254171428571429e-06,
+      "loss": 0.6918,
+      "step": 36660
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.253902857142857e-06,
+      "loss": 1.2725,
+      "step": 36661
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.253634285714286e-06,
+      "loss": 1.2281,
+      "step": 36662
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.253365714285714e-06,
+      "loss": 1.2088,
+      "step": 36663
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.253097142857143e-06,
+      "loss": 1.2643,
+      "step": 36664
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.252828571428571e-06,
+      "loss": 0.7292,
+      "step": 36665
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2525599999999995e-06,
+      "loss": 1.2588,
+      "step": 36666
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2522914285714285e-06,
+      "loss": 1.1921,
+      "step": 36667
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2520228571428575e-06,
+      "loss": 1.459,
+      "step": 36668
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2517542857142866e-06,
+      "loss": 1.1072,
+      "step": 36669
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.251485714285715e-06,
+      "loss": 1.1453,
+      "step": 36670
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.251217142857143e-06,
+      "loss": 0.912,
+      "step": 36671
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.250948571428572e-06,
+      "loss": 1.1831,
+      "step": 36672
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.25068e-06,
+      "loss": 1.2366,
+      "step": 36673
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.250411428571429e-06,
+      "loss": 1.192,
+      "step": 36674
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.250142857142857e-06,
+      "loss": 1.0961,
+      "step": 36675
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2498742857142855e-06,
+      "loss": 0.9626,
+      "step": 36676
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2496057142857145e-06,
+      "loss": 0.9194,
+      "step": 36677
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.249337142857143e-06,
+      "loss": 0.9525,
+      "step": 36678
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.249068571428571e-06,
+      "loss": 1.119,
+      "step": 36679
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2488e-06,
+      "loss": 0.9045,
+      "step": 36680
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.248531428571429e-06,
+      "loss": 1.2123,
+      "step": 36681
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.248262857142858e-06,
+      "loss": 1.1239,
+      "step": 36682
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.247994285714286e-06,
+      "loss": 0.9374,
+      "step": 36683
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.247725714285714e-06,
+      "loss": 1.2662,
+      "step": 36684
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.247457142857143e-06,
+      "loss": 1.5845,
+      "step": 36685
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2471885714285715e-06,
+      "loss": 1.0384,
+      "step": 36686
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2469200000000005e-06,
+      "loss": 0.7788,
+      "step": 36687
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.246651428571429e-06,
+      "loss": 0.7593,
+      "step": 36688
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.246382857142857e-06,
+      "loss": 1.06,
+      "step": 36689
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.246114285714286e-06,
+      "loss": 1.3894,
+      "step": 36690
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.245845714285714e-06,
+      "loss": 0.5786,
+      "step": 36691
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.245577142857143e-06,
+      "loss": 1.3822,
+      "step": 36692
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.245308571428571e-06,
+      "loss": 1.1626,
+      "step": 36693
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.24504e-06,
+      "loss": 0.562,
+      "step": 36694
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.244771428571429e-06,
+      "loss": 1.1624,
+      "step": 36695
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2445028571428575e-06,
+      "loss": 0.8429,
+      "step": 36696
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.244234285714286e-06,
+      "loss": 1.1836,
+      "step": 36697
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.243965714285715e-06,
+      "loss": 1.0688,
+      "step": 36698
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.243697142857143e-06,
+      "loss": 1.3425,
+      "step": 36699
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.243428571428572e-06,
+      "loss": 1.0381,
+      "step": 36700
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.24316e-06,
+      "loss": 0.8671,
+      "step": 36701
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.242891428571428e-06,
+      "loss": 1.2268,
+      "step": 36702
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.242622857142857e-06,
+      "loss": 0.9352,
+      "step": 36703
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.242354285714285e-06,
+      "loss": 1.0728,
+      "step": 36704
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2420857142857144e-06,
+      "loss": 0.9106,
+      "step": 36705
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.241817142857143e-06,
+      "loss": 1.531,
+      "step": 36706
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.241548571428572e-06,
+      "loss": 1.0149,
+      "step": 36707
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.241280000000001e-06,
+      "loss": 1.4266,
+      "step": 36708
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.241011428571429e-06,
+      "loss": 1.1,
+      "step": 36709
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.240742857142858e-06,
+      "loss": 0.8199,
+      "step": 36710
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.240474285714286e-06,
+      "loss": 1.3339,
+      "step": 36711
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.240205714285714e-06,
+      "loss": 1.2955,
+      "step": 36712
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.239937142857143e-06,
+      "loss": 1.0007,
+      "step": 36713
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.239668571428571e-06,
+      "loss": 0.8869,
+      "step": 36714
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2394000000000004e-06,
+      "loss": 1.3989,
+      "step": 36715
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.239131428571429e-06,
+      "loss": 1.1963,
+      "step": 36716
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.238862857142857e-06,
+      "loss": 0.8949,
+      "step": 36717
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.238594285714286e-06,
+      "loss": 0.6901,
+      "step": 36718
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.238325714285714e-06,
+      "loss": 0.7685,
+      "step": 36719
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.238057142857143e-06,
+      "loss": 1.375,
+      "step": 36720
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.237788571428572e-06,
+      "loss": 1.5114,
+      "step": 36721
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.23752e-06,
+      "loss": 1.6507,
+      "step": 36722
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.237251428571429e-06,
+      "loss": 0.5871,
+      "step": 36723
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.236982857142857e-06,
+      "loss": 0.7983,
+      "step": 36724
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.236714285714286e-06,
+      "loss": 1.359,
+      "step": 36725
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.236445714285715e-06,
+      "loss": 1.2226,
+      "step": 36726
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.236177142857143e-06,
+      "loss": 0.9545,
+      "step": 36727
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.235908571428572e-06,
+      "loss": 1.0344,
+      "step": 36728
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.23564e-06,
+      "loss": 1.1174,
+      "step": 36729
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.235371428571428e-06,
+      "loss": 0.7089,
+      "step": 36730
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.235102857142857e-06,
+      "loss": 1.639,
+      "step": 36731
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.234834285714285e-06,
+      "loss": 1.0537,
+      "step": 36732
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.234565714285714e-06,
+      "loss": 0.8569,
+      "step": 36733
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.234297142857143e-06,
+      "loss": 1.2592,
+      "step": 36734
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.234028571428572e-06,
+      "loss": 1.4726,
+      "step": 36735
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.233760000000001e-06,
+      "loss": 1.1785,
+      "step": 36736
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.233491428571429e-06,
+      "loss": 1.3854,
+      "step": 36737
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.233222857142858e-06,
+      "loss": 1.2737,
+      "step": 36738
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.232954285714286e-06,
+      "loss": 1.163,
+      "step": 36739
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.232685714285714e-06,
+      "loss": 1.0702,
+      "step": 36740
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.232417142857143e-06,
+      "loss": 1.272,
+      "step": 36741
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.232148571428571e-06,
+      "loss": 1.4001,
+      "step": 36742
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2318799999999995e-06,
+      "loss": 1.3202,
+      "step": 36743
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2316114285714286e-06,
+      "loss": 0.7969,
+      "step": 36744
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.231342857142857e-06,
+      "loss": 1.376,
+      "step": 36745
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.231074285714286e-06,
+      "loss": 1.4891,
+      "step": 36746
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.230805714285715e-06,
+      "loss": 0.9873,
+      "step": 36747
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.230537142857143e-06,
+      "loss": 0.7578,
+      "step": 36748
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.230268571428572e-06,
+      "loss": 0.6149,
+      "step": 36749
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.23e-06,
+      "loss": 0.9919,
+      "step": 36750
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.229731428571429e-06,
+      "loss": 1.18,
+      "step": 36751
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.229462857142857e-06,
+      "loss": 1.0805,
+      "step": 36752
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2291942857142855e-06,
+      "loss": 1.4996,
+      "step": 36753
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2289257142857146e-06,
+      "loss": 0.8691,
+      "step": 36754
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.228657142857143e-06,
+      "loss": 1.5051,
+      "step": 36755
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.228388571428572e-06,
+      "loss": 1.2951,
+      "step": 36756
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.22812e-06,
+      "loss": 0.9811,
+      "step": 36757
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.227851428571428e-06,
+      "loss": 1.0004,
+      "step": 36758
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.227582857142857e-06,
+      "loss": 0.7716,
+      "step": 36759
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.227314285714286e-06,
+      "loss": 1.0891,
+      "step": 36760
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.227045714285714e-06,
+      "loss": 0.9192,
+      "step": 36761
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.226777142857143e-06,
+      "loss": 0.8529,
+      "step": 36762
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2265085714285715e-06,
+      "loss": 1.2594,
+      "step": 36763
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2262400000000006e-06,
+      "loss": 1.1125,
+      "step": 36764
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.225971428571429e-06,
+      "loss": 0.9595,
+      "step": 36765
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.225702857142857e-06,
+      "loss": 1.3418,
+      "step": 36766
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.225434285714286e-06,
+      "loss": 1.3206,
+      "step": 36767
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.225165714285714e-06,
+      "loss": 0.921,
+      "step": 36768
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.224897142857143e-06,
+      "loss": 1.4382,
+      "step": 36769
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.224628571428571e-06,
+      "loss": 1.0255,
+      "step": 36770
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2243599999999995e-06,
+      "loss": 1.1274,
+      "step": 36771
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2240914285714285e-06,
+      "loss": 0.7363,
+      "step": 36772
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2238228571428575e-06,
+      "loss": 0.9947,
+      "step": 36773
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2235542857142865e-06,
+      "loss": 1.1211,
+      "step": 36774
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.223285714285715e-06,
+      "loss": 1.324,
+      "step": 36775
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.223017142857143e-06,
+      "loss": 0.968,
+      "step": 36776
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.222748571428572e-06,
+      "loss": 1.137,
+      "step": 36777
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.22248e-06,
+      "loss": 0.8509,
+      "step": 36778
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.222211428571429e-06,
+      "loss": 1.2472,
+      "step": 36779
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.221942857142857e-06,
+      "loss": 0.9246,
+      "step": 36780
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2216742857142855e-06,
+      "loss": 1.2424,
+      "step": 36781
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2214057142857145e-06,
+      "loss": 1.1714,
+      "step": 36782
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.221137142857143e-06,
+      "loss": 0.7768,
+      "step": 36783
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.220868571428571e-06,
+      "loss": 0.8832,
+      "step": 36784
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2206e-06,
+      "loss": 0.7708,
+      "step": 36785
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.220331428571429e-06,
+      "loss": 0.8997,
+      "step": 36786
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.220062857142858e-06,
+      "loss": 0.5974,
+      "step": 36787
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.219794285714286e-06,
+      "loss": 1.0405,
+      "step": 36788
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.219525714285714e-06,
+      "loss": 1.1938,
+      "step": 36789
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.219257142857143e-06,
+      "loss": 1.0859,
+      "step": 36790
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2189885714285715e-06,
+      "loss": 1.1595,
+      "step": 36791
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2187200000000005e-06,
+      "loss": 0.8797,
+      "step": 36792
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.218451428571429e-06,
+      "loss": 0.9526,
+      "step": 36793
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.218182857142857e-06,
+      "loss": 1.0002,
+      "step": 36794
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.217914285714286e-06,
+      "loss": 1.148,
+      "step": 36795
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.217645714285714e-06,
+      "loss": 1.1191,
+      "step": 36796
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.217377142857143e-06,
+      "loss": 1.2398,
+      "step": 36797
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.217108571428571e-06,
+      "loss": 0.9224,
+      "step": 36798
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.21684e-06,
+      "loss": 0.8288,
+      "step": 36799
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.216571428571429e-06,
+      "loss": 1.232,
+      "step": 36800
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2163028571428575e-06,
+      "loss": 0.9528,
+      "step": 36801
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2160342857142865e-06,
+      "loss": 1.0313,
+      "step": 36802
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.215765714285715e-06,
+      "loss": 1.1521,
+      "step": 36803
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.215497142857143e-06,
+      "loss": 1.022,
+      "step": 36804
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.215228571428572e-06,
+      "loss": 1.3993,
+      "step": 36805
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.21496e-06,
+      "loss": 1.0619,
+      "step": 36806
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.214691428571428e-06,
+      "loss": 1.0273,
+      "step": 36807
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.214422857142857e-06,
+      "loss": 0.9192,
+      "step": 36808
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.214154285714285e-06,
+      "loss": 1.1676,
+      "step": 36809
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2138857142857144e-06,
+      "loss": 1.0381,
+      "step": 36810
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.213617142857143e-06,
+      "loss": 1.4189,
+      "step": 36811
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.213348571428572e-06,
+      "loss": 1.0459,
+      "step": 36812
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.213080000000001e-06,
+      "loss": 1.0664,
+      "step": 36813
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.212811428571429e-06,
+      "loss": 1.0236,
+      "step": 36814
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.212542857142858e-06,
+      "loss": 0.992,
+      "step": 36815
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.212274285714286e-06,
+      "loss": 0.9629,
+      "step": 36816
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.212005714285714e-06,
+      "loss": 1.3554,
+      "step": 36817
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.211737142857143e-06,
+      "loss": 0.8,
+      "step": 36818
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.211468571428571e-06,
+      "loss": 0.9785,
+      "step": 36819
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.2112000000000004e-06,
+      "loss": 1.234,
+      "step": 36820
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.210931428571429e-06,
+      "loss": 1.0395,
+      "step": 36821
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.210662857142857e-06,
+      "loss": 1.3387,
+      "step": 36822
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.210394285714286e-06,
+      "loss": 1.5507,
+      "step": 36823
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.210125714285714e-06,
+      "loss": 1.5037,
+      "step": 36824
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.209857142857143e-06,
+      "loss": 1.1555,
+      "step": 36825
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.209588571428572e-06,
+      "loss": 1.1033,
+      "step": 36826
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.20932e-06,
+      "loss": 0.7409,
+      "step": 36827
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.209051428571429e-06,
+      "loss": 1.2235,
+      "step": 36828
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.208782857142857e-06,
+      "loss": 0.8119,
+      "step": 36829
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.208514285714286e-06,
+      "loss": 0.9023,
+      "step": 36830
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.208245714285715e-06,
+      "loss": 1.0945,
+      "step": 36831
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.207977142857143e-06,
+      "loss": 1.2492,
+      "step": 36832
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.207708571428572e-06,
+      "loss": 0.8266,
+      "step": 36833
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.20744e-06,
+      "loss": 1.4754,
+      "step": 36834
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.207171428571428e-06,
+      "loss": 0.9805,
+      "step": 36835
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.206902857142857e-06,
+      "loss": 0.9992,
+      "step": 36836
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.206634285714285e-06,
+      "loss": 0.9745,
+      "step": 36837
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.206365714285714e-06,
+      "loss": 1.1745,
+      "step": 36838
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.206097142857143e-06,
+      "loss": 1.1826,
+      "step": 36839
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.205828571428572e-06,
+      "loss": 1.0334,
+      "step": 36840
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.205560000000001e-06,
+      "loss": 0.9215,
+      "step": 36841
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.205291428571429e-06,
+      "loss": 0.9877,
+      "step": 36842
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.205022857142858e-06,
+      "loss": 0.6891,
+      "step": 36843
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.204754285714286e-06,
+      "loss": 1.1228,
+      "step": 36844
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.204485714285714e-06,
+      "loss": 1.3349,
+      "step": 36845
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.204217142857143e-06,
+      "loss": 0.7916,
+      "step": 36846
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.203948571428571e-06,
+      "loss": 1.2132,
+      "step": 36847
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.2036799999999995e-06,
+      "loss": 1.4274,
+      "step": 36848
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.2034114285714286e-06,
+      "loss": 1.0508,
+      "step": 36849
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.203142857142857e-06,
+      "loss": 1.2871,
+      "step": 36850
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.202874285714286e-06,
+      "loss": 1.123,
+      "step": 36851
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.202605714285715e-06,
+      "loss": 0.977,
+      "step": 36852
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.202337142857143e-06,
+      "loss": 0.9954,
+      "step": 36853
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.202068571428572e-06,
+      "loss": 0.8771,
+      "step": 36854
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.2018e-06,
+      "loss": 0.8264,
+      "step": 36855
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.201531428571429e-06,
+      "loss": 1.0832,
+      "step": 36856
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.201262857142857e-06,
+      "loss": 1.1582,
+      "step": 36857
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.2009942857142855e-06,
+      "loss": 1.2644,
+      "step": 36858
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.2007257142857146e-06,
+      "loss": 0.6917,
+      "step": 36859
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.200457142857143e-06,
+      "loss": 1.0146,
+      "step": 36860
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.200188571428572e-06,
+      "loss": 0.9962,
+      "step": 36861
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.19992e-06,
+      "loss": 1.342,
+      "step": 36862
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.199651428571428e-06,
+      "loss": 1.0497,
+      "step": 36863
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.199382857142857e-06,
+      "loss": 1.1291,
+      "step": 36864
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.199114285714286e-06,
+      "loss": 1.0935,
+      "step": 36865
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.198845714285715e-06,
+      "loss": 1.3266,
+      "step": 36866
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.198577142857143e-06,
+      "loss": 0.9054,
+      "step": 36867
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1983085714285715e-06,
+      "loss": 1.3325,
+      "step": 36868
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1980400000000006e-06,
+      "loss": 1.1555,
+      "step": 36869
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.197771428571429e-06,
+      "loss": 1.2221,
+      "step": 36870
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.197502857142857e-06,
+      "loss": 0.8091,
+      "step": 36871
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.197234285714286e-06,
+      "loss": 1.0026,
+      "step": 36872
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.196965714285714e-06,
+      "loss": 1.2254,
+      "step": 36873
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.196697142857143e-06,
+      "loss": 1.1013,
+      "step": 36874
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.196428571428571e-06,
+      "loss": 0.8294,
+      "step": 36875
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1961599999999995e-06,
+      "loss": 0.9639,
+      "step": 36876
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1958914285714285e-06,
+      "loss": 1.4342,
+      "step": 36877
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1956228571428575e-06,
+      "loss": 0.9566,
+      "step": 36878
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1953542857142865e-06,
+      "loss": 1.0241,
+      "step": 36879
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.195085714285715e-06,
+      "loss": 1.4228,
+      "step": 36880
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.194817142857143e-06,
+      "loss": 1.268,
+      "step": 36881
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.194548571428572e-06,
+      "loss": 1.0238,
+      "step": 36882
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.19428e-06,
+      "loss": 1.0911,
+      "step": 36883
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.194011428571429e-06,
+      "loss": 0.9007,
+      "step": 36884
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.193742857142857e-06,
+      "loss": 1.3281,
+      "step": 36885
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1934742857142855e-06,
+      "loss": 1.2919,
+      "step": 36886
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1932057142857145e-06,
+      "loss": 1.173,
+      "step": 36887
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.192937142857143e-06,
+      "loss": 1.3334,
+      "step": 36888
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.192668571428572e-06,
+      "loss": 0.9173,
+      "step": 36889
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1924e-06,
+      "loss": 1.2763,
+      "step": 36890
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.192131428571429e-06,
+      "loss": 1.048,
+      "step": 36891
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.191862857142858e-06,
+      "loss": 1.0405,
+      "step": 36892
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.191594285714286e-06,
+      "loss": 1.4083,
+      "step": 36893
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.191325714285714e-06,
+      "loss": 0.8762,
+      "step": 36894
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.191057142857143e-06,
+      "loss": 0.6694,
+      "step": 36895
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1907885714285715e-06,
+      "loss": 1.1187,
+      "step": 36896
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1905200000000005e-06,
+      "loss": 1.0375,
+      "step": 36897
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.190251428571429e-06,
+      "loss": 1.2079,
+      "step": 36898
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.189982857142857e-06,
+      "loss": 1.277,
+      "step": 36899
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.189714285714286e-06,
+      "loss": 1.1979,
+      "step": 36900
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.189445714285714e-06,
+      "loss": 0.9248,
+      "step": 36901
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.189177142857143e-06,
+      "loss": 1.6751,
+      "step": 36902
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.188908571428571e-06,
+      "loss": 1.3467,
+      "step": 36903
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.18864e-06,
+      "loss": 1.1086,
+      "step": 36904
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.188371428571429e-06,
+      "loss": 1.0599,
+      "step": 36905
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1881028571428575e-06,
+      "loss": 0.6316,
+      "step": 36906
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1878342857142865e-06,
+      "loss": 1.1862,
+      "step": 36907
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.187565714285715e-06,
+      "loss": 1.2904,
+      "step": 36908
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.187297142857143e-06,
+      "loss": 1.4422,
+      "step": 36909
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.187028571428572e-06,
+      "loss": 1.2581,
+      "step": 36910
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.18676e-06,
+      "loss": 1.4019,
+      "step": 36911
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.186491428571429e-06,
+      "loss": 1.0951,
+      "step": 36912
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.186222857142857e-06,
+      "loss": 0.7558,
+      "step": 36913
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.185954285714285e-06,
+      "loss": 1.3444,
+      "step": 36914
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1856857142857144e-06,
+      "loss": 0.7478,
+      "step": 36915
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.185417142857143e-06,
+      "loss": 1.0673,
+      "step": 36916
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.185148571428572e-06,
+      "loss": 1.4583,
+      "step": 36917
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.184880000000001e-06,
+      "loss": 1.2626,
+      "step": 36918
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.184611428571429e-06,
+      "loss": 1.1662,
+      "step": 36919
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.184342857142858e-06,
+      "loss": 1.0015,
+      "step": 36920
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.184074285714286e-06,
+      "loss": 0.8397,
+      "step": 36921
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.183805714285714e-06,
+      "loss": 1.1454,
+      "step": 36922
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.183537142857143e-06,
+      "loss": 1.1414,
+      "step": 36923
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.183268571428571e-06,
+      "loss": 1.8201,
+      "step": 36924
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1830000000000004e-06,
+      "loss": 1.0592,
+      "step": 36925
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.182731428571429e-06,
+      "loss": 1.0263,
+      "step": 36926
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.182462857142857e-06,
+      "loss": 1.2918,
+      "step": 36927
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.182194285714286e-06,
+      "loss": 1.257,
+      "step": 36928
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.181925714285714e-06,
+      "loss": 1.1676,
+      "step": 36929
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.181657142857143e-06,
+      "loss": 1.3616,
+      "step": 36930
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.181388571428572e-06,
+      "loss": 0.938,
+      "step": 36931
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.18112e-06,
+      "loss": 0.9406,
+      "step": 36932
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.180851428571429e-06,
+      "loss": 1.1742,
+      "step": 36933
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.180582857142857e-06,
+      "loss": 1.0318,
+      "step": 36934
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.180314285714286e-06,
+      "loss": 1.2684,
+      "step": 36935
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.180045714285715e-06,
+      "loss": 1.4648,
+      "step": 36936
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.179777142857143e-06,
+      "loss": 1.0925,
+      "step": 36937
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.179508571428572e-06,
+      "loss": 1.4574,
+      "step": 36938
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.17924e-06,
+      "loss": 1.3155,
+      "step": 36939
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.178971428571428e-06,
+      "loss": 1.4124,
+      "step": 36940
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.178702857142857e-06,
+      "loss": 1.4131,
+      "step": 36941
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.178434285714285e-06,
+      "loss": 0.9065,
+      "step": 36942
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.178165714285714e-06,
+      "loss": 1.5708,
+      "step": 36943
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.177897142857143e-06,
+      "loss": 1.4896,
+      "step": 36944
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.177628571428572e-06,
+      "loss": 0.9488,
+      "step": 36945
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.177360000000001e-06,
+      "loss": 1.0772,
+      "step": 36946
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.177091428571429e-06,
+      "loss": 1.3004,
+      "step": 36947
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.176822857142858e-06,
+      "loss": 1.0693,
+      "step": 36948
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.176554285714286e-06,
+      "loss": 1.171,
+      "step": 36949
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.176285714285714e-06,
+      "loss": 0.9593,
+      "step": 36950
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.176017142857143e-06,
+      "loss": 0.7253,
+      "step": 36951
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.175748571428571e-06,
+      "loss": 1.1292,
+      "step": 36952
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.17548e-06,
+      "loss": 0.9164,
+      "step": 36953
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1752114285714286e-06,
+      "loss": 0.9505,
+      "step": 36954
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.174942857142857e-06,
+      "loss": 1.2467,
+      "step": 36955
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.174674285714286e-06,
+      "loss": 1.5372,
+      "step": 36956
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.174405714285715e-06,
+      "loss": 1.1108,
+      "step": 36957
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.174137142857143e-06,
+      "loss": 0.949,
+      "step": 36958
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.173868571428572e-06,
+      "loss": 0.8126,
+      "step": 36959
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1736e-06,
+      "loss": 1.3103,
+      "step": 36960
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.173331428571429e-06,
+      "loss": 1.2004,
+      "step": 36961
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.173062857142857e-06,
+      "loss": 1.3895,
+      "step": 36962
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1727942857142855e-06,
+      "loss": 0.9796,
+      "step": 36963
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1725257142857146e-06,
+      "loss": 0.8492,
+      "step": 36964
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.172257142857143e-06,
+      "loss": 1.2073,
+      "step": 36965
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.171988571428572e-06,
+      "loss": 1.3315,
+      "step": 36966
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.17172e-06,
+      "loss": 1.0764,
+      "step": 36967
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.171451428571428e-06,
+      "loss": 1.228,
+      "step": 36968
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.171182857142857e-06,
+      "loss": 1.4338,
+      "step": 36969
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.170914285714286e-06,
+      "loss": 0.6182,
+      "step": 36970
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.170645714285715e-06,
+      "loss": 0.8844,
+      "step": 36971
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.170377142857143e-06,
+      "loss": 0.8806,
+      "step": 36972
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1701085714285715e-06,
+      "loss": 0.7127,
+      "step": 36973
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1698400000000006e-06,
+      "loss": 1.6699,
+      "step": 36974
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.169571428571429e-06,
+      "loss": 1.4508,
+      "step": 36975
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.169302857142858e-06,
+      "loss": 1.3916,
+      "step": 36976
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.169034285714286e-06,
+      "loss": 0.7961,
+      "step": 36977
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.168765714285714e-06,
+      "loss": 1.2179,
+      "step": 36978
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.168497142857143e-06,
+      "loss": 1.1856,
+      "step": 36979
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.168228571428571e-06,
+      "loss": 0.9346,
+      "step": 36980
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1679599999999995e-06,
+      "loss": 1.1612,
+      "step": 36981
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1676914285714285e-06,
+      "loss": 1.2997,
+      "step": 36982
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1674228571428575e-06,
+      "loss": 1.1951,
+      "step": 36983
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1671542857142865e-06,
+      "loss": 1.1089,
+      "step": 36984
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.166885714285715e-06,
+      "loss": 1.1408,
+      "step": 36985
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.166617142857143e-06,
+      "loss": 0.9959,
+      "step": 36986
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.166348571428572e-06,
+      "loss": 1.1124,
+      "step": 36987
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.16608e-06,
+      "loss": 1.0606,
+      "step": 36988
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.165811428571429e-06,
+      "loss": 1.3067,
+      "step": 36989
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.165542857142857e-06,
+      "loss": 1.052,
+      "step": 36990
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1652742857142855e-06,
+      "loss": 1.0877,
+      "step": 36991
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1650057142857145e-06,
+      "loss": 1.2751,
+      "step": 36992
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.164737142857143e-06,
+      "loss": 0.6945,
+      "step": 36993
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.164468571428572e-06,
+      "loss": 1.56,
+      "step": 36994
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1642e-06,
+      "loss": 1.0775,
+      "step": 36995
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.163931428571428e-06,
+      "loss": 0.7571,
+      "step": 36996
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.163662857142858e-06,
+      "loss": 0.7413,
+      "step": 36997
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.163394285714286e-06,
+      "loss": 0.9285,
+      "step": 36998
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.163125714285715e-06,
+      "loss": 1.0262,
+      "step": 36999
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.162857142857143e-06,
+      "loss": 1.1271,
+      "step": 37000
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1625885714285715e-06,
+      "loss": 1.0037,
+      "step": 37001
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1623200000000005e-06,
+      "loss": 1.3442,
+      "step": 37002
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.162051428571429e-06,
+      "loss": 1.3173,
+      "step": 37003
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.161782857142857e-06,
+      "loss": 0.8751,
+      "step": 37004
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.161514285714286e-06,
+      "loss": 1.0413,
+      "step": 37005
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.161245714285714e-06,
+      "loss": 1.0278,
+      "step": 37006
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.160977142857143e-06,
+      "loss": 1.0466,
+      "step": 37007
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.160708571428571e-06,
+      "loss": 1.2888,
+      "step": 37008
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.160439999999999e-06,
+      "loss": 1.0442,
+      "step": 37009
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.160171428571429e-06,
+      "loss": 1.5181,
+      "step": 37010
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1599028571428575e-06,
+      "loss": 1.3366,
+      "step": 37011
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.1596342857142865e-06,
+      "loss": 0.7633,
+      "step": 37012
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.159365714285715e-06,
+      "loss": 0.7692,
+      "step": 37013
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.159097142857143e-06,
+      "loss": 1.1366,
+      "step": 37014
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.158828571428572e-06,
+      "loss": 0.9875,
+      "step": 37015
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.15856e-06,
+      "loss": 0.8569,
+      "step": 37016
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.158291428571429e-06,
+      "loss": 0.8041,
+      "step": 37017
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.158022857142857e-06,
+      "loss": 0.8684,
+      "step": 37018
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.157754285714285e-06,
+      "loss": 1.0443,
+      "step": 37019
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1574857142857144e-06,
+      "loss": 0.8684,
+      "step": 37020
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.157217142857143e-06,
+      "loss": 1.2023,
+      "step": 37021
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.156948571428571e-06,
+      "loss": 1.2736,
+      "step": 37022
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.156680000000001e-06,
+      "loss": 0.5903,
+      "step": 37023
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.156411428571429e-06,
+      "loss": 1.1782,
+      "step": 37024
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.156142857142858e-06,
+      "loss": 0.9572,
+      "step": 37025
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.155874285714286e-06,
+      "loss": 1.0655,
+      "step": 37026
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.155605714285714e-06,
+      "loss": 1.0434,
+      "step": 37027
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.155337142857143e-06,
+      "loss": 1.0481,
+      "step": 37028
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.155068571428571e-06,
+      "loss": 1.2456,
+      "step": 37029
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1548000000000004e-06,
+      "loss": 1.0336,
+      "step": 37030
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.154531428571429e-06,
+      "loss": 1.34,
+      "step": 37031
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.154262857142857e-06,
+      "loss": 1.0239,
+      "step": 37032
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.153994285714286e-06,
+      "loss": 1.2226,
+      "step": 37033
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.153725714285714e-06,
+      "loss": 1.0661,
+      "step": 37034
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.153457142857143e-06,
+      "loss": 0.5581,
+      "step": 37035
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.153188571428572e-06,
+      "loss": 0.8659,
+      "step": 37036
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.15292e-06,
+      "loss": 0.999,
+      "step": 37037
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.152651428571429e-06,
+      "loss": 0.9464,
+      "step": 37038
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.152382857142857e-06,
+      "loss": 1.1226,
+      "step": 37039
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1521142857142864e-06,
+      "loss": 1.6407,
+      "step": 37040
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.151845714285715e-06,
+      "loss": 1.1627,
+      "step": 37041
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.151577142857143e-06,
+      "loss": 0.8441,
+      "step": 37042
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.151308571428572e-06,
+      "loss": 1.23,
+      "step": 37043
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.15104e-06,
+      "loss": 0.9939,
+      "step": 37044
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.150771428571428e-06,
+      "loss": 1.3396,
+      "step": 37045
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.150502857142857e-06,
+      "loss": 1.3009,
+      "step": 37046
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.150234285714285e-06,
+      "loss": 1.0972,
+      "step": 37047
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.149965714285714e-06,
+      "loss": 1.0923,
+      "step": 37048
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1496971428571426e-06,
+      "loss": 1.3558,
+      "step": 37049
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.149428571428572e-06,
+      "loss": 1.0815,
+      "step": 37050
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.149160000000001e-06,
+      "loss": 1.4498,
+      "step": 37051
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.148891428571429e-06,
+      "loss": 1.0365,
+      "step": 37052
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.148622857142858e-06,
+      "loss": 1.6459,
+      "step": 37053
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.148354285714286e-06,
+      "loss": 1.0229,
+      "step": 37054
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.148085714285714e-06,
+      "loss": 0.7874,
+      "step": 37055
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.147817142857143e-06,
+      "loss": 1.2755,
+      "step": 37056
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.147548571428571e-06,
+      "loss": 0.9796,
+      "step": 37057
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.14728e-06,
+      "loss": 1.0606,
+      "step": 37058
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1470114285714286e-06,
+      "loss": 1.0644,
+      "step": 37059
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.146742857142857e-06,
+      "loss": 1.0991,
+      "step": 37060
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.146474285714286e-06,
+      "loss": 0.9857,
+      "step": 37061
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.146205714285714e-06,
+      "loss": 1.1843,
+      "step": 37062
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.145937142857144e-06,
+      "loss": 1.1226,
+      "step": 37063
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.145668571428572e-06,
+      "loss": 1.0481,
+      "step": 37064
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1454e-06,
+      "loss": 1.1332,
+      "step": 37065
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.145131428571429e-06,
+      "loss": 0.8169,
+      "step": 37066
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.144862857142857e-06,
+      "loss": 0.6659,
+      "step": 37067
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1445942857142855e-06,
+      "loss": 0.869,
+      "step": 37068
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1443257142857146e-06,
+      "loss": 1.4572,
+      "step": 37069
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.144057142857143e-06,
+      "loss": 1.4615,
+      "step": 37070
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.143788571428572e-06,
+      "loss": 1.6584,
+      "step": 37071
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.14352e-06,
+      "loss": 1.1169,
+      "step": 37072
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.143251428571428e-06,
+      "loss": 0.7236,
+      "step": 37073
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.142982857142857e-06,
+      "loss": 1.698,
+      "step": 37074
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.142714285714285e-06,
+      "loss": 0.6082,
+      "step": 37075
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.142445714285715e-06,
+      "loss": 1.3451,
+      "step": 37076
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.142177142857143e-06,
+      "loss": 1.1454,
+      "step": 37077
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1419085714285715e-06,
+      "loss": 1.0379,
+      "step": 37078
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1416400000000005e-06,
+      "loss": 1.2872,
+      "step": 37079
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.141371428571429e-06,
+      "loss": 1.5966,
+      "step": 37080
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.141102857142858e-06,
+      "loss": 1.1398,
+      "step": 37081
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.140834285714286e-06,
+      "loss": 1.268,
+      "step": 37082
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.140565714285714e-06,
+      "loss": 1.1407,
+      "step": 37083
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.140297142857143e-06,
+      "loss": 1.0751,
+      "step": 37084
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.140028571428571e-06,
+      "loss": 1.1739,
+      "step": 37085
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.13976e-06,
+      "loss": 1.2454,
+      "step": 37086
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1394914285714285e-06,
+      "loss": 1.3754,
+      "step": 37087
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.139222857142857e-06,
+      "loss": 0.9943,
+      "step": 37088
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1389542857142865e-06,
+      "loss": 1.1846,
+      "step": 37089
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.138685714285715e-06,
+      "loss": 1.1126,
+      "step": 37090
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.138417142857143e-06,
+      "loss": 1.1809,
+      "step": 37091
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.138148571428572e-06,
+      "loss": 1.1572,
+      "step": 37092
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.13788e-06,
+      "loss": 1.2251,
+      "step": 37093
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.137611428571429e-06,
+      "loss": 1.3421,
+      "step": 37094
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.137342857142857e-06,
+      "loss": 1.5189,
+      "step": 37095
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1370742857142855e-06,
+      "loss": 1.0024,
+      "step": 37096
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1368057142857145e-06,
+      "loss": 0.8273,
+      "step": 37097
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.136537142857143e-06,
+      "loss": 1.1425,
+      "step": 37098
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.136268571428572e-06,
+      "loss": 0.9167,
+      "step": 37099
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.136e-06,
+      "loss": 0.8977,
+      "step": 37100
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.135731428571428e-06,
+      "loss": 1.2109,
+      "step": 37101
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.135462857142857e-06,
+      "loss": 0.9851,
+      "step": 37102
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.135194285714286e-06,
+      "loss": 1.3441,
+      "step": 37103
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.134925714285715e-06,
+      "loss": 1.0757,
+      "step": 37104
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.134657142857143e-06,
+      "loss": 1.3616,
+      "step": 37105
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1343885714285715e-06,
+      "loss": 0.6688,
+      "step": 37106
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1341200000000005e-06,
+      "loss": 1.0221,
+      "step": 37107
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.133851428571429e-06,
+      "loss": 1.0046,
+      "step": 37108
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.133582857142857e-06,
+      "loss": 1.4072,
+      "step": 37109
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.133314285714286e-06,
+      "loss": 0.8697,
+      "step": 37110
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.133045714285714e-06,
+      "loss": 1.3135,
+      "step": 37111
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.132777142857143e-06,
+      "loss": 1.2191,
+      "step": 37112
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.132508571428571e-06,
+      "loss": 0.9541,
+      "step": 37113
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.132239999999999e-06,
+      "loss": 1.0013,
+      "step": 37114
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1319714285714284e-06,
+      "loss": 1.042,
+      "step": 37115
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1317028571428575e-06,
+      "loss": 1.1688,
+      "step": 37116
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1314342857142865e-06,
+      "loss": 1.2431,
+      "step": 37117
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.131165714285715e-06,
+      "loss": 1.0058,
+      "step": 37118
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.130897142857143e-06,
+      "loss": 1.342,
+      "step": 37119
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.130628571428572e-06,
+      "loss": 0.877,
+      "step": 37120
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.13036e-06,
+      "loss": 1.3328,
+      "step": 37121
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.130091428571429e-06,
+      "loss": 1.6673,
+      "step": 37122
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.129822857142857e-06,
+      "loss": 1.3854,
+      "step": 37123
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.129554285714285e-06,
+      "loss": 0.9707,
+      "step": 37124
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1292857142857144e-06,
+      "loss": 1.2462,
+      "step": 37125
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.129017142857143e-06,
+      "loss": 1.2744,
+      "step": 37126
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.128748571428572e-06,
+      "loss": 1.3203,
+      "step": 37127
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.12848e-06,
+      "loss": 1.0729,
+      "step": 37128
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.128211428571429e-06,
+      "loss": 1.1526,
+      "step": 37129
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.127942857142858e-06,
+      "loss": 0.9315,
+      "step": 37130
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.127674285714286e-06,
+      "loss": 0.925,
+      "step": 37131
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.127405714285714e-06,
+      "loss": 1.1439,
+      "step": 37132
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.127137142857143e-06,
+      "loss": 1.0831,
+      "step": 37133
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.126868571428571e-06,
+      "loss": 1.327,
+      "step": 37134
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1266000000000004e-06,
+      "loss": 1.0126,
+      "step": 37135
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.126331428571429e-06,
+      "loss": 0.9865,
+      "step": 37136
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.126062857142857e-06,
+      "loss": 0.8867,
+      "step": 37137
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.125794285714286e-06,
+      "loss": 0.8016,
+      "step": 37138
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.125525714285714e-06,
+      "loss": 0.6901,
+      "step": 37139
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.125257142857143e-06,
+      "loss": 1.1337,
+      "step": 37140
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.124988571428571e-06,
+      "loss": 1.1099,
+      "step": 37141
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.12472e-06,
+      "loss": 0.8916,
+      "step": 37142
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.124451428571429e-06,
+      "loss": 1.0231,
+      "step": 37143
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.124182857142857e-06,
+      "loss": 0.8743,
+      "step": 37144
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1239142857142864e-06,
+      "loss": 0.8489,
+      "step": 37145
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.123645714285715e-06,
+      "loss": 1.1665,
+      "step": 37146
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.123377142857143e-06,
+      "loss": 0.8226,
+      "step": 37147
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.123108571428572e-06,
+      "loss": 1.0099,
+      "step": 37148
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.12284e-06,
+      "loss": 1.4958,
+      "step": 37149
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.122571428571429e-06,
+      "loss": 1.4631,
+      "step": 37150
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.122302857142857e-06,
+      "loss": 0.9292,
+      "step": 37151
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.122034285714285e-06,
+      "loss": 1.1093,
+      "step": 37152
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.121765714285714e-06,
+      "loss": 1.0029,
+      "step": 37153
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1214971428571426e-06,
+      "loss": 1.0215,
+      "step": 37154
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.121228571428572e-06,
+      "loss": 1.0443,
+      "step": 37155
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.120960000000001e-06,
+      "loss": 1.3682,
+      "step": 37156
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.120691428571429e-06,
+      "loss": 1.1593,
+      "step": 37157
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.120422857142858e-06,
+      "loss": 1.2173,
+      "step": 37158
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.120154285714286e-06,
+      "loss": 1.0453,
+      "step": 37159
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.119885714285714e-06,
+      "loss": 1.5298,
+      "step": 37160
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.119617142857143e-06,
+      "loss": 1.3051,
+      "step": 37161
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.119348571428571e-06,
+      "loss": 1.1467,
+      "step": 37162
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.11908e-06,
+      "loss": 1.3763,
+      "step": 37163
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1188114285714286e-06,
+      "loss": 1.3338,
+      "step": 37164
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.118542857142857e-06,
+      "loss": 1.1292,
+      "step": 37165
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.118274285714286e-06,
+      "loss": 1.408,
+      "step": 37166
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.118005714285714e-06,
+      "loss": 1.1303,
+      "step": 37167
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.117737142857143e-06,
+      "loss": 0.7978,
+      "step": 37168
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.117468571428572e-06,
+      "loss": 1.1889,
+      "step": 37169
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1172e-06,
+      "loss": 1.0336,
+      "step": 37170
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.116931428571429e-06,
+      "loss": 1.3797,
+      "step": 37171
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.116662857142857e-06,
+      "loss": 0.9479,
+      "step": 37172
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.116394285714286e-06,
+      "loss": 0.8554,
+      "step": 37173
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1161257142857146e-06,
+      "loss": 0.946,
+      "step": 37174
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.115857142857143e-06,
+      "loss": 1.4805,
+      "step": 37175
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.115588571428572e-06,
+      "loss": 1.0872,
+      "step": 37176
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.11532e-06,
+      "loss": 1.4422,
+      "step": 37177
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.115051428571428e-06,
+      "loss": 0.8846,
+      "step": 37178
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.114782857142857e-06,
+      "loss": 0.9535,
+      "step": 37179
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.114514285714285e-06,
+      "loss": 1.0213,
+      "step": 37180
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.114245714285714e-06,
+      "loss": 1.6299,
+      "step": 37181
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.113977142857143e-06,
+      "loss": 0.9297,
+      "step": 37182
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1137085714285715e-06,
+      "loss": 1.3336,
+      "step": 37183
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.1134400000000005e-06,
+      "loss": 0.8572,
+      "step": 37184
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.113171428571429e-06,
+      "loss": 0.9192,
+      "step": 37185
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.112902857142858e-06,
+      "loss": 1.5985,
+      "step": 37186
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.112634285714286e-06,
+      "loss": 1.0682,
+      "step": 37187
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.112365714285714e-06,
+      "loss": 1.1962,
+      "step": 37188
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.112097142857143e-06,
+      "loss": 1.2571,
+      "step": 37189
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.111828571428571e-06,
+      "loss": 1.164,
+      "step": 37190
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.11156e-06,
+      "loss": 1.1432,
+      "step": 37191
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1112914285714285e-06,
+      "loss": 0.8752,
+      "step": 37192
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.111022857142857e-06,
+      "loss": 1.025,
+      "step": 37193
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.110754285714286e-06,
+      "loss": 1.1174,
+      "step": 37194
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.110485714285715e-06,
+      "loss": 1.3475,
+      "step": 37195
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.110217142857143e-06,
+      "loss": 1.0846,
+      "step": 37196
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.109948571428572e-06,
+      "loss": 1.1359,
+      "step": 37197
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.10968e-06,
+      "loss": 0.8947,
+      "step": 37198
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.109411428571429e-06,
+      "loss": 1.1237,
+      "step": 37199
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.109142857142857e-06,
+      "loss": 0.9652,
+      "step": 37200
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1088742857142855e-06,
+      "loss": 1.2507,
+      "step": 37201
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1086057142857145e-06,
+      "loss": 0.8544,
+      "step": 37202
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.108337142857143e-06,
+      "loss": 0.8781,
+      "step": 37203
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.108068571428572e-06,
+      "loss": 1.5182,
+      "step": 37204
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1078e-06,
+      "loss": 1.3887,
+      "step": 37205
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.107531428571428e-06,
+      "loss": 1.1874,
+      "step": 37206
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.107262857142857e-06,
+      "loss": 0.8706,
+      "step": 37207
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.106994285714286e-06,
+      "loss": 0.8821,
+      "step": 37208
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.106725714285715e-06,
+      "loss": 0.9462,
+      "step": 37209
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.106457142857143e-06,
+      "loss": 1.0355,
+      "step": 37210
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1061885714285715e-06,
+      "loss": 0.7627,
+      "step": 37211
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1059200000000005e-06,
+      "loss": 1.0263,
+      "step": 37212
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.105651428571429e-06,
+      "loss": 1.6098,
+      "step": 37213
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.105382857142858e-06,
+      "loss": 1.4907,
+      "step": 37214
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.105114285714286e-06,
+      "loss": 1.1283,
+      "step": 37215
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.104845714285714e-06,
+      "loss": 1.2692,
+      "step": 37216
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.104577142857143e-06,
+      "loss": 0.8762,
+      "step": 37217
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.104308571428571e-06,
+      "loss": 1.3287,
+      "step": 37218
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.104039999999999e-06,
+      "loss": 0.9964,
+      "step": 37219
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1037714285714284e-06,
+      "loss": 1.1968,
+      "step": 37220
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1035028571428575e-06,
+      "loss": 1.0138,
+      "step": 37221
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1032342857142865e-06,
+      "loss": 1.3086,
+      "step": 37222
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.102965714285715e-06,
+      "loss": 1.3216,
+      "step": 37223
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.102697142857143e-06,
+      "loss": 1.345,
+      "step": 37224
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.102428571428572e-06,
+      "loss": 1.5519,
+      "step": 37225
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.10216e-06,
+      "loss": 1.3647,
+      "step": 37226
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.101891428571429e-06,
+      "loss": 1.0106,
+      "step": 37227
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.101622857142857e-06,
+      "loss": 1.2169,
+      "step": 37228
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.101354285714285e-06,
+      "loss": 1.1932,
+      "step": 37229
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.1010857142857144e-06,
+      "loss": 1.0192,
+      "step": 37230
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.100817142857143e-06,
+      "loss": 1.0929,
+      "step": 37231
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.100548571428572e-06,
+      "loss": 1.3816,
+      "step": 37232
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.10028e-06,
+      "loss": 1.3566,
+      "step": 37233
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.100011428571429e-06,
+      "loss": 0.8931,
+      "step": 37234
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.099742857142858e-06,
+      "loss": 1.2871,
+      "step": 37235
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.099474285714286e-06,
+      "loss": 1.0403,
+      "step": 37236
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.099205714285715e-06,
+      "loss": 0.8595,
+      "step": 37237
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.098937142857143e-06,
+      "loss": 0.8947,
+      "step": 37238
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.098668571428571e-06,
+      "loss": 1.2281,
+      "step": 37239
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0984000000000004e-06,
+      "loss": 1.1677,
+      "step": 37240
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.098131428571429e-06,
+      "loss": 0.9686,
+      "step": 37241
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.097862857142857e-06,
+      "loss": 1.0255,
+      "step": 37242
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.097594285714286e-06,
+      "loss": 1.0567,
+      "step": 37243
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.097325714285714e-06,
+      "loss": 0.8876,
+      "step": 37244
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.097057142857143e-06,
+      "loss": 1.3736,
+      "step": 37245
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.096788571428571e-06,
+      "loss": 1.0685,
+      "step": 37246
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.09652e-06,
+      "loss": 1.5752,
+      "step": 37247
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.096251428571429e-06,
+      "loss": 1.2105,
+      "step": 37248
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.095982857142857e-06,
+      "loss": 1.2712,
+      "step": 37249
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0957142857142864e-06,
+      "loss": 1.2116,
+      "step": 37250
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.095445714285715e-06,
+      "loss": 1.0046,
+      "step": 37251
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.095177142857143e-06,
+      "loss": 1.0055,
+      "step": 37252
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.094908571428572e-06,
+      "loss": 1.2418,
+      "step": 37253
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.09464e-06,
+      "loss": 0.7298,
+      "step": 37254
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.094371428571429e-06,
+      "loss": 0.7797,
+      "step": 37255
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.094102857142857e-06,
+      "loss": 0.9287,
+      "step": 37256
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.093834285714285e-06,
+      "loss": 0.8199,
+      "step": 37257
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.093565714285714e-06,
+      "loss": 0.9555,
+      "step": 37258
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0932971428571426e-06,
+      "loss": 0.6458,
+      "step": 37259
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.093028571428572e-06,
+      "loss": 1.2292,
+      "step": 37260
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.092760000000001e-06,
+      "loss": 1.0826,
+      "step": 37261
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.092491428571429e-06,
+      "loss": 1.0871,
+      "step": 37262
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.092222857142858e-06,
+      "loss": 1.3142,
+      "step": 37263
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.091954285714286e-06,
+      "loss": 0.919,
+      "step": 37264
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.091685714285714e-06,
+      "loss": 1.2599,
+      "step": 37265
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.091417142857143e-06,
+      "loss": 1.0228,
+      "step": 37266
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.091148571428571e-06,
+      "loss": 0.9942,
+      "step": 37267
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.09088e-06,
+      "loss": 0.7835,
+      "step": 37268
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0906114285714286e-06,
+      "loss": 0.5324,
+      "step": 37269
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.090342857142857e-06,
+      "loss": 1.0587,
+      "step": 37270
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.090074285714286e-06,
+      "loss": 1.0951,
+      "step": 37271
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.089805714285714e-06,
+      "loss": 1.3899,
+      "step": 37272
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.089537142857143e-06,
+      "loss": 1.1749,
+      "step": 37273
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.089268571428572e-06,
+      "loss": 0.7993,
+      "step": 37274
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.089e-06,
+      "loss": 1.3823,
+      "step": 37275
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.088731428571429e-06,
+      "loss": 1.0968,
+      "step": 37276
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.088462857142857e-06,
+      "loss": 1.1831,
+      "step": 37277
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.088194285714286e-06,
+      "loss": 1.2752,
+      "step": 37278
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0879257142857146e-06,
+      "loss": 1.2291,
+      "step": 37279
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.087657142857143e-06,
+      "loss": 0.9324,
+      "step": 37280
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.087388571428572e-06,
+      "loss": 1.1679,
+      "step": 37281
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.08712e-06,
+      "loss": 0.7261,
+      "step": 37282
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.086851428571428e-06,
+      "loss": 1.3133,
+      "step": 37283
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.086582857142857e-06,
+      "loss": 1.3901,
+      "step": 37284
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.086314285714285e-06,
+      "loss": 0.9137,
+      "step": 37285
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.086045714285714e-06,
+      "loss": 1.1931,
+      "step": 37286
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.085777142857143e-06,
+      "loss": 1.1835,
+      "step": 37287
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0855085714285715e-06,
+      "loss": 1.0128,
+      "step": 37288
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0852400000000005e-06,
+      "loss": 1.2116,
+      "step": 37289
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.084971428571429e-06,
+      "loss": 1.0076,
+      "step": 37290
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.084702857142858e-06,
+      "loss": 0.9694,
+      "step": 37291
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.084434285714286e-06,
+      "loss": 1.4006,
+      "step": 37292
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.084165714285714e-06,
+      "loss": 1.1643,
+      "step": 37293
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.083897142857143e-06,
+      "loss": 1.0386,
+      "step": 37294
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.083628571428571e-06,
+      "loss": 1.1091,
+      "step": 37295
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.08336e-06,
+      "loss": 1.0885,
+      "step": 37296
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0830914285714285e-06,
+      "loss": 1.0081,
+      "step": 37297
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.082822857142857e-06,
+      "loss": 1.0023,
+      "step": 37298
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.082554285714286e-06,
+      "loss": 0.9348,
+      "step": 37299
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.082285714285715e-06,
+      "loss": 0.8684,
+      "step": 37300
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.082017142857144e-06,
+      "loss": 1.3152,
+      "step": 37301
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.081748571428572e-06,
+      "loss": 1.3523,
+      "step": 37302
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.08148e-06,
+      "loss": 0.7508,
+      "step": 37303
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.081211428571429e-06,
+      "loss": 0.9718,
+      "step": 37304
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.080942857142857e-06,
+      "loss": 0.8764,
+      "step": 37305
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0806742857142855e-06,
+      "loss": 1.1318,
+      "step": 37306
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0804057142857145e-06,
+      "loss": 1.307,
+      "step": 37307
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.080137142857143e-06,
+      "loss": 1.459,
+      "step": 37308
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.079868571428572e-06,
+      "loss": 0.924,
+      "step": 37309
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0796e-06,
+      "loss": 1.2666,
+      "step": 37310
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.079331428571428e-06,
+      "loss": 0.9036,
+      "step": 37311
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.079062857142857e-06,
+      "loss": 1.1721,
+      "step": 37312
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.078794285714286e-06,
+      "loss": 1.1544,
+      "step": 37313
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.078525714285715e-06,
+      "loss": 1.524,
+      "step": 37314
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.078257142857143e-06,
+      "loss": 1.3006,
+      "step": 37315
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0779885714285715e-06,
+      "loss": 0.8942,
+      "step": 37316
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0777200000000005e-06,
+      "loss": 1.1974,
+      "step": 37317
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.077451428571429e-06,
+      "loss": 0.8299,
+      "step": 37318
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.077182857142858e-06,
+      "loss": 1.322,
+      "step": 37319
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.076914285714286e-06,
+      "loss": 0.9876,
+      "step": 37320
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.076645714285714e-06,
+      "loss": 1.1556,
+      "step": 37321
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.076377142857143e-06,
+      "loss": 1.2026,
+      "step": 37322
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.076108571428571e-06,
+      "loss": 0.8557,
+      "step": 37323
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.07584e-06,
+      "loss": 1.1824,
+      "step": 37324
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0755714285714284e-06,
+      "loss": 1.1395,
+      "step": 37325
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0753028571428575e-06,
+      "loss": 1.3272,
+      "step": 37326
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0750342857142865e-06,
+      "loss": 0.886,
+      "step": 37327
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.074765714285715e-06,
+      "loss": 0.6938,
+      "step": 37328
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.074497142857143e-06,
+      "loss": 1.469,
+      "step": 37329
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.074228571428572e-06,
+      "loss": 0.9777,
+      "step": 37330
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.07396e-06,
+      "loss": 0.9096,
+      "step": 37331
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.073691428571429e-06,
+      "loss": 1.264,
+      "step": 37332
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.073422857142857e-06,
+      "loss": 1.1027,
+      "step": 37333
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.073154285714285e-06,
+      "loss": 1.2967,
+      "step": 37334
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0728857142857144e-06,
+      "loss": 1.3199,
+      "step": 37335
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.072617142857143e-06,
+      "loss": 1.2862,
+      "step": 37336
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.072348571428572e-06,
+      "loss": 1.3899,
+      "step": 37337
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.07208e-06,
+      "loss": 1.2948,
+      "step": 37338
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.071811428571429e-06,
+      "loss": 1.1412,
+      "step": 37339
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.071542857142858e-06,
+      "loss": 0.9788,
+      "step": 37340
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.071274285714286e-06,
+      "loss": 0.8539,
+      "step": 37341
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.071005714285715e-06,
+      "loss": 1.3349,
+      "step": 37342
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.070737142857143e-06,
+      "loss": 1.2383,
+      "step": 37343
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.070468571428571e-06,
+      "loss": 1.4754,
+      "step": 37344
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0702000000000004e-06,
+      "loss": 1.4135,
+      "step": 37345
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.069931428571429e-06,
+      "loss": 1.3841,
+      "step": 37346
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.069662857142857e-06,
+      "loss": 1.549,
+      "step": 37347
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.069394285714286e-06,
+      "loss": 0.9991,
+      "step": 37348
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.069125714285714e-06,
+      "loss": 1.2665,
+      "step": 37349
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.068857142857143e-06,
+      "loss": 1.0523,
+      "step": 37350
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.068588571428571e-06,
+      "loss": 1.3362,
+      "step": 37351
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.06832e-06,
+      "loss": 0.7049,
+      "step": 37352
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.068051428571429e-06,
+      "loss": 0.7919,
+      "step": 37353
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.067782857142857e-06,
+      "loss": 0.8718,
+      "step": 37354
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.0675142857142864e-06,
+      "loss": 0.8046,
+      "step": 37355
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.067245714285715e-06,
+      "loss": 1.2348,
+      "step": 37356
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.066977142857143e-06,
+      "loss": 1.3366,
+      "step": 37357
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.066708571428572e-06,
+      "loss": 0.9703,
+      "step": 37358
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.06644e-06,
+      "loss": 1.319,
+      "step": 37359
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.066171428571429e-06,
+      "loss": 1.4393,
+      "step": 37360
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.065902857142857e-06,
+      "loss": 1.77,
+      "step": 37361
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.065634285714285e-06,
+      "loss": 1.1407,
+      "step": 37362
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.065365714285714e-06,
+      "loss": 0.7967,
+      "step": 37363
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0650971428571426e-06,
+      "loss": 1.1555,
+      "step": 37364
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.064828571428572e-06,
+      "loss": 0.7854,
+      "step": 37365
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.064560000000001e-06,
+      "loss": 1.154,
+      "step": 37366
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.064291428571429e-06,
+      "loss": 1.3995,
+      "step": 37367
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.064022857142858e-06,
+      "loss": 0.7711,
+      "step": 37368
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.063754285714286e-06,
+      "loss": 0.808,
+      "step": 37369
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.063485714285714e-06,
+      "loss": 0.79,
+      "step": 37370
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.063217142857143e-06,
+      "loss": 1.2799,
+      "step": 37371
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.062948571428571e-06,
+      "loss": 1.2576,
+      "step": 37372
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.06268e-06,
+      "loss": 0.8238,
+      "step": 37373
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0624114285714286e-06,
+      "loss": 0.7418,
+      "step": 37374
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.062142857142857e-06,
+      "loss": 1.3086,
+      "step": 37375
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.061874285714286e-06,
+      "loss": 1.0373,
+      "step": 37376
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.061605714285714e-06,
+      "loss": 0.9237,
+      "step": 37377
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.061337142857143e-06,
+      "loss": 1.0432,
+      "step": 37378
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.061068571428572e-06,
+      "loss": 1.2126,
+      "step": 37379
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0608e-06,
+      "loss": 0.9917,
+      "step": 37380
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.060531428571429e-06,
+      "loss": 1.2942,
+      "step": 37381
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.060262857142857e-06,
+      "loss": 0.9176,
+      "step": 37382
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.059994285714286e-06,
+      "loss": 0.9092,
+      "step": 37383
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0597257142857146e-06,
+      "loss": 1.1028,
+      "step": 37384
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.059457142857143e-06,
+      "loss": 0.8152,
+      "step": 37385
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.059188571428572e-06,
+      "loss": 1.7248,
+      "step": 37386
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.05892e-06,
+      "loss": 0.8334,
+      "step": 37387
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.058651428571429e-06,
+      "loss": 1.5639,
+      "step": 37388
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.058382857142857e-06,
+      "loss": 1.2752,
+      "step": 37389
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.058114285714285e-06,
+      "loss": 0.9778,
+      "step": 37390
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.057845714285714e-06,
+      "loss": 1.0492,
+      "step": 37391
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.057577142857143e-06,
+      "loss": 0.7141,
+      "step": 37392
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0573085714285715e-06,
+      "loss": 1.1963,
+      "step": 37393
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0570400000000005e-06,
+      "loss": 1.2671,
+      "step": 37394
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.056771428571429e-06,
+      "loss": 1.3083,
+      "step": 37395
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.056502857142858e-06,
+      "loss": 1.4807,
+      "step": 37396
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.056234285714286e-06,
+      "loss": 1.2098,
+      "step": 37397
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.055965714285714e-06,
+      "loss": 0.9353,
+      "step": 37398
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.055697142857143e-06,
+      "loss": 1.2132,
+      "step": 37399
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.055428571428571e-06,
+      "loss": 1.0643,
+      "step": 37400
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.05516e-06,
+      "loss": 1.3867,
+      "step": 37401
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0548914285714285e-06,
+      "loss": 1.4345,
+      "step": 37402
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.054622857142857e-06,
+      "loss": 0.7879,
+      "step": 37403
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.054354285714286e-06,
+      "loss": 1.0394,
+      "step": 37404
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.054085714285715e-06,
+      "loss": 1.1987,
+      "step": 37405
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.053817142857144e-06,
+      "loss": 1.0635,
+      "step": 37406
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.053548571428572e-06,
+      "loss": 0.8146,
+      "step": 37407
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.05328e-06,
+      "loss": 0.7767,
+      "step": 37408
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.053011428571429e-06,
+      "loss": 1.217,
+      "step": 37409
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.052742857142857e-06,
+      "loss": 1.0209,
+      "step": 37410
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.052474285714286e-06,
+      "loss": 1.1668,
+      "step": 37411
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0522057142857145e-06,
+      "loss": 1.041,
+      "step": 37412
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.051937142857143e-06,
+      "loss": 1.4978,
+      "step": 37413
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.051668571428572e-06,
+      "loss": 1.0602,
+      "step": 37414
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0514e-06,
+      "loss": 0.7275,
+      "step": 37415
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.051131428571428e-06,
+      "loss": 1.1487,
+      "step": 37416
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.050862857142857e-06,
+      "loss": 1.4311,
+      "step": 37417
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.050594285714286e-06,
+      "loss": 0.9848,
+      "step": 37418
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.050325714285715e-06,
+      "loss": 1.7154,
+      "step": 37419
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.050057142857143e-06,
+      "loss": 0.9339,
+      "step": 37420
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0497885714285715e-06,
+      "loss": 0.8848,
+      "step": 37421
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0495200000000005e-06,
+      "loss": 0.9633,
+      "step": 37422
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.049251428571429e-06,
+      "loss": 0.6858,
+      "step": 37423
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.048982857142858e-06,
+      "loss": 1.3023,
+      "step": 37424
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.048714285714286e-06,
+      "loss": 1.2187,
+      "step": 37425
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.048445714285714e-06,
+      "loss": 0.8046,
+      "step": 37426
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.048177142857143e-06,
+      "loss": 1.1889,
+      "step": 37427
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.047908571428571e-06,
+      "loss": 1.0213,
+      "step": 37428
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.04764e-06,
+      "loss": 0.9087,
+      "step": 37429
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0473714285714284e-06,
+      "loss": 1.1179,
+      "step": 37430
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0471028571428575e-06,
+      "loss": 0.8778,
+      "step": 37431
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0468342857142865e-06,
+      "loss": 1.6249,
+      "step": 37432
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.046565714285715e-06,
+      "loss": 1.2004,
+      "step": 37433
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.046297142857143e-06,
+      "loss": 1.3694,
+      "step": 37434
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.046028571428572e-06,
+      "loss": 1.4235,
+      "step": 37435
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.04576e-06,
+      "loss": 1.0453,
+      "step": 37436
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.045491428571429e-06,
+      "loss": 1.5762,
+      "step": 37437
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.045222857142857e-06,
+      "loss": 1.561,
+      "step": 37438
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.044954285714285e-06,
+      "loss": 0.6352,
+      "step": 37439
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0446857142857144e-06,
+      "loss": 0.9675,
+      "step": 37440
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.044417142857143e-06,
+      "loss": 0.947,
+      "step": 37441
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.044148571428572e-06,
+      "loss": 0.9491,
+      "step": 37442
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.04388e-06,
+      "loss": 1.4004,
+      "step": 37443
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.043611428571428e-06,
+      "loss": 0.9036,
+      "step": 37444
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.043342857142858e-06,
+      "loss": 1.406,
+      "step": 37445
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.043074285714286e-06,
+      "loss": 1.265,
+      "step": 37446
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.042805714285715e-06,
+      "loss": 1.165,
+      "step": 37447
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.042537142857143e-06,
+      "loss": 1.3626,
+      "step": 37448
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.042268571428571e-06,
+      "loss": 1.1944,
+      "step": 37449
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0420000000000004e-06,
+      "loss": 1.3955,
+      "step": 37450
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.041731428571429e-06,
+      "loss": 0.8956,
+      "step": 37451
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.041462857142858e-06,
+      "loss": 1.1088,
+      "step": 37452
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.041194285714286e-06,
+      "loss": 1.3668,
+      "step": 37453
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.040925714285714e-06,
+      "loss": 1.2674,
+      "step": 37454
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.040657142857143e-06,
+      "loss": 1.3104,
+      "step": 37455
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.040388571428571e-06,
+      "loss": 0.9187,
+      "step": 37456
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.040119999999999e-06,
+      "loss": 0.6861,
+      "step": 37457
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.039851428571429e-06,
+      "loss": 1.2345,
+      "step": 37458
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.039582857142857e-06,
+      "loss": 0.8061,
+      "step": 37459
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0393142857142864e-06,
+      "loss": 1.0997,
+      "step": 37460
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.039045714285715e-06,
+      "loss": 0.9976,
+      "step": 37461
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.038777142857143e-06,
+      "loss": 1.0831,
+      "step": 37462
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.038508571428572e-06,
+      "loss": 0.9179,
+      "step": 37463
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.03824e-06,
+      "loss": 1.2198,
+      "step": 37464
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.037971428571429e-06,
+      "loss": 0.8646,
+      "step": 37465
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.037702857142857e-06,
+      "loss": 1.0212,
+      "step": 37466
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.037434285714285e-06,
+      "loss": 0.686,
+      "step": 37467
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.037165714285714e-06,
+      "loss": 1.0439,
+      "step": 37468
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0368971428571426e-06,
+      "loss": 1.2703,
+      "step": 37469
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.036628571428572e-06,
+      "loss": 1.1937,
+      "step": 37470
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.036360000000001e-06,
+      "loss": 1.1696,
+      "step": 37471
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.036091428571429e-06,
+      "loss": 1.0544,
+      "step": 37472
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.035822857142858e-06,
+      "loss": 1.4472,
+      "step": 37473
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.035554285714286e-06,
+      "loss": 1.4688,
+      "step": 37474
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.035285714285715e-06,
+      "loss": 0.882,
+      "step": 37475
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.035017142857143e-06,
+      "loss": 1.0604,
+      "step": 37476
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.034748571428571e-06,
+      "loss": 0.746,
+      "step": 37477
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.03448e-06,
+      "loss": 1.65,
+      "step": 37478
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0342114285714286e-06,
+      "loss": 1.5213,
+      "step": 37479
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.033942857142857e-06,
+      "loss": 1.5246,
+      "step": 37480
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.033674285714286e-06,
+      "loss": 1.4335,
+      "step": 37481
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.033405714285714e-06,
+      "loss": 1.285,
+      "step": 37482
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.033137142857143e-06,
+      "loss": 0.8976,
+      "step": 37483
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.032868571428572e-06,
+      "loss": 1.1177,
+      "step": 37484
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0326e-06,
+      "loss": 1.3892,
+      "step": 37485
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.032331428571429e-06,
+      "loss": 0.9748,
+      "step": 37486
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.032062857142857e-06,
+      "loss": 1.1508,
+      "step": 37487
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.031794285714286e-06,
+      "loss": 1.2629,
+      "step": 37488
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0315257142857145e-06,
+      "loss": 0.9318,
+      "step": 37489
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.031257142857143e-06,
+      "loss": 1.2797,
+      "step": 37490
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.030988571428572e-06,
+      "loss": 0.9807,
+      "step": 37491
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.03072e-06,
+      "loss": 1.4362,
+      "step": 37492
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.030451428571429e-06,
+      "loss": 1.0896,
+      "step": 37493
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.030182857142857e-06,
+      "loss": 1.3017,
+      "step": 37494
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.029914285714285e-06,
+      "loss": 1.3947,
+      "step": 37495
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.029645714285714e-06,
+      "loss": 0.8684,
+      "step": 37496
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0293771428571425e-06,
+      "loss": 1.2324,
+      "step": 37497
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.029108571428572e-06,
+      "loss": 1.0075,
+      "step": 37498
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0288400000000005e-06,
+      "loss": 1.2701,
+      "step": 37499
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.028571428571429e-06,
+      "loss": 1.3383,
+      "step": 37500
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.028302857142858e-06,
+      "loss": 0.8177,
+      "step": 37501
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.028034285714286e-06,
+      "loss": 1.3773,
+      "step": 37502
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.027765714285714e-06,
+      "loss": 1.3295,
+      "step": 37503
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.027497142857143e-06,
+      "loss": 1.2025,
+      "step": 37504
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.027228571428571e-06,
+      "loss": 1.191,
+      "step": 37505
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.02696e-06,
+      "loss": 1.2729,
+      "step": 37506
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0266914285714285e-06,
+      "loss": 0.693,
+      "step": 37507
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.026422857142857e-06,
+      "loss": 1.4665,
+      "step": 37508
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.026154285714286e-06,
+      "loss": 1.0777,
+      "step": 37509
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.025885714285714e-06,
+      "loss": 0.5893,
+      "step": 37510
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.025617142857144e-06,
+      "loss": 0.967,
+      "step": 37511
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.025348571428572e-06,
+      "loss": 1.0813,
+      "step": 37512
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.02508e-06,
+      "loss": 1.451,
+      "step": 37513
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.024811428571429e-06,
+      "loss": 1.1647,
+      "step": 37514
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.024542857142857e-06,
+      "loss": 0.9405,
+      "step": 37515
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.024274285714286e-06,
+      "loss": 0.5856,
+      "step": 37516
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0240057142857145e-06,
+      "loss": 1.1256,
+      "step": 37517
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.023737142857143e-06,
+      "loss": 1.2905,
+      "step": 37518
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.023468571428572e-06,
+      "loss": 1.22,
+      "step": 37519
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0232e-06,
+      "loss": 1.4458,
+      "step": 37520
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.022931428571428e-06,
+      "loss": 1.3028,
+      "step": 37521
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.022662857142857e-06,
+      "loss": 1.0367,
+      "step": 37522
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.022394285714285e-06,
+      "loss": 1.0847,
+      "step": 37523
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.022125714285715e-06,
+      "loss": 1.1391,
+      "step": 37524
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.021857142857143e-06,
+      "loss": 1.3788,
+      "step": 37525
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0215885714285715e-06,
+      "loss": 1.0126,
+      "step": 37526
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0213200000000005e-06,
+      "loss": 1.644,
+      "step": 37527
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.021051428571429e-06,
+      "loss": 1.434,
+      "step": 37528
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.020782857142858e-06,
+      "loss": 1.1623,
+      "step": 37529
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.020514285714286e-06,
+      "loss": 1.1891,
+      "step": 37530
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.020245714285714e-06,
+      "loss": 1.2807,
+      "step": 37531
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.019977142857143e-06,
+      "loss": 1.4137,
+      "step": 37532
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.019708571428571e-06,
+      "loss": 1.1432,
+      "step": 37533
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.01944e-06,
+      "loss": 1.4399,
+      "step": 37534
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0191714285714284e-06,
+      "loss": 1.1652,
+      "step": 37535
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.018902857142857e-06,
+      "loss": 1.0902,
+      "step": 37536
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.0186342857142865e-06,
+      "loss": 1.6419,
+      "step": 37537
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.018365714285715e-06,
+      "loss": 1.2374,
+      "step": 37538
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.018097142857144e-06,
+      "loss": 1.1246,
+      "step": 37539
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.017828571428572e-06,
+      "loss": 0.9619,
+      "step": 37540
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.01756e-06,
+      "loss": 1.0754,
+      "step": 37541
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.017291428571429e-06,
+      "loss": 0.7948,
+      "step": 37542
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.017022857142857e-06,
+      "loss": 1.6234,
+      "step": 37543
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.016754285714285e-06,
+      "loss": 0.8674,
+      "step": 37544
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0164857142857144e-06,
+      "loss": 1.1412,
+      "step": 37545
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.016217142857143e-06,
+      "loss": 1.5918,
+      "step": 37546
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.015948571428572e-06,
+      "loss": 1.5687,
+      "step": 37547
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.01568e-06,
+      "loss": 1.0619,
+      "step": 37548
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.015411428571428e-06,
+      "loss": 1.2835,
+      "step": 37549
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.015142857142857e-06,
+      "loss": 1.2191,
+      "step": 37550
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.014874285714286e-06,
+      "loss": 1.1785,
+      "step": 37551
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.014605714285715e-06,
+      "loss": 1.1331,
+      "step": 37552
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.014337142857143e-06,
+      "loss": 1.033,
+      "step": 37553
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.014068571428571e-06,
+      "loss": 1.4509,
+      "step": 37554
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0138000000000004e-06,
+      "loss": 1.0146,
+      "step": 37555
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.013531428571429e-06,
+      "loss": 1.2199,
+      "step": 37556
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.013262857142858e-06,
+      "loss": 0.8706,
+      "step": 37557
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.012994285714286e-06,
+      "loss": 0.8565,
+      "step": 37558
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.012725714285714e-06,
+      "loss": 1.1564,
+      "step": 37559
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.012457142857143e-06,
+      "loss": 0.7338,
+      "step": 37560
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.012188571428571e-06,
+      "loss": 1.367,
+      "step": 37561
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.01192e-06,
+      "loss": 0.9401,
+      "step": 37562
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.011651428571428e-06,
+      "loss": 1.0051,
+      "step": 37563
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.011382857142857e-06,
+      "loss": 1.2546,
+      "step": 37564
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0111142857142864e-06,
+      "loss": 1.2307,
+      "step": 37565
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.010845714285715e-06,
+      "loss": 1.1969,
+      "step": 37566
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.010577142857143e-06,
+      "loss": 1.5653,
+      "step": 37567
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.010308571428572e-06,
+      "loss": 1.6277,
+      "step": 37568
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.01004e-06,
+      "loss": 1.2022,
+      "step": 37569
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.009771428571429e-06,
+      "loss": 1.3819,
+      "step": 37570
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.009502857142857e-06,
+      "loss": 1.2341,
+      "step": 37571
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.009234285714285e-06,
+      "loss": 1.1262,
+      "step": 37572
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.008965714285714e-06,
+      "loss": 1.0838,
+      "step": 37573
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0086971428571426e-06,
+      "loss": 1.2216,
+      "step": 37574
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.008428571428572e-06,
+      "loss": 1.0432,
+      "step": 37575
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.00816e-06,
+      "loss": 1.4584,
+      "step": 37576
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.007891428571429e-06,
+      "loss": 1.0597,
+      "step": 37577
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.007622857142858e-06,
+      "loss": 1.1171,
+      "step": 37578
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.007354285714286e-06,
+      "loss": 1.3005,
+      "step": 37579
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.007085714285715e-06,
+      "loss": 0.9716,
+      "step": 37580
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.006817142857143e-06,
+      "loss": 1.6964,
+      "step": 37581
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.006548571428571e-06,
+      "loss": 1.0355,
+      "step": 37582
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.00628e-06,
+      "loss": 1.4601,
+      "step": 37583
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0060114285714286e-06,
+      "loss": 1.1115,
+      "step": 37584
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0057428571428576e-06,
+      "loss": 1.2001,
+      "step": 37585
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.005474285714286e-06,
+      "loss": 1.5407,
+      "step": 37586
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.005205714285714e-06,
+      "loss": 1.1914,
+      "step": 37587
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.004937142857143e-06,
+      "loss": 1.1085,
+      "step": 37588
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.004668571428571e-06,
+      "loss": 1.4552,
+      "step": 37589
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0044e-06,
+      "loss": 1.1984,
+      "step": 37590
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.004131428571429e-06,
+      "loss": 0.9398,
+      "step": 37591
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.003862857142857e-06,
+      "loss": 1.1987,
+      "step": 37592
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.003594285714286e-06,
+      "loss": 1.3702,
+      "step": 37593
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0033257142857145e-06,
+      "loss": 1.1071,
+      "step": 37594
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.003057142857143e-06,
+      "loss": 1.2956,
+      "step": 37595
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.002788571428572e-06,
+      "loss": 1.0775,
+      "step": 37596
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.00252e-06,
+      "loss": 0.9537,
+      "step": 37597
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.002251428571429e-06,
+      "loss": 0.9508,
+      "step": 37598
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.001982857142857e-06,
+      "loss": 1.1605,
+      "step": 37599
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.001714285714285e-06,
+      "loss": 1.371,
+      "step": 37600
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.001445714285714e-06,
+      "loss": 0.8397,
+      "step": 37601
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0011771428571425e-06,
+      "loss": 1.0899,
+      "step": 37602
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0009085714285715e-06,
+      "loss": 0.9906,
+      "step": 37603
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.0006400000000005e-06,
+      "loss": 1.2878,
+      "step": 37604
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.000371428571429e-06,
+      "loss": 1.4624,
+      "step": 37605
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.000102857142858e-06,
+      "loss": 1.2456,
+      "step": 37606
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.999834285714286e-06,
+      "loss": 1.1089,
+      "step": 37607
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.999565714285714e-06,
+      "loss": 1.3117,
+      "step": 37608
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.999297142857143e-06,
+      "loss": 1.1424,
+      "step": 37609
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.999028571428571e-06,
+      "loss": 1.3772,
+      "step": 37610
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.99876e-06,
+      "loss": 1.3031,
+      "step": 37611
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9984914285714285e-06,
+      "loss": 1.1219,
+      "step": 37612
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.998222857142857e-06,
+      "loss": 1.1022,
+      "step": 37613
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.997954285714286e-06,
+      "loss": 1.1008,
+      "step": 37614
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.997685714285714e-06,
+      "loss": 1.2307,
+      "step": 37615
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.997417142857143e-06,
+      "loss": 1.523,
+      "step": 37616
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.997148571428572e-06,
+      "loss": 1.4647,
+      "step": 37617
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.99688e-06,
+      "loss": 0.9412,
+      "step": 37618
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.996611428571429e-06,
+      "loss": 1.3178,
+      "step": 37619
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.996342857142857e-06,
+      "loss": 1.0432,
+      "step": 37620
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.996074285714286e-06,
+      "loss": 1.4759,
+      "step": 37621
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9958057142857145e-06,
+      "loss": 1.3045,
+      "step": 37622
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.995537142857143e-06,
+      "loss": 1.2272,
+      "step": 37623
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.995268571428572e-06,
+      "loss": 0.6616,
+      "step": 37624
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.995e-06,
+      "loss": 1.0281,
+      "step": 37625
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.994731428571429e-06,
+      "loss": 1.0736,
+      "step": 37626
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.994462857142857e-06,
+      "loss": 0.8672,
+      "step": 37627
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.994194285714285e-06,
+      "loss": 0.9965,
+      "step": 37628
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.993925714285714e-06,
+      "loss": 1.2645,
+      "step": 37629
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.993657142857143e-06,
+      "loss": 0.7624,
+      "step": 37630
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9933885714285715e-06,
+      "loss": 1.5355,
+      "step": 37631
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9931200000000005e-06,
+      "loss": 1.1318,
+      "step": 37632
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.992851428571429e-06,
+      "loss": 1.4859,
+      "step": 37633
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.992582857142858e-06,
+      "loss": 1.2649,
+      "step": 37634
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.992314285714286e-06,
+      "loss": 1.233,
+      "step": 37635
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.992045714285714e-06,
+      "loss": 1.3284,
+      "step": 37636
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.991777142857143e-06,
+      "loss": 0.8289,
+      "step": 37637
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.991508571428571e-06,
+      "loss": 1.0902,
+      "step": 37638
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.99124e-06,
+      "loss": 0.8815,
+      "step": 37639
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9909714285714284e-06,
+      "loss": 1.0752,
+      "step": 37640
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.990702857142857e-06,
+      "loss": 0.8984,
+      "step": 37641
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.990434285714286e-06,
+      "loss": 0.8254,
+      "step": 37642
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.990165714285715e-06,
+      "loss": 0.6511,
+      "step": 37643
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.989897142857144e-06,
+      "loss": 1.2833,
+      "step": 37644
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.989628571428572e-06,
+      "loss": 1.1068,
+      "step": 37645
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.98936e-06,
+      "loss": 1.551,
+      "step": 37646
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.989091428571429e-06,
+      "loss": 1.3581,
+      "step": 37647
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.988822857142857e-06,
+      "loss": 1.0571,
+      "step": 37648
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.988554285714286e-06,
+      "loss": 1.0855,
+      "step": 37649
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9882857142857144e-06,
+      "loss": 1.0621,
+      "step": 37650
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.988017142857143e-06,
+      "loss": 1.2351,
+      "step": 37651
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.987748571428572e-06,
+      "loss": 1.2335,
+      "step": 37652
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.98748e-06,
+      "loss": 1.2025,
+      "step": 37653
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.987211428571428e-06,
+      "loss": 0.814,
+      "step": 37654
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.986942857142857e-06,
+      "loss": 1.3487,
+      "step": 37655
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.986674285714286e-06,
+      "loss": 1.065,
+      "step": 37656
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.986405714285715e-06,
+      "loss": 0.9041,
+      "step": 37657
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.986137142857143e-06,
+      "loss": 1.2755,
+      "step": 37658
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.985868571428571e-06,
+      "loss": 1.1148,
+      "step": 37659
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9856000000000004e-06,
+      "loss": 0.8332,
+      "step": 37660
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.985331428571429e-06,
+      "loss": 1.4977,
+      "step": 37661
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.985062857142858e-06,
+      "loss": 1.0639,
+      "step": 37662
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.984794285714286e-06,
+      "loss": 0.9935,
+      "step": 37663
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.984525714285714e-06,
+      "loss": 1.042,
+      "step": 37664
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.984257142857143e-06,
+      "loss": 1.6069,
+      "step": 37665
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.983988571428571e-06,
+      "loss": 1.5159,
+      "step": 37666
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.98372e-06,
+      "loss": 1.1239,
+      "step": 37667
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.983451428571428e-06,
+      "loss": 1.5413,
+      "step": 37668
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.983182857142857e-06,
+      "loss": 1.0106,
+      "step": 37669
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9829142857142864e-06,
+      "loss": 1.3476,
+      "step": 37670
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.982645714285715e-06,
+      "loss": 1.1324,
+      "step": 37671
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.982377142857144e-06,
+      "loss": 1.2384,
+      "step": 37672
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.982108571428572e-06,
+      "loss": 1.1357,
+      "step": 37673
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.98184e-06,
+      "loss": 1.4708,
+      "step": 37674
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.981571428571429e-06,
+      "loss": 1.2993,
+      "step": 37675
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.981302857142857e-06,
+      "loss": 1.4628,
+      "step": 37676
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.981034285714285e-06,
+      "loss": 1.4687,
+      "step": 37677
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.980765714285714e-06,
+      "loss": 1.3952,
+      "step": 37678
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9804971428571426e-06,
+      "loss": 1.3343,
+      "step": 37679
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.980228571428572e-06,
+      "loss": 0.897,
+      "step": 37680
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.97996e-06,
+      "loss": 1.2784,
+      "step": 37681
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.979691428571429e-06,
+      "loss": 1.5155,
+      "step": 37682
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.979422857142858e-06,
+      "loss": 1.4696,
+      "step": 37683
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.979154285714286e-06,
+      "loss": 1.1801,
+      "step": 37684
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.978885714285715e-06,
+      "loss": 1.1927,
+      "step": 37685
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.978617142857143e-06,
+      "loss": 1.5529,
+      "step": 37686
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.978348571428571e-06,
+      "loss": 1.3662,
+      "step": 37687
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.97808e-06,
+      "loss": 1.1676,
+      "step": 37688
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9778114285714286e-06,
+      "loss": 1.2967,
+      "step": 37689
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9775428571428576e-06,
+      "loss": 1.2768,
+      "step": 37690
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.977274285714286e-06,
+      "loss": 1.1161,
+      "step": 37691
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.977005714285714e-06,
+      "loss": 1.2684,
+      "step": 37692
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.976737142857143e-06,
+      "loss": 1.5908,
+      "step": 37693
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.976468571428571e-06,
+      "loss": 1.0488,
+      "step": 37694
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9762e-06,
+      "loss": 0.8713,
+      "step": 37695
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.975931428571429e-06,
+      "loss": 1.2039,
+      "step": 37696
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.975662857142857e-06,
+      "loss": 1.4223,
+      "step": 37697
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.975394285714286e-06,
+      "loss": 1.1456,
+      "step": 37698
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9751257142857145e-06,
+      "loss": 0.928,
+      "step": 37699
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.974857142857143e-06,
+      "loss": 1.1114,
+      "step": 37700
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.974588571428572e-06,
+      "loss": 1.251,
+      "step": 37701
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.97432e-06,
+      "loss": 1.3387,
+      "step": 37702
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.974051428571429e-06,
+      "loss": 1.2457,
+      "step": 37703
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.973782857142857e-06,
+      "loss": 1.3065,
+      "step": 37704
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.973514285714285e-06,
+      "loss": 1.0227,
+      "step": 37705
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.973245714285714e-06,
+      "loss": 1.2749,
+      "step": 37706
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9729771428571425e-06,
+      "loss": 1.1234,
+      "step": 37707
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9727085714285715e-06,
+      "loss": 1.2664,
+      "step": 37708
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.9724400000000005e-06,
+      "loss": 1.344,
+      "step": 37709
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.972171428571429e-06,
+      "loss": 1.3547,
+      "step": 37710
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.971902857142858e-06,
+      "loss": 1.0769,
+      "step": 37711
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.971634285714286e-06,
+      "loss": 1.3607,
+      "step": 37712
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.971365714285715e-06,
+      "loss": 1.3862,
+      "step": 37713
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.971097142857143e-06,
+      "loss": 0.7865,
+      "step": 37714
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.970828571428571e-06,
+      "loss": 1.0618,
+      "step": 37715
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.97056e-06,
+      "loss": 1.3693,
+      "step": 37716
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9702914285714285e-06,
+      "loss": 0.7893,
+      "step": 37717
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.970022857142857e-06,
+      "loss": 1.0799,
+      "step": 37718
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.969754285714286e-06,
+      "loss": 1.0648,
+      "step": 37719
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.969485714285714e-06,
+      "loss": 1.1913,
+      "step": 37720
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.969217142857143e-06,
+      "loss": 0.9674,
+      "step": 37721
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.968948571428572e-06,
+      "loss": 1.4551,
+      "step": 37722
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.96868e-06,
+      "loss": 1.2186,
+      "step": 37723
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.968411428571429e-06,
+      "loss": 1.2673,
+      "step": 37724
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.968142857142857e-06,
+      "loss": 1.4016,
+      "step": 37725
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.967874285714286e-06,
+      "loss": 0.8544,
+      "step": 37726
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9676057142857145e-06,
+      "loss": 1.3238,
+      "step": 37727
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.967337142857143e-06,
+      "loss": 0.9506,
+      "step": 37728
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.967068571428572e-06,
+      "loss": 1.1614,
+      "step": 37729
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9668e-06,
+      "loss": 1.3564,
+      "step": 37730
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.966531428571429e-06,
+      "loss": 1.0358,
+      "step": 37731
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.966262857142857e-06,
+      "loss": 1.2643,
+      "step": 37732
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.965994285714285e-06,
+      "loss": 0.9843,
+      "step": 37733
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.965725714285714e-06,
+      "loss": 1.0872,
+      "step": 37734
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.965457142857143e-06,
+      "loss": 1.4544,
+      "step": 37735
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.965188571428572e-06,
+      "loss": 1.1308,
+      "step": 37736
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9649200000000005e-06,
+      "loss": 1.3636,
+      "step": 37737
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.964651428571429e-06,
+      "loss": 1.4994,
+      "step": 37738
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.964382857142858e-06,
+      "loss": 0.9388,
+      "step": 37739
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.964114285714286e-06,
+      "loss": 0.9633,
+      "step": 37740
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.963845714285714e-06,
+      "loss": 1.242,
+      "step": 37741
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.963577142857143e-06,
+      "loss": 1.1813,
+      "step": 37742
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.963308571428571e-06,
+      "loss": 1.097,
+      "step": 37743
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.96304e-06,
+      "loss": 1.2437,
+      "step": 37744
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9627714285714284e-06,
+      "loss": 0.6116,
+      "step": 37745
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.962502857142857e-06,
+      "loss": 1.0312,
+      "step": 37746
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.962234285714286e-06,
+      "loss": 1.5537,
+      "step": 37747
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.961965714285715e-06,
+      "loss": 0.9754,
+      "step": 37748
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.961697142857144e-06,
+      "loss": 1.0335,
+      "step": 37749
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.961428571428572e-06,
+      "loss": 0.8411,
+      "step": 37750
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.96116e-06,
+      "loss": 1.302,
+      "step": 37751
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.960891428571429e-06,
+      "loss": 0.7998,
+      "step": 37752
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.960622857142857e-06,
+      "loss": 1.0249,
+      "step": 37753
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.960354285714286e-06,
+      "loss": 0.7845,
+      "step": 37754
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9600857142857144e-06,
+      "loss": 0.8768,
+      "step": 37755
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.959817142857143e-06,
+      "loss": 0.9815,
+      "step": 37756
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.959548571428572e-06,
+      "loss": 1.0304,
+      "step": 37757
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.95928e-06,
+      "loss": 1.3918,
+      "step": 37758
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.959011428571428e-06,
+      "loss": 1.0981,
+      "step": 37759
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.958742857142857e-06,
+      "loss": 1.3272,
+      "step": 37760
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.958474285714286e-06,
+      "loss": 0.9854,
+      "step": 37761
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.958205714285715e-06,
+      "loss": 1.1231,
+      "step": 37762
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.957937142857143e-06,
+      "loss": 1.1949,
+      "step": 37763
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.957668571428571e-06,
+      "loss": 0.9509,
+      "step": 37764
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9574000000000004e-06,
+      "loss": 0.9133,
+      "step": 37765
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.957131428571429e-06,
+      "loss": 0.9522,
+      "step": 37766
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.956862857142858e-06,
+      "loss": 1.1188,
+      "step": 37767
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.956594285714286e-06,
+      "loss": 1.1953,
+      "step": 37768
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.956325714285714e-06,
+      "loss": 1.1646,
+      "step": 37769
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.956057142857143e-06,
+      "loss": 0.9871,
+      "step": 37770
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.955788571428571e-06,
+      "loss": 1.1479,
+      "step": 37771
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.95552e-06,
+      "loss": 1.2303,
+      "step": 37772
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.955251428571428e-06,
+      "loss": 1.1339,
+      "step": 37773
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.954982857142857e-06,
+      "loss": 0.9308,
+      "step": 37774
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9547142857142864e-06,
+      "loss": 1.2054,
+      "step": 37775
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.954445714285715e-06,
+      "loss": 1.5871,
+      "step": 37776
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.954177142857144e-06,
+      "loss": 1.4765,
+      "step": 37777
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.953908571428572e-06,
+      "loss": 0.9689,
+      "step": 37778
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.95364e-06,
+      "loss": 1.2055,
+      "step": 37779
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.953371428571429e-06,
+      "loss": 1.3667,
+      "step": 37780
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.953102857142857e-06,
+      "loss": 1.1019,
+      "step": 37781
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.952834285714285e-06,
+      "loss": 1.112,
+      "step": 37782
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.952565714285714e-06,
+      "loss": 0.911,
+      "step": 37783
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9522971428571426e-06,
+      "loss": 1.0269,
+      "step": 37784
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9520285714285716e-06,
+      "loss": 0.8658,
+      "step": 37785
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.95176e-06,
+      "loss": 0.9455,
+      "step": 37786
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.951491428571429e-06,
+      "loss": 1.0611,
+      "step": 37787
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.951222857142858e-06,
+      "loss": 1.0362,
+      "step": 37788
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.950954285714286e-06,
+      "loss": 1.3507,
+      "step": 37789
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.950685714285715e-06,
+      "loss": 1.1593,
+      "step": 37790
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.950417142857143e-06,
+      "loss": 0.9839,
+      "step": 37791
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.950148571428571e-06,
+      "loss": 1.421,
+      "step": 37792
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.94988e-06,
+      "loss": 1.1272,
+      "step": 37793
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9496114285714285e-06,
+      "loss": 0.7204,
+      "step": 37794
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9493428571428576e-06,
+      "loss": 1.021,
+      "step": 37795
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.949074285714286e-06,
+      "loss": 1.2219,
+      "step": 37796
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.948805714285714e-06,
+      "loss": 0.7829,
+      "step": 37797
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.948537142857143e-06,
+      "loss": 0.7314,
+      "step": 37798
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.948268571428571e-06,
+      "loss": 1.2047,
+      "step": 37799
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.948e-06,
+      "loss": 1.3003,
+      "step": 37800
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.947731428571429e-06,
+      "loss": 1.039,
+      "step": 37801
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.947462857142857e-06,
+      "loss": 0.6228,
+      "step": 37802
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.947194285714286e-06,
+      "loss": 1.6393,
+      "step": 37803
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9469257142857145e-06,
+      "loss": 1.0552,
+      "step": 37804
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.946657142857143e-06,
+      "loss": 1.3925,
+      "step": 37805
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.946388571428572e-06,
+      "loss": 1.3351,
+      "step": 37806
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.94612e-06,
+      "loss": 1.058,
+      "step": 37807
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.945851428571429e-06,
+      "loss": 0.9757,
+      "step": 37808
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.945582857142857e-06,
+      "loss": 1.2123,
+      "step": 37809
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.945314285714285e-06,
+      "loss": 0.8892,
+      "step": 37810
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.945045714285714e-06,
+      "loss": 1.1365,
+      "step": 37811
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9447771428571425e-06,
+      "loss": 0.8491,
+      "step": 37812
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9445085714285715e-06,
+      "loss": 0.8632,
+      "step": 37813
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9442400000000005e-06,
+      "loss": 1.128,
+      "step": 37814
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.943971428571429e-06,
+      "loss": 0.8645,
+      "step": 37815
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.943702857142858e-06,
+      "loss": 1.5055,
+      "step": 37816
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.943434285714286e-06,
+      "loss": 1.0094,
+      "step": 37817
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.943165714285715e-06,
+      "loss": 1.2129,
+      "step": 37818
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.942897142857143e-06,
+      "loss": 0.9587,
+      "step": 37819
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.942628571428571e-06,
+      "loss": 1.2559,
+      "step": 37820
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.94236e-06,
+      "loss": 1.0344,
+      "step": 37821
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9420914285714285e-06,
+      "loss": 1.4301,
+      "step": 37822
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9418228571428575e-06,
+      "loss": 0.9901,
+      "step": 37823
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.941554285714286e-06,
+      "loss": 0.8618,
+      "step": 37824
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.941285714285714e-06,
+      "loss": 1.1898,
+      "step": 37825
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.941017142857143e-06,
+      "loss": 1.4704,
+      "step": 37826
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.940748571428572e-06,
+      "loss": 1.3312,
+      "step": 37827
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.94048e-06,
+      "loss": 1.2842,
+      "step": 37828
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.940211428571429e-06,
+      "loss": 0.8181,
+      "step": 37829
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.939942857142857e-06,
+      "loss": 1.7007,
+      "step": 37830
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.939674285714286e-06,
+      "loss": 1.185,
+      "step": 37831
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9394057142857145e-06,
+      "loss": 1.2647,
+      "step": 37832
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.939137142857143e-06,
+      "loss": 0.8169,
+      "step": 37833
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.938868571428572e-06,
+      "loss": 0.948,
+      "step": 37834
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9386e-06,
+      "loss": 1.1667,
+      "step": 37835
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.938331428571429e-06,
+      "loss": 0.7783,
+      "step": 37836
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.938062857142857e-06,
+      "loss": 1.4366,
+      "step": 37837
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.937794285714285e-06,
+      "loss": 1.0985,
+      "step": 37838
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.937525714285714e-06,
+      "loss": 1.1822,
+      "step": 37839
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.937257142857143e-06,
+      "loss": 1.0281,
+      "step": 37840
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.936988571428572e-06,
+      "loss": 1.0665,
+      "step": 37841
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9367200000000005e-06,
+      "loss": 1.4452,
+      "step": 37842
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.936451428571429e-06,
+      "loss": 1.4195,
+      "step": 37843
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.936182857142858e-06,
+      "loss": 1.0642,
+      "step": 37844
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.935914285714286e-06,
+      "loss": 1.1172,
+      "step": 37845
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.935645714285714e-06,
+      "loss": 1.4916,
+      "step": 37846
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.935377142857143e-06,
+      "loss": 1.1724,
+      "step": 37847
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.935108571428571e-06,
+      "loss": 1.176,
+      "step": 37848
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.93484e-06,
+      "loss": 1.2041,
+      "step": 37849
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9345714285714284e-06,
+      "loss": 1.1968,
+      "step": 37850
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.934302857142857e-06,
+      "loss": 1.0936,
+      "step": 37851
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.934034285714286e-06,
+      "loss": 0.8355,
+      "step": 37852
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.933765714285715e-06,
+      "loss": 1.3493,
+      "step": 37853
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.933497142857144e-06,
+      "loss": 0.9172,
+      "step": 37854
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.933228571428572e-06,
+      "loss": 1.4788,
+      "step": 37855
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.93296e-06,
+      "loss": 1.3387,
+      "step": 37856
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.932691428571429e-06,
+      "loss": 1.5312,
+      "step": 37857
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.932422857142857e-06,
+      "loss": 0.7917,
+      "step": 37858
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.932154285714286e-06,
+      "loss": 1.3287,
+      "step": 37859
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9318857142857144e-06,
+      "loss": 1.2543,
+      "step": 37860
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.931617142857143e-06,
+      "loss": 0.8949,
+      "step": 37861
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.931348571428572e-06,
+      "loss": 1.3959,
+      "step": 37862
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.93108e-06,
+      "loss": 1.5863,
+      "step": 37863
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.930811428571429e-06,
+      "loss": 0.9562,
+      "step": 37864
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.930542857142857e-06,
+      "loss": 1.0782,
+      "step": 37865
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.930274285714286e-06,
+      "loss": 1.1828,
+      "step": 37866
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.930005714285715e-06,
+      "loss": 1.0754,
+      "step": 37867
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.929737142857143e-06,
+      "loss": 0.9985,
+      "step": 37868
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.929468571428571e-06,
+      "loss": 1.2534,
+      "step": 37869
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9292000000000004e-06,
+      "loss": 1.3185,
+      "step": 37870
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.928931428571429e-06,
+      "loss": 0.9964,
+      "step": 37871
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.928662857142858e-06,
+      "loss": 1.5026,
+      "step": 37872
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.928394285714286e-06,
+      "loss": 1.3269,
+      "step": 37873
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.928125714285714e-06,
+      "loss": 0.9727,
+      "step": 37874
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.927857142857143e-06,
+      "loss": 1.4513,
+      "step": 37875
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.927588571428571e-06,
+      "loss": 1.6997,
+      "step": 37876
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.92732e-06,
+      "loss": 0.7709,
+      "step": 37877
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.927051428571428e-06,
+      "loss": 1.2368,
+      "step": 37878
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.926782857142857e-06,
+      "loss": 0.9904,
+      "step": 37879
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.9265142857142864e-06,
+      "loss": 1.2752,
+      "step": 37880
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.926245714285715e-06,
+      "loss": 0.7395,
+      "step": 37881
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.925977142857144e-06,
+      "loss": 1.1036,
+      "step": 37882
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.925708571428572e-06,
+      "loss": 0.9671,
+      "step": 37883
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.92544e-06,
+      "loss": 1.4679,
+      "step": 37884
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.925171428571429e-06,
+      "loss": 0.9134,
+      "step": 37885
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.924902857142857e-06,
+      "loss": 1.4606,
+      "step": 37886
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.924634285714286e-06,
+      "loss": 1.4764,
+      "step": 37887
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.924365714285714e-06,
+      "loss": 1.0866,
+      "step": 37888
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9240971428571426e-06,
+      "loss": 1.2047,
+      "step": 37889
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9238285714285716e-06,
+      "loss": 1.1785,
+      "step": 37890
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.92356e-06,
+      "loss": 1.6094,
+      "step": 37891
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.923291428571428e-06,
+      "loss": 1.0567,
+      "step": 37892
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.923022857142858e-06,
+      "loss": 1.1246,
+      "step": 37893
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.922754285714286e-06,
+      "loss": 0.8555,
+      "step": 37894
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.922485714285715e-06,
+      "loss": 1.279,
+      "step": 37895
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.922217142857143e-06,
+      "loss": 1.2013,
+      "step": 37896
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.921948571428571e-06,
+      "loss": 1.1288,
+      "step": 37897
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.92168e-06,
+      "loss": 0.6918,
+      "step": 37898
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9214114285714285e-06,
+      "loss": 1.1702,
+      "step": 37899
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9211428571428576e-06,
+      "loss": 1.079,
+      "step": 37900
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.920874285714286e-06,
+      "loss": 1.0645,
+      "step": 37901
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.920605714285714e-06,
+      "loss": 1.1577,
+      "step": 37902
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.920337142857143e-06,
+      "loss": 1.2295,
+      "step": 37903
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.920068571428571e-06,
+      "loss": 0.8656,
+      "step": 37904
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9198e-06,
+      "loss": 1.2634,
+      "step": 37905
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.919531428571429e-06,
+      "loss": 1.2189,
+      "step": 37906
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.919262857142857e-06,
+      "loss": 1.2265,
+      "step": 37907
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.918994285714286e-06,
+      "loss": 0.9138,
+      "step": 37908
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9187257142857145e-06,
+      "loss": 0.6919,
+      "step": 37909
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9184571428571436e-06,
+      "loss": 1.3193,
+      "step": 37910
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.918188571428572e-06,
+      "loss": 1.1252,
+      "step": 37911
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.91792e-06,
+      "loss": 1.0691,
+      "step": 37912
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.917651428571429e-06,
+      "loss": 1.1253,
+      "step": 37913
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.917382857142857e-06,
+      "loss": 1.0966,
+      "step": 37914
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.917114285714285e-06,
+      "loss": 0.6121,
+      "step": 37915
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.916845714285714e-06,
+      "loss": 1.101,
+      "step": 37916
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9165771428571425e-06,
+      "loss": 1.2459,
+      "step": 37917
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9163085714285715e-06,
+      "loss": 1.0256,
+      "step": 37918
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9160400000000005e-06,
+      "loss": 0.9103,
+      "step": 37919
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.915771428571429e-06,
+      "loss": 1.285,
+      "step": 37920
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.915502857142858e-06,
+      "loss": 1.1482,
+      "step": 37921
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.915234285714286e-06,
+      "loss": 0.8809,
+      "step": 37922
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.914965714285715e-06,
+      "loss": 1.0354,
+      "step": 37923
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.914697142857143e-06,
+      "loss": 1.3174,
+      "step": 37924
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.914428571428571e-06,
+      "loss": 1.1622,
+      "step": 37925
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.91416e-06,
+      "loss": 0.8921,
+      "step": 37926
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9138914285714285e-06,
+      "loss": 0.8358,
+      "step": 37927
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9136228571428575e-06,
+      "loss": 1.0186,
+      "step": 37928
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.913354285714286e-06,
+      "loss": 1.1412,
+      "step": 37929
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.913085714285714e-06,
+      "loss": 1.3354,
+      "step": 37930
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.912817142857143e-06,
+      "loss": 1.0762,
+      "step": 37931
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.912548571428572e-06,
+      "loss": 0.9044,
+      "step": 37932
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.91228e-06,
+      "loss": 1.6026,
+      "step": 37933
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.912011428571429e-06,
+      "loss": 1.1358,
+      "step": 37934
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.911742857142857e-06,
+      "loss": 1.2051,
+      "step": 37935
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.911474285714286e-06,
+      "loss": 0.9185,
+      "step": 37936
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9112057142857145e-06,
+      "loss": 1.2943,
+      "step": 37937
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.910937142857143e-06,
+      "loss": 1.4175,
+      "step": 37938
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.910668571428572e-06,
+      "loss": 1.3228,
+      "step": 37939
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9104e-06,
+      "loss": 1.5432,
+      "step": 37940
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.910131428571429e-06,
+      "loss": 0.936,
+      "step": 37941
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.909862857142857e-06,
+      "loss": 1.1844,
+      "step": 37942
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.909594285714285e-06,
+      "loss": 1.2205,
+      "step": 37943
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.909325714285714e-06,
+      "loss": 0.6649,
+      "step": 37944
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9090571428571424e-06,
+      "loss": 1.0257,
+      "step": 37945
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.908788571428572e-06,
+      "loss": 1.5127,
+      "step": 37946
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9085200000000005e-06,
+      "loss": 1.3554,
+      "step": 37947
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.908251428571429e-06,
+      "loss": 1.2828,
+      "step": 37948
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.907982857142858e-06,
+      "loss": 0.7377,
+      "step": 37949
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.907714285714286e-06,
+      "loss": 0.9072,
+      "step": 37950
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.907445714285715e-06,
+      "loss": 1.4488,
+      "step": 37951
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.907177142857143e-06,
+      "loss": 1.0111,
+      "step": 37952
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.906908571428571e-06,
+      "loss": 0.9856,
+      "step": 37953
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.90664e-06,
+      "loss": 1.1538,
+      "step": 37954
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9063714285714284e-06,
+      "loss": 1.4208,
+      "step": 37955
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.906102857142857e-06,
+      "loss": 1.2414,
+      "step": 37956
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.905834285714286e-06,
+      "loss": 0.6472,
+      "step": 37957
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.905565714285714e-06,
+      "loss": 0.9974,
+      "step": 37958
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.905297142857144e-06,
+      "loss": 0.951,
+      "step": 37959
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.905028571428572e-06,
+      "loss": 0.9849,
+      "step": 37960
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.90476e-06,
+      "loss": 1.1839,
+      "step": 37961
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.904491428571429e-06,
+      "loss": 1.3053,
+      "step": 37962
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.904222857142857e-06,
+      "loss": 0.9913,
+      "step": 37963
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.903954285714286e-06,
+      "loss": 1.0501,
+      "step": 37964
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9036857142857144e-06,
+      "loss": 1.1454,
+      "step": 37965
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.903417142857143e-06,
+      "loss": 1.1253,
+      "step": 37966
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.903148571428572e-06,
+      "loss": 0.9644,
+      "step": 37967
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.90288e-06,
+      "loss": 0.9573,
+      "step": 37968
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.902611428571429e-06,
+      "loss": 0.8417,
+      "step": 37969
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.902342857142857e-06,
+      "loss": 0.9438,
+      "step": 37970
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.902074285714285e-06,
+      "loss": 0.9252,
+      "step": 37971
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.901805714285715e-06,
+      "loss": 1.1952,
+      "step": 37972
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.901537142857143e-06,
+      "loss": 0.7119,
+      "step": 37973
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.901268571428572e-06,
+      "loss": 1.4326,
+      "step": 37974
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9010000000000004e-06,
+      "loss": 1.049,
+      "step": 37975
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.900731428571429e-06,
+      "loss": 1.0984,
+      "step": 37976
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.900462857142858e-06,
+      "loss": 1.124,
+      "step": 37977
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.900194285714286e-06,
+      "loss": 1.4481,
+      "step": 37978
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.899925714285714e-06,
+      "loss": 0.8277,
+      "step": 37979
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.899657142857143e-06,
+      "loss": 1.268,
+      "step": 37980
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.899388571428571e-06,
+      "loss": 1.2251,
+      "step": 37981
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.89912e-06,
+      "loss": 1.2747,
+      "step": 37982
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.898851428571428e-06,
+      "loss": 0.6991,
+      "step": 37983
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8985828571428566e-06,
+      "loss": 1.0262,
+      "step": 37984
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8983142857142864e-06,
+      "loss": 1.015,
+      "step": 37985
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.898045714285715e-06,
+      "loss": 1.1451,
+      "step": 37986
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.897777142857144e-06,
+      "loss": 1.192,
+      "step": 37987
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.897508571428572e-06,
+      "loss": 0.8247,
+      "step": 37988
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.89724e-06,
+      "loss": 1.1011,
+      "step": 37989
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.896971428571429e-06,
+      "loss": 1.3826,
+      "step": 37990
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.896702857142857e-06,
+      "loss": 1.289,
+      "step": 37991
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.896434285714286e-06,
+      "loss": 0.7379,
+      "step": 37992
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.896165714285714e-06,
+      "loss": 1.4207,
+      "step": 37993
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8958971428571426e-06,
+      "loss": 0.9461,
+      "step": 37994
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8956285714285716e-06,
+      "loss": 1.0773,
+      "step": 37995
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.89536e-06,
+      "loss": 1.0029,
+      "step": 37996
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.895091428571429e-06,
+      "loss": 1.2229,
+      "step": 37997
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.894822857142857e-06,
+      "loss": 0.8764,
+      "step": 37998
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.894554285714286e-06,
+      "loss": 1.1509,
+      "step": 37999
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.894285714285715e-06,
+      "loss": 1.4357,
+      "step": 38000
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.894017142857143e-06,
+      "loss": 1.4118,
+      "step": 38001
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.893748571428571e-06,
+      "loss": 1.0649,
+      "step": 38002
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.89348e-06,
+      "loss": 0.8596,
+      "step": 38003
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8932114285714285e-06,
+      "loss": 1.1773,
+      "step": 38004
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8929428571428576e-06,
+      "loss": 1.2145,
+      "step": 38005
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.892674285714286e-06,
+      "loss": 1.1294,
+      "step": 38006
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.892405714285714e-06,
+      "loss": 1.238,
+      "step": 38007
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.892137142857143e-06,
+      "loss": 1.0536,
+      "step": 38008
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.891868571428571e-06,
+      "loss": 1.0666,
+      "step": 38009
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8916e-06,
+      "loss": 1.1381,
+      "step": 38010
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.891331428571428e-06,
+      "loss": 1.6765,
+      "step": 38011
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.891062857142857e-06,
+      "loss": 1.0446,
+      "step": 38012
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.890794285714286e-06,
+      "loss": 0.8681,
+      "step": 38013
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8905257142857145e-06,
+      "loss": 0.7501,
+      "step": 38014
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8902571428571436e-06,
+      "loss": 1.0616,
+      "step": 38015
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.889988571428572e-06,
+      "loss": 1.1172,
+      "step": 38016
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.88972e-06,
+      "loss": 1.3304,
+      "step": 38017
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.889451428571429e-06,
+      "loss": 1.5305,
+      "step": 38018
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.889182857142857e-06,
+      "loss": 1.1341,
+      "step": 38019
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.888914285714285e-06,
+      "loss": 0.6308,
+      "step": 38020
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.888645714285714e-06,
+      "loss": 1.07,
+      "step": 38021
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8883771428571425e-06,
+      "loss": 1.137,
+      "step": 38022
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8881085714285715e-06,
+      "loss": 1.1624,
+      "step": 38023
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.88784e-06,
+      "loss": 0.9235,
+      "step": 38024
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.887571428571429e-06,
+      "loss": 1.1042,
+      "step": 38025
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.887302857142858e-06,
+      "loss": 0.9943,
+      "step": 38026
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.887034285714286e-06,
+      "loss": 1.1863,
+      "step": 38027
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.886765714285715e-06,
+      "loss": 1.0439,
+      "step": 38028
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.886497142857143e-06,
+      "loss": 1.2528,
+      "step": 38029
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.886228571428571e-06,
+      "loss": 1.0911,
+      "step": 38030
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.88596e-06,
+      "loss": 1.3288,
+      "step": 38031
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8856914285714285e-06,
+      "loss": 1.1026,
+      "step": 38032
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8854228571428575e-06,
+      "loss": 1.5541,
+      "step": 38033
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.885154285714286e-06,
+      "loss": 1.3174,
+      "step": 38034
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.884885714285714e-06,
+      "loss": 1.5011,
+      "step": 38035
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.884617142857143e-06,
+      "loss": 0.8343,
+      "step": 38036
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.884348571428571e-06,
+      "loss": 1.201,
+      "step": 38037
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.884080000000001e-06,
+      "loss": 1.2681,
+      "step": 38038
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.883811428571429e-06,
+      "loss": 0.9749,
+      "step": 38039
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.883542857142857e-06,
+      "loss": 1.3286,
+      "step": 38040
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.883274285714286e-06,
+      "loss": 1.0276,
+      "step": 38041
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8830057142857145e-06,
+      "loss": 0.6451,
+      "step": 38042
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.882737142857143e-06,
+      "loss": 0.907,
+      "step": 38043
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.882468571428572e-06,
+      "loss": 1.137,
+      "step": 38044
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8822e-06,
+      "loss": 1.1805,
+      "step": 38045
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.881931428571429e-06,
+      "loss": 1.0193,
+      "step": 38046
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.881662857142857e-06,
+      "loss": 1.4477,
+      "step": 38047
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.881394285714285e-06,
+      "loss": 0.9719,
+      "step": 38048
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.881125714285714e-06,
+      "loss": 0.9084,
+      "step": 38049
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8808571428571424e-06,
+      "loss": 0.9936,
+      "step": 38050
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8805885714285715e-06,
+      "loss": 1.2357,
+      "step": 38051
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8803200000000005e-06,
+      "loss": 1.1969,
+      "step": 38052
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.880051428571429e-06,
+      "loss": 1.2599,
+      "step": 38053
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.879782857142858e-06,
+      "loss": 1.3443,
+      "step": 38054
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.879514285714286e-06,
+      "loss": 0.9173,
+      "step": 38055
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.879245714285715e-06,
+      "loss": 0.659,
+      "step": 38056
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.878977142857143e-06,
+      "loss": 1.445,
+      "step": 38057
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.878708571428571e-06,
+      "loss": 1.0984,
+      "step": 38058
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.87844e-06,
+      "loss": 0.8968,
+      "step": 38059
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8781714285714284e-06,
+      "loss": 1.2719,
+      "step": 38060
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.8779028571428575e-06,
+      "loss": 1.1748,
+      "step": 38061
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.877634285714286e-06,
+      "loss": 0.9609,
+      "step": 38062
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.877365714285714e-06,
+      "loss": 1.2215,
+      "step": 38063
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.877097142857143e-06,
+      "loss": 0.806,
+      "step": 38064
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.876828571428572e-06,
+      "loss": 0.9044,
+      "step": 38065
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.87656e-06,
+      "loss": 1.6243,
+      "step": 38066
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.876291428571429e-06,
+      "loss": 0.5954,
+      "step": 38067
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.876022857142857e-06,
+      "loss": 1.561,
+      "step": 38068
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.875754285714286e-06,
+      "loss": 1.6168,
+      "step": 38069
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8754857142857144e-06,
+      "loss": 1.1972,
+      "step": 38070
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.875217142857143e-06,
+      "loss": 0.9087,
+      "step": 38071
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.874948571428572e-06,
+      "loss": 0.9614,
+      "step": 38072
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.87468e-06,
+      "loss": 0.8315,
+      "step": 38073
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.874411428571429e-06,
+      "loss": 0.861,
+      "step": 38074
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.874142857142857e-06,
+      "loss": 0.9337,
+      "step": 38075
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.873874285714285e-06,
+      "loss": 1.283,
+      "step": 38076
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.873605714285714e-06,
+      "loss": 1.2297,
+      "step": 38077
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.873337142857143e-06,
+      "loss": 0.7848,
+      "step": 38078
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.873068571428572e-06,
+      "loss": 1.5289,
+      "step": 38079
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8728000000000004e-06,
+      "loss": 1.2142,
+      "step": 38080
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.872531428571429e-06,
+      "loss": 1.0628,
+      "step": 38081
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.872262857142858e-06,
+      "loss": 1.2281,
+      "step": 38082
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.871994285714286e-06,
+      "loss": 1.1528,
+      "step": 38083
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.871725714285715e-06,
+      "loss": 1.0357,
+      "step": 38084
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.871457142857143e-06,
+      "loss": 1.1561,
+      "step": 38085
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.871188571428571e-06,
+      "loss": 0.851,
+      "step": 38086
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.87092e-06,
+      "loss": 1.0274,
+      "step": 38087
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.870651428571428e-06,
+      "loss": 1.106,
+      "step": 38088
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8703828571428566e-06,
+      "loss": 1.2359,
+      "step": 38089
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.870114285714286e-06,
+      "loss": 1.2074,
+      "step": 38090
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.869845714285715e-06,
+      "loss": 1.4676,
+      "step": 38091
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.869577142857144e-06,
+      "loss": 1.0588,
+      "step": 38092
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.869308571428572e-06,
+      "loss": 1.155,
+      "step": 38093
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.86904e-06,
+      "loss": 0.9612,
+      "step": 38094
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.868771428571429e-06,
+      "loss": 1.1955,
+      "step": 38095
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.868502857142857e-06,
+      "loss": 1.2844,
+      "step": 38096
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.868234285714286e-06,
+      "loss": 1.3795,
+      "step": 38097
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.867965714285714e-06,
+      "loss": 0.6385,
+      "step": 38098
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8676971428571425e-06,
+      "loss": 1.1539,
+      "step": 38099
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8674285714285716e-06,
+      "loss": 0.8832,
+      "step": 38100
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.86716e-06,
+      "loss": 1.1567,
+      "step": 38101
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.866891428571429e-06,
+      "loss": 1.0387,
+      "step": 38102
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.866622857142857e-06,
+      "loss": 0.7503,
+      "step": 38103
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.866354285714286e-06,
+      "loss": 1.0997,
+      "step": 38104
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.866085714285715e-06,
+      "loss": 1.6583,
+      "step": 38105
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.865817142857143e-06,
+      "loss": 1.2088,
+      "step": 38106
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.865548571428571e-06,
+      "loss": 1.1763,
+      "step": 38107
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.86528e-06,
+      "loss": 1.4877,
+      "step": 38108
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8650114285714285e-06,
+      "loss": 1.4731,
+      "step": 38109
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8647428571428576e-06,
+      "loss": 1.5122,
+      "step": 38110
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.864474285714286e-06,
+      "loss": 0.8169,
+      "step": 38111
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.864205714285714e-06,
+      "loss": 1.1103,
+      "step": 38112
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.863937142857143e-06,
+      "loss": 1.3005,
+      "step": 38113
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.863668571428571e-06,
+      "loss": 1.3043,
+      "step": 38114
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8634e-06,
+      "loss": 0.7141,
+      "step": 38115
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.863131428571428e-06,
+      "loss": 0.9607,
+      "step": 38116
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.862862857142857e-06,
+      "loss": 1.2385,
+      "step": 38117
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.862594285714286e-06,
+      "loss": 0.7943,
+      "step": 38118
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8623257142857145e-06,
+      "loss": 0.9083,
+      "step": 38119
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8620571428571436e-06,
+      "loss": 0.984,
+      "step": 38120
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.861788571428572e-06,
+      "loss": 1.1189,
+      "step": 38121
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.86152e-06,
+      "loss": 1.3192,
+      "step": 38122
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.861251428571429e-06,
+      "loss": 1.0458,
+      "step": 38123
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.860982857142857e-06,
+      "loss": 1.0595,
+      "step": 38124
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.860714285714286e-06,
+      "loss": 1.3415,
+      "step": 38125
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.860445714285714e-06,
+      "loss": 1.1389,
+      "step": 38126
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8601771428571425e-06,
+      "loss": 1.2422,
+      "step": 38127
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8599085714285715e-06,
+      "loss": 1.3888,
+      "step": 38128
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.85964e-06,
+      "loss": 1.2043,
+      "step": 38129
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.859371428571429e-06,
+      "loss": 1.3074,
+      "step": 38130
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.859102857142858e-06,
+      "loss": 1.1418,
+      "step": 38131
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.858834285714286e-06,
+      "loss": 1.4046,
+      "step": 38132
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.858565714285715e-06,
+      "loss": 1.2098,
+      "step": 38133
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.858297142857143e-06,
+      "loss": 0.8859,
+      "step": 38134
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.858028571428571e-06,
+      "loss": 0.8938,
+      "step": 38135
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.85776e-06,
+      "loss": 1.0121,
+      "step": 38136
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8574914285714285e-06,
+      "loss": 0.8966,
+      "step": 38137
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8572228571428575e-06,
+      "loss": 1.1997,
+      "step": 38138
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.856954285714286e-06,
+      "loss": 1.4062,
+      "step": 38139
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.856685714285714e-06,
+      "loss": 1.1803,
+      "step": 38140
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.856417142857143e-06,
+      "loss": 1.2604,
+      "step": 38141
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.856148571428571e-06,
+      "loss": 0.6341,
+      "step": 38142
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.85588e-06,
+      "loss": 1.0416,
+      "step": 38143
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.855611428571429e-06,
+      "loss": 1.4848,
+      "step": 38144
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.855342857142857e-06,
+      "loss": 1.4795,
+      "step": 38145
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.855074285714286e-06,
+      "loss": 0.9274,
+      "step": 38146
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8548057142857145e-06,
+      "loss": 1.1132,
+      "step": 38147
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8545371428571435e-06,
+      "loss": 1.4626,
+      "step": 38148
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.854268571428572e-06,
+      "loss": 1.7616,
+      "step": 38149
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.854e-06,
+      "loss": 1.7108,
+      "step": 38150
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.853731428571429e-06,
+      "loss": 0.8021,
+      "step": 38151
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.853462857142857e-06,
+      "loss": 1.4747,
+      "step": 38152
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.853194285714285e-06,
+      "loss": 1.0072,
+      "step": 38153
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.852925714285714e-06,
+      "loss": 1.4533,
+      "step": 38154
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8526571428571424e-06,
+      "loss": 1.3918,
+      "step": 38155
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8523885714285715e-06,
+      "loss": 0.857,
+      "step": 38156
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8521200000000005e-06,
+      "loss": 1.0038,
+      "step": 38157
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.851851428571429e-06,
+      "loss": 1.3676,
+      "step": 38158
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.851582857142858e-06,
+      "loss": 1.0754,
+      "step": 38159
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.851314285714286e-06,
+      "loss": 1.3197,
+      "step": 38160
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.851045714285715e-06,
+      "loss": 0.9939,
+      "step": 38161
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.850777142857143e-06,
+      "loss": 1.2381,
+      "step": 38162
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.850508571428571e-06,
+      "loss": 1.0802,
+      "step": 38163
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.85024e-06,
+      "loss": 1.4713,
+      "step": 38164
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8499714285714284e-06,
+      "loss": 1.1804,
+      "step": 38165
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8497028571428575e-06,
+      "loss": 0.9061,
+      "step": 38166
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.849434285714286e-06,
+      "loss": 1.3593,
+      "step": 38167
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.849165714285714e-06,
+      "loss": 0.9797,
+      "step": 38168
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.848897142857143e-06,
+      "loss": 0.8768,
+      "step": 38169
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.848628571428572e-06,
+      "loss": 1.1207,
+      "step": 38170
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.84836e-06,
+      "loss": 1.0451,
+      "step": 38171
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.848091428571429e-06,
+      "loss": 0.8436,
+      "step": 38172
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.847822857142857e-06,
+      "loss": 1.3524,
+      "step": 38173
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.847554285714286e-06,
+      "loss": 0.9168,
+      "step": 38174
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8472857142857144e-06,
+      "loss": 1.1179,
+      "step": 38175
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.847017142857143e-06,
+      "loss": 1.1009,
+      "step": 38176
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.846748571428572e-06,
+      "loss": 1.4555,
+      "step": 38177
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.84648e-06,
+      "loss": 0.8734,
+      "step": 38178
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.846211428571429e-06,
+      "loss": 1.0581,
+      "step": 38179
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.845942857142857e-06,
+      "loss": 0.8401,
+      "step": 38180
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.845674285714285e-06,
+      "loss": 1.2811,
+      "step": 38181
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.845405714285714e-06,
+      "loss": 0.8875,
+      "step": 38182
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.845137142857143e-06,
+      "loss": 1.1721,
+      "step": 38183
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.844868571428572e-06,
+      "loss": 1.0225,
+      "step": 38184
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8446000000000004e-06,
+      "loss": 1.5089,
+      "step": 38185
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.844331428571429e-06,
+      "loss": 1.2189,
+      "step": 38186
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.844062857142858e-06,
+      "loss": 0.9549,
+      "step": 38187
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.843794285714286e-06,
+      "loss": 1.3629,
+      "step": 38188
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.843525714285715e-06,
+      "loss": 1.0118,
+      "step": 38189
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.843257142857143e-06,
+      "loss": 1.4859,
+      "step": 38190
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.842988571428571e-06,
+      "loss": 1.1683,
+      "step": 38191
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.84272e-06,
+      "loss": 1.3801,
+      "step": 38192
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.842451428571428e-06,
+      "loss": 1.3332,
+      "step": 38193
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8421828571428566e-06,
+      "loss": 0.6293,
+      "step": 38194
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8419142857142856e-06,
+      "loss": 0.9818,
+      "step": 38195
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.841645714285715e-06,
+      "loss": 0.9901,
+      "step": 38196
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.841377142857144e-06,
+      "loss": 0.8,
+      "step": 38197
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.841108571428572e-06,
+      "loss": 1.1171,
+      "step": 38198
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.84084e-06,
+      "loss": 0.8208,
+      "step": 38199
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.840571428571429e-06,
+      "loss": 1.0124,
+      "step": 38200
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.840302857142857e-06,
+      "loss": 1.4505,
+      "step": 38201
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.840034285714286e-06,
+      "loss": 1.1031,
+      "step": 38202
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.839765714285714e-06,
+      "loss": 0.707,
+      "step": 38203
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8394971428571425e-06,
+      "loss": 0.953,
+      "step": 38204
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8392285714285716e-06,
+      "loss": 0.6659,
+      "step": 38205
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.83896e-06,
+      "loss": 1.2683,
+      "step": 38206
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.838691428571429e-06,
+      "loss": 1.258,
+      "step": 38207
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.838422857142857e-06,
+      "loss": 1.1997,
+      "step": 38208
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.838154285714286e-06,
+      "loss": 1.4118,
+      "step": 38209
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.837885714285715e-06,
+      "loss": 0.9413,
+      "step": 38210
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.837617142857143e-06,
+      "loss": 0.9025,
+      "step": 38211
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.837348571428572e-06,
+      "loss": 1.0341,
+      "step": 38212
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.83708e-06,
+      "loss": 1.1573,
+      "step": 38213
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8368114285714285e-06,
+      "loss": 0.8485,
+      "step": 38214
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8365428571428576e-06,
+      "loss": 1.0561,
+      "step": 38215
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.836274285714286e-06,
+      "loss": 0.9092,
+      "step": 38216
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.836005714285714e-06,
+      "loss": 1.2305,
+      "step": 38217
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.835737142857143e-06,
+      "loss": 0.9891,
+      "step": 38218
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.835468571428571e-06,
+      "loss": 1.0675,
+      "step": 38219
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8352e-06,
+      "loss": 1.0358,
+      "step": 38220
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.834931428571428e-06,
+      "loss": 1.1188,
+      "step": 38221
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.834662857142857e-06,
+      "loss": 1.3328,
+      "step": 38222
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.834394285714286e-06,
+      "loss": 0.9142,
+      "step": 38223
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8341257142857145e-06,
+      "loss": 1.0689,
+      "step": 38224
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8338571428571436e-06,
+      "loss": 1.2344,
+      "step": 38225
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.833588571428572e-06,
+      "loss": 1.4388,
+      "step": 38226
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.83332e-06,
+      "loss": 1.0493,
+      "step": 38227
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.833051428571429e-06,
+      "loss": 1.039,
+      "step": 38228
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.832782857142857e-06,
+      "loss": 0.9182,
+      "step": 38229
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.832514285714286e-06,
+      "loss": 1.1175,
+      "step": 38230
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.832245714285714e-06,
+      "loss": 1.2395,
+      "step": 38231
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8319771428571425e-06,
+      "loss": 1.1735,
+      "step": 38232
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.8317085714285715e-06,
+      "loss": 0.8908,
+      "step": 38233
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.83144e-06,
+      "loss": 1.5279,
+      "step": 38234
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.831171428571429e-06,
+      "loss": 1.2162,
+      "step": 38235
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.830902857142858e-06,
+      "loss": 1.0912,
+      "step": 38236
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.830634285714286e-06,
+      "loss": 1.0567,
+      "step": 38237
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.830365714285715e-06,
+      "loss": 1.1683,
+      "step": 38238
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.830097142857143e-06,
+      "loss": 1.2455,
+      "step": 38239
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.829828571428571e-06,
+      "loss": 0.8618,
+      "step": 38240
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.82956e-06,
+      "loss": 1.3211,
+      "step": 38241
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8292914285714285e-06,
+      "loss": 1.5037,
+      "step": 38242
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8290228571428575e-06,
+      "loss": 1.3397,
+      "step": 38243
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.828754285714286e-06,
+      "loss": 1.0545,
+      "step": 38244
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.828485714285714e-06,
+      "loss": 0.6658,
+      "step": 38245
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.828217142857143e-06,
+      "loss": 0.9424,
+      "step": 38246
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.827948571428571e-06,
+      "loss": 1.1122,
+      "step": 38247
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.82768e-06,
+      "loss": 1.4043,
+      "step": 38248
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.827411428571429e-06,
+      "loss": 1.3374,
+      "step": 38249
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.827142857142857e-06,
+      "loss": 1.2273,
+      "step": 38250
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.826874285714286e-06,
+      "loss": 1.2288,
+      "step": 38251
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8266057142857145e-06,
+      "loss": 0.8784,
+      "step": 38252
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8263371428571435e-06,
+      "loss": 1.3056,
+      "step": 38253
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.826068571428572e-06,
+      "loss": 0.9756,
+      "step": 38254
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8258e-06,
+      "loss": 0.7264,
+      "step": 38255
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.825531428571429e-06,
+      "loss": 1.1945,
+      "step": 38256
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.825262857142857e-06,
+      "loss": 0.6861,
+      "step": 38257
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.824994285714285e-06,
+      "loss": 0.9793,
+      "step": 38258
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.824725714285714e-06,
+      "loss": 1.4485,
+      "step": 38259
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8244571428571424e-06,
+      "loss": 1.3489,
+      "step": 38260
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8241885714285715e-06,
+      "loss": 1.0306,
+      "step": 38261
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8239200000000005e-06,
+      "loss": 1.4441,
+      "step": 38262
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.823651428571429e-06,
+      "loss": 1.1806,
+      "step": 38263
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.823382857142858e-06,
+      "loss": 0.992,
+      "step": 38264
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.823114285714286e-06,
+      "loss": 1.0574,
+      "step": 38265
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.822845714285715e-06,
+      "loss": 0.9225,
+      "step": 38266
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.822577142857143e-06,
+      "loss": 0.8837,
+      "step": 38267
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.822308571428571e-06,
+      "loss": 1.0326,
+      "step": 38268
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.82204e-06,
+      "loss": 1.2813,
+      "step": 38269
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8217714285714284e-06,
+      "loss": 1.2007,
+      "step": 38270
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8215028571428575e-06,
+      "loss": 1.2463,
+      "step": 38271
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.821234285714286e-06,
+      "loss": 1.2884,
+      "step": 38272
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.820965714285714e-06,
+      "loss": 1.3975,
+      "step": 38273
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.820697142857143e-06,
+      "loss": 0.7062,
+      "step": 38274
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.820428571428572e-06,
+      "loss": 0.975,
+      "step": 38275
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.820160000000001e-06,
+      "loss": 0.8003,
+      "step": 38276
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.819891428571429e-06,
+      "loss": 1.2171,
+      "step": 38277
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.819622857142857e-06,
+      "loss": 0.9998,
+      "step": 38278
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.819354285714286e-06,
+      "loss": 1.5124,
+      "step": 38279
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8190857142857144e-06,
+      "loss": 0.7858,
+      "step": 38280
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.818817142857143e-06,
+      "loss": 1.1527,
+      "step": 38281
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.818548571428572e-06,
+      "loss": 1.2937,
+      "step": 38282
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.81828e-06,
+      "loss": 1.0162,
+      "step": 38283
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.818011428571429e-06,
+      "loss": 0.8972,
+      "step": 38284
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.817742857142857e-06,
+      "loss": 1.4054,
+      "step": 38285
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.817474285714285e-06,
+      "loss": 1.0143,
+      "step": 38286
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.817205714285714e-06,
+      "loss": 1.3471,
+      "step": 38287
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.816937142857143e-06,
+      "loss": 1.2629,
+      "step": 38288
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.816668571428572e-06,
+      "loss": 1.134,
+      "step": 38289
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8164000000000004e-06,
+      "loss": 1.0363,
+      "step": 38290
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.816131428571429e-06,
+      "loss": 1.3556,
+      "step": 38291
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.815862857142858e-06,
+      "loss": 0.8238,
+      "step": 38292
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.815594285714286e-06,
+      "loss": 1.0957,
+      "step": 38293
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.815325714285715e-06,
+      "loss": 1.3617,
+      "step": 38294
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.815057142857143e-06,
+      "loss": 0.9575,
+      "step": 38295
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.814788571428571e-06,
+      "loss": 0.8486,
+      "step": 38296
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.81452e-06,
+      "loss": 0.7188,
+      "step": 38297
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.814251428571429e-06,
+      "loss": 1.3674,
+      "step": 38298
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.813982857142858e-06,
+      "loss": 0.9485,
+      "step": 38299
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.813714285714286e-06,
+      "loss": 1.2358,
+      "step": 38300
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.813445714285714e-06,
+      "loss": 1.3857,
+      "step": 38301
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.813177142857143e-06,
+      "loss": 1.6473,
+      "step": 38302
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8129085714285714e-06,
+      "loss": 1.4288,
+      "step": 38303
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.81264e-06,
+      "loss": 1.7682,
+      "step": 38304
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.812371428571429e-06,
+      "loss": 0.8659,
+      "step": 38305
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.812102857142857e-06,
+      "loss": 1.4316,
+      "step": 38306
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.811834285714286e-06,
+      "loss": 0.7847,
+      "step": 38307
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8115657142857144e-06,
+      "loss": 1.4243,
+      "step": 38308
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8112971428571425e-06,
+      "loss": 0.9554,
+      "step": 38309
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8110285714285716e-06,
+      "loss": 0.9192,
+      "step": 38310
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.81076e-06,
+      "loss": 1.1286,
+      "step": 38311
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.810491428571429e-06,
+      "loss": 1.0654,
+      "step": 38312
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8102228571428574e-06,
+      "loss": 1.3096,
+      "step": 38313
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8099542857142855e-06,
+      "loss": 0.9902,
+      "step": 38314
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8096857142857146e-06,
+      "loss": 0.7279,
+      "step": 38315
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8094171428571427e-06,
+      "loss": 1.142,
+      "step": 38316
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8091485714285718e-06,
+      "loss": 0.8525,
+      "step": 38317
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8088800000000004e-06,
+      "loss": 0.9859,
+      "step": 38318
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8086114285714285e-06,
+      "loss": 0.4666,
+      "step": 38319
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8083428571428576e-06,
+      "loss": 1.8464,
+      "step": 38320
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8080742857142857e-06,
+      "loss": 1.0697,
+      "step": 38321
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8078057142857148e-06,
+      "loss": 1.2264,
+      "step": 38322
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.807537142857143e-06,
+      "loss": 0.9178,
+      "step": 38323
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8072685714285715e-06,
+      "loss": 0.8548,
+      "step": 38324
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8070000000000006e-06,
+      "loss": 1.13,
+      "step": 38325
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8067314285714287e-06,
+      "loss": 1.3643,
+      "step": 38326
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.806462857142857e-06,
+      "loss": 1.3982,
+      "step": 38327
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.806194285714286e-06,
+      "loss": 1.0549,
+      "step": 38328
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.805925714285714e-06,
+      "loss": 0.6776,
+      "step": 38329
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.805657142857143e-06,
+      "loss": 0.7431,
+      "step": 38330
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8053885714285717e-06,
+      "loss": 1.1203,
+      "step": 38331
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.80512e-06,
+      "loss": 1.3727,
+      "step": 38332
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.804851428571429e-06,
+      "loss": 1.1003,
+      "step": 38333
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.804582857142857e-06,
+      "loss": 0.916,
+      "step": 38334
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.804314285714286e-06,
+      "loss": 1.4762,
+      "step": 38335
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8040457142857143e-06,
+      "loss": 1.3538,
+      "step": 38336
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.803777142857143e-06,
+      "loss": 1.621,
+      "step": 38337
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.803508571428572e-06,
+      "loss": 1.0087,
+      "step": 38338
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.80324e-06,
+      "loss": 1.4363,
+      "step": 38339
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.802971428571429e-06,
+      "loss": 1.3012,
+      "step": 38340
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8027028571428573e-06,
+      "loss": 0.7591,
+      "step": 38341
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8024342857142855e-06,
+      "loss": 1.0412,
+      "step": 38342
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8021657142857145e-06,
+      "loss": 1.5062,
+      "step": 38343
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.801897142857143e-06,
+      "loss": 0.6597,
+      "step": 38344
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8016285714285713e-06,
+      "loss": 1.3484,
+      "step": 38345
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8013600000000003e-06,
+      "loss": 1.1662,
+      "step": 38346
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8010914285714285e-06,
+      "loss": 0.7526,
+      "step": 38347
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8008228571428575e-06,
+      "loss": 1.2463,
+      "step": 38348
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8005542857142857e-06,
+      "loss": 1.2603,
+      "step": 38349
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8002857142857143e-06,
+      "loss": 1.3093,
+      "step": 38350
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.8000171428571433e-06,
+      "loss": 0.8507,
+      "step": 38351
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7997485714285715e-06,
+      "loss": 1.2146,
+      "step": 38352
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7994800000000005e-06,
+      "loss": 1.1742,
+      "step": 38353
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7992114285714287e-06,
+      "loss": 1.0096,
+      "step": 38354
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.798942857142857e-06,
+      "loss": 1.1137,
+      "step": 38355
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.798674285714286e-06,
+      "loss": 1.1816,
+      "step": 38356
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7984057142857145e-06,
+      "loss": 0.7677,
+      "step": 38357
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7981371428571435e-06,
+      "loss": 1.3199,
+      "step": 38358
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7978685714285717e-06,
+      "loss": 1.4848,
+      "step": 38359
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7976e-06,
+      "loss": 0.8361,
+      "step": 38360
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.797331428571429e-06,
+      "loss": 0.8604,
+      "step": 38361
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.797062857142857e-06,
+      "loss": 0.9258,
+      "step": 38362
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.796794285714286e-06,
+      "loss": 0.7658,
+      "step": 38363
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7965257142857147e-06,
+      "loss": 1.1751,
+      "step": 38364
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.796257142857143e-06,
+      "loss": 1.0881,
+      "step": 38365
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.795988571428572e-06,
+      "loss": 1.673,
+      "step": 38366
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.79572e-06,
+      "loss": 1.1635,
+      "step": 38367
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7954514285714282e-06,
+      "loss": 1.4087,
+      "step": 38368
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7951828571428573e-06,
+      "loss": 0.7728,
+      "step": 38369
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.794914285714286e-06,
+      "loss": 1.3764,
+      "step": 38370
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.794645714285715e-06,
+      "loss": 0.6803,
+      "step": 38371
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.794377142857143e-06,
+      "loss": 1.2111,
+      "step": 38372
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7941085714285712e-06,
+      "loss": 1.5061,
+      "step": 38373
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7938400000000003e-06,
+      "loss": 1.1051,
+      "step": 38374
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7935714285714284e-06,
+      "loss": 1.055,
+      "step": 38375
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7933028571428575e-06,
+      "loss": 0.8911,
+      "step": 38376
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.793034285714286e-06,
+      "loss": 1.2889,
+      "step": 38377
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7927657142857142e-06,
+      "loss": 1.0524,
+      "step": 38378
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7924971428571433e-06,
+      "loss": 1.2558,
+      "step": 38379
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7922285714285714e-06,
+      "loss": 1.2066,
+      "step": 38380
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7919600000000005e-06,
+      "loss": 1.3046,
+      "step": 38381
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7916914285714286e-06,
+      "loss": 1.4015,
+      "step": 38382
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.791422857142857e-06,
+      "loss": 1.1954,
+      "step": 38383
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7911542857142863e-06,
+      "loss": 1.2016,
+      "step": 38384
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7908857142857144e-06,
+      "loss": 1.4095,
+      "step": 38385
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7906171428571435e-06,
+      "loss": 1.2308,
+      "step": 38386
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7903485714285716e-06,
+      "loss": 1.1758,
+      "step": 38387
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.79008e-06,
+      "loss": 1.3579,
+      "step": 38388
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.789811428571429e-06,
+      "loss": 1.5696,
+      "step": 38389
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7895428571428574e-06,
+      "loss": 1.2707,
+      "step": 38390
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7892742857142856e-06,
+      "loss": 0.9008,
+      "step": 38391
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7890057142857146e-06,
+      "loss": 0.6761,
+      "step": 38392
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.788737142857143e-06,
+      "loss": 0.8758,
+      "step": 38393
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.788468571428572e-06,
+      "loss": 0.9929,
+      "step": 38394
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7882e-06,
+      "loss": 1.0331,
+      "step": 38395
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.787931428571428e-06,
+      "loss": 0.9624,
+      "step": 38396
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7876628571428576e-06,
+      "loss": 0.9768,
+      "step": 38397
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.787394285714286e-06,
+      "loss": 1.0489,
+      "step": 38398
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.787125714285715e-06,
+      "loss": 1.2348,
+      "step": 38399
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.786857142857143e-06,
+      "loss": 1.5365,
+      "step": 38400
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.786588571428571e-06,
+      "loss": 1.2601,
+      "step": 38401
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.78632e-06,
+      "loss": 1.7434,
+      "step": 38402
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.786051428571429e-06,
+      "loss": 0.9953,
+      "step": 38403
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.785782857142858e-06,
+      "loss": 0.9693,
+      "step": 38404
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.785514285714286e-06,
+      "loss": 1.1547,
+      "step": 38405
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.785245714285714e-06,
+      "loss": 1.4652,
+      "step": 38406
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.784977142857143e-06,
+      "loss": 1.1234,
+      "step": 38407
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7847085714285714e-06,
+      "loss": 0.6026,
+      "step": 38408
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7844400000000004e-06,
+      "loss": 1.3194,
+      "step": 38409
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.784171428571429e-06,
+      "loss": 1.0876,
+      "step": 38410
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.783902857142857e-06,
+      "loss": 0.9215,
+      "step": 38411
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.783634285714286e-06,
+      "loss": 1.1567,
+      "step": 38412
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7833657142857144e-06,
+      "loss": 1.2835,
+      "step": 38413
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7830971428571425e-06,
+      "loss": 1.4495,
+      "step": 38414
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7828285714285716e-06,
+      "loss": 1.0755,
+      "step": 38415
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7825599999999997e-06,
+      "loss": 1.2546,
+      "step": 38416
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.782291428571429e-06,
+      "loss": 1.0558,
+      "step": 38417
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7820228571428574e-06,
+      "loss": 1.4272,
+      "step": 38418
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7817542857142855e-06,
+      "loss": 1.3699,
+      "step": 38419
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7814857142857146e-06,
+      "loss": 0.723,
+      "step": 38420
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7812171428571427e-06,
+      "loss": 1.0016,
+      "step": 38421
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7809485714285718e-06,
+      "loss": 1.1578,
+      "step": 38422
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7806800000000004e-06,
+      "loss": 1.0831,
+      "step": 38423
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7804114285714285e-06,
+      "loss": 1.1032,
+      "step": 38424
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7801428571428576e-06,
+      "loss": 1.0933,
+      "step": 38425
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7798742857142857e-06,
+      "loss": 1.4617,
+      "step": 38426
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7796057142857148e-06,
+      "loss": 1.03,
+      "step": 38427
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.779337142857143e-06,
+      "loss": 0.8284,
+      "step": 38428
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.779068571428571e-06,
+      "loss": 1.1226,
+      "step": 38429
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7788000000000006e-06,
+      "loss": 1.1067,
+      "step": 38430
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7785314285714287e-06,
+      "loss": 1.0934,
+      "step": 38431
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.778262857142857e-06,
+      "loss": 1.1248,
+      "step": 38432
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.777994285714286e-06,
+      "loss": 0.8709,
+      "step": 38433
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.777725714285714e-06,
+      "loss": 0.8265,
+      "step": 38434
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.777457142857143e-06,
+      "loss": 1.4469,
+      "step": 38435
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7771885714285713e-06,
+      "loss": 1.0071,
+      "step": 38436
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.77692e-06,
+      "loss": 1.1034,
+      "step": 38437
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.776651428571429e-06,
+      "loss": 1.1207,
+      "step": 38438
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.776382857142857e-06,
+      "loss": 0.7853,
+      "step": 38439
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.776114285714286e-06,
+      "loss": 1.3141,
+      "step": 38440
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7758457142857143e-06,
+      "loss": 1.2238,
+      "step": 38441
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7755771428571425e-06,
+      "loss": 0.9241,
+      "step": 38442
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.775308571428572e-06,
+      "loss": 1.4359,
+      "step": 38443
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.77504e-06,
+      "loss": 0.9844,
+      "step": 38444
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.774771428571429e-06,
+      "loss": 0.8246,
+      "step": 38445
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7745028571428573e-06,
+      "loss": 0.9012,
+      "step": 38446
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7742342857142855e-06,
+      "loss": 1.4165,
+      "step": 38447
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7739657142857145e-06,
+      "loss": 1.096,
+      "step": 38448
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7736971428571427e-06,
+      "loss": 1.0431,
+      "step": 38449
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.773428571428572e-06,
+      "loss": 0.8709,
+      "step": 38450
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7731600000000003e-06,
+      "loss": 1.2732,
+      "step": 38451
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7728914285714285e-06,
+      "loss": 0.8866,
+      "step": 38452
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7726228571428575e-06,
+      "loss": 1.2723,
+      "step": 38453
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7723542857142857e-06,
+      "loss": 1.0358,
+      "step": 38454
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.772085714285714e-06,
+      "loss": 0.9472,
+      "step": 38455
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7718171428571433e-06,
+      "loss": 0.7295,
+      "step": 38456
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7715485714285715e-06,
+      "loss": 1.0031,
+      "step": 38457
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7712800000000005e-06,
+      "loss": 1.1404,
+      "step": 38458
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7710114285714287e-06,
+      "loss": 1.0116,
+      "step": 38459
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.770742857142857e-06,
+      "loss": 1.1663,
+      "step": 38460
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.770474285714286e-06,
+      "loss": 1.29,
+      "step": 38461
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.770205714285714e-06,
+      "loss": 0.8697,
+      "step": 38462
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7699371428571435e-06,
+      "loss": 0.7353,
+      "step": 38463
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7696685714285717e-06,
+      "loss": 1.4474,
+      "step": 38464
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7694e-06,
+      "loss": 1.0853,
+      "step": 38465
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.769131428571429e-06,
+      "loss": 1.1238,
+      "step": 38466
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.768862857142857e-06,
+      "loss": 1.1596,
+      "step": 38467
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.768594285714286e-06,
+      "loss": 0.7653,
+      "step": 38468
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7683257142857143e-06,
+      "loss": 1.0892,
+      "step": 38469
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.768057142857143e-06,
+      "loss": 0.9977,
+      "step": 38470
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.767788571428572e-06,
+      "loss": 0.9334,
+      "step": 38471
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.76752e-06,
+      "loss": 0.8663,
+      "step": 38472
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.767251428571429e-06,
+      "loss": 1.6982,
+      "step": 38473
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7669828571428573e-06,
+      "loss": 1.1622,
+      "step": 38474
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7667142857142854e-06,
+      "loss": 0.9666,
+      "step": 38475
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.766445714285715e-06,
+      "loss": 1.5367,
+      "step": 38476
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.766177142857143e-06,
+      "loss": 1.6666,
+      "step": 38477
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7659085714285712e-06,
+      "loss": 1.2071,
+      "step": 38478
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7656400000000003e-06,
+      "loss": 1.2168,
+      "step": 38479
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7653714285714284e-06,
+      "loss": 1.6284,
+      "step": 38480
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7651028571428575e-06,
+      "loss": 0.6133,
+      "step": 38481
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7648342857142856e-06,
+      "loss": 0.7794,
+      "step": 38482
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7645657142857142e-06,
+      "loss": 1.0465,
+      "step": 38483
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7642971428571433e-06,
+      "loss": 1.0356,
+      "step": 38484
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7640285714285714e-06,
+      "loss": 0.8895,
+      "step": 38485
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7637600000000005e-06,
+      "loss": 1.075,
+      "step": 38486
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7634914285714286e-06,
+      "loss": 1.218,
+      "step": 38487
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.763222857142857e-06,
+      "loss": 0.9768,
+      "step": 38488
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.762954285714286e-06,
+      "loss": 1.4386,
+      "step": 38489
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7626857142857144e-06,
+      "loss": 1.3713,
+      "step": 38490
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7624171428571434e-06,
+      "loss": 1.2981,
+      "step": 38491
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7621485714285716e-06,
+      "loss": 1.4378,
+      "step": 38492
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.76188e-06,
+      "loss": 0.9639,
+      "step": 38493
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.761611428571429e-06,
+      "loss": 1.3937,
+      "step": 38494
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.761342857142857e-06,
+      "loss": 1.254,
+      "step": 38495
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7610742857142864e-06,
+      "loss": 0.9629,
+      "step": 38496
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7608057142857146e-06,
+      "loss": 0.666,
+      "step": 38497
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.760537142857143e-06,
+      "loss": 1.1112,
+      "step": 38498
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.760268571428572e-06,
+      "loss": 0.8628,
+      "step": 38499
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.76e-06,
+      "loss": 1.2216,
+      "step": 38500
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.759731428571428e-06,
+      "loss": 1.1355,
+      "step": 38501
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.759462857142857e-06,
+      "loss": 0.7525,
+      "step": 38502
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.759194285714286e-06,
+      "loss": 1.363,
+      "step": 38503
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.758925714285715e-06,
+      "loss": 1.079,
+      "step": 38504
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.758657142857143e-06,
+      "loss": 0.9113,
+      "step": 38505
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.758388571428571e-06,
+      "loss": 1.2175,
+      "step": 38506
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.75812e-06,
+      "loss": 0.8921,
+      "step": 38507
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7578514285714284e-06,
+      "loss": 1.3364,
+      "step": 38508
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.757582857142858e-06,
+      "loss": 1.5538,
+      "step": 38509
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.757314285714286e-06,
+      "loss": 0.9649,
+      "step": 38510
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.757045714285714e-06,
+      "loss": 1.1574,
+      "step": 38511
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.756777142857143e-06,
+      "loss": 1.3002,
+      "step": 38512
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7565085714285714e-06,
+      "loss": 0.9547,
+      "step": 38513
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7562400000000004e-06,
+      "loss": 0.7062,
+      "step": 38514
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7559714285714286e-06,
+      "loss": 0.9474,
+      "step": 38515
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.755702857142857e-06,
+      "loss": 0.7496,
+      "step": 38516
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.755434285714286e-06,
+      "loss": 0.9605,
+      "step": 38517
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7551657142857144e-06,
+      "loss": 1.5513,
+      "step": 38518
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7548971428571425e-06,
+      "loss": 1.4469,
+      "step": 38519
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7546285714285716e-06,
+      "loss": 1.0889,
+      "step": 38520
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7543599999999997e-06,
+      "loss": 0.7391,
+      "step": 38521
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7540914285714288e-06,
+      "loss": 1.0692,
+      "step": 38522
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7538228571428574e-06,
+      "loss": 1.2572,
+      "step": 38523
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7535542857142855e-06,
+      "loss": 1.5718,
+      "step": 38524
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7532857142857146e-06,
+      "loss": 0.8182,
+      "step": 38525
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7530171428571427e-06,
+      "loss": 0.9594,
+      "step": 38526
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7527485714285718e-06,
+      "loss": 1.1733,
+      "step": 38527
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.75248e-06,
+      "loss": 1.0239,
+      "step": 38528
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7522114285714285e-06,
+      "loss": 1.2195,
+      "step": 38529
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7519428571428576e-06,
+      "loss": 1.3931,
+      "step": 38530
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7516742857142857e-06,
+      "loss": 1.0452,
+      "step": 38531
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7514057142857148e-06,
+      "loss": 1.559,
+      "step": 38532
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.751137142857143e-06,
+      "loss": 1.2002,
+      "step": 38533
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.750868571428571e-06,
+      "loss": 1.0323,
+      "step": 38534
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7506e-06,
+      "loss": 1.3813,
+      "step": 38535
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7503314285714287e-06,
+      "loss": 1.1975,
+      "step": 38536
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7500628571428578e-06,
+      "loss": 1.1903,
+      "step": 38537
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.749794285714286e-06,
+      "loss": 1.2614,
+      "step": 38538
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.749525714285714e-06,
+      "loss": 0.8022,
+      "step": 38539
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.749257142857143e-06,
+      "loss": 0.9553,
+      "step": 38540
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7489885714285713e-06,
+      "loss": 1.2418,
+      "step": 38541
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.74872e-06,
+      "loss": 0.9463,
+      "step": 38542
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.748451428571429e-06,
+      "loss": 1.3299,
+      "step": 38543
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.748182857142857e-06,
+      "loss": 1.4253,
+      "step": 38544
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.747914285714286e-06,
+      "loss": 1.3906,
+      "step": 38545
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7476457142857143e-06,
+      "loss": 1.2072,
+      "step": 38546
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7473771428571425e-06,
+      "loss": 1.3182,
+      "step": 38547
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7471085714285715e-06,
+      "loss": 1.0088,
+      "step": 38548
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.74684e-06,
+      "loss": 0.9588,
+      "step": 38549
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.746571428571429e-06,
+      "loss": 1.1829,
+      "step": 38550
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7463028571428573e-06,
+      "loss": 1.4088,
+      "step": 38551
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7460342857142855e-06,
+      "loss": 1.3041,
+      "step": 38552
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7457657142857145e-06,
+      "loss": 1.0156,
+      "step": 38553
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7454971428571427e-06,
+      "loss": 1.2069,
+      "step": 38554
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7452285714285717e-06,
+      "loss": 1.3848,
+      "step": 38555
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7449600000000003e-06,
+      "loss": 1.0748,
+      "step": 38556
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7446914285714285e-06,
+      "loss": 0.9853,
+      "step": 38557
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7444228571428575e-06,
+      "loss": 1.3043,
+      "step": 38558
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7441542857142857e-06,
+      "loss": 1.2537,
+      "step": 38559
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7438857142857147e-06,
+      "loss": 1.1504,
+      "step": 38560
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.743617142857143e-06,
+      "loss": 1.0775,
+      "step": 38561
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7433485714285715e-06,
+      "loss": 1.064,
+      "step": 38562
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7430800000000005e-06,
+      "loss": 0.9272,
+      "step": 38563
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7428114285714287e-06,
+      "loss": 1.3002,
+      "step": 38564
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.742542857142857e-06,
+      "loss": 0.8008,
+      "step": 38565
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.742274285714286e-06,
+      "loss": 1.2069,
+      "step": 38566
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.742005714285714e-06,
+      "loss": 1.2787,
+      "step": 38567
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.741737142857143e-06,
+      "loss": 1.2466,
+      "step": 38568
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7414685714285717e-06,
+      "loss": 1.3109,
+      "step": 38569
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7412e-06,
+      "loss": 1.0359,
+      "step": 38570
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.740931428571429e-06,
+      "loss": 1.0876,
+      "step": 38571
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.740662857142857e-06,
+      "loss": 1.1192,
+      "step": 38572
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.740394285714286e-06,
+      "loss": 1.5447,
+      "step": 38573
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7401257142857143e-06,
+      "loss": 0.899,
+      "step": 38574
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.739857142857143e-06,
+      "loss": 1.137,
+      "step": 38575
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.739588571428572e-06,
+      "loss": 1.454,
+      "step": 38576
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.73932e-06,
+      "loss": 1.1846,
+      "step": 38577
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.739051428571429e-06,
+      "loss": 0.8063,
+      "step": 38578
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7387828571428573e-06,
+      "loss": 1.0879,
+      "step": 38579
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7385142857142854e-06,
+      "loss": 1.2203,
+      "step": 38580
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7382457142857145e-06,
+      "loss": 1.0631,
+      "step": 38581
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.737977142857143e-06,
+      "loss": 0.9404,
+      "step": 38582
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.737708571428572e-06,
+      "loss": 1.2348,
+      "step": 38583
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7374400000000003e-06,
+      "loss": 1.0095,
+      "step": 38584
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7371714285714284e-06,
+      "loss": 0.8382,
+      "step": 38585
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7369028571428575e-06,
+      "loss": 1.3381,
+      "step": 38586
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.7366342857142856e-06,
+      "loss": 0.8893,
+      "step": 38587
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7363657142857142e-06,
+      "loss": 1.1898,
+      "step": 38588
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7360971428571433e-06,
+      "loss": 1.1844,
+      "step": 38589
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7358285714285714e-06,
+      "loss": 0.9401,
+      "step": 38590
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7355600000000005e-06,
+      "loss": 1.386,
+      "step": 38591
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7352914285714286e-06,
+      "loss": 0.5396,
+      "step": 38592
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.735022857142857e-06,
+      "loss": 0.9658,
+      "step": 38593
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.734754285714286e-06,
+      "loss": 0.8547,
+      "step": 38594
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7344857142857144e-06,
+      "loss": 1.1119,
+      "step": 38595
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7342171428571434e-06,
+      "loss": 0.9995,
+      "step": 38596
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7339485714285716e-06,
+      "loss": 0.9983,
+      "step": 38597
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.73368e-06,
+      "loss": 1.0048,
+      "step": 38598
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.733411428571429e-06,
+      "loss": 0.8179,
+      "step": 38599
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.733142857142857e-06,
+      "loss": 0.706,
+      "step": 38600
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.732874285714286e-06,
+      "loss": 1.003,
+      "step": 38601
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7326057142857146e-06,
+      "loss": 0.6885,
+      "step": 38602
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.732337142857143e-06,
+      "loss": 1.4492,
+      "step": 38603
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.732068571428572e-06,
+      "loss": 1.1781,
+      "step": 38604
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7318e-06,
+      "loss": 1.2384,
+      "step": 38605
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.731531428571428e-06,
+      "loss": 1.5864,
+      "step": 38606
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.731262857142857e-06,
+      "loss": 0.8276,
+      "step": 38607
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.730994285714286e-06,
+      "loss": 1.186,
+      "step": 38608
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.730725714285715e-06,
+      "loss": 1.0112,
+      "step": 38609
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.730457142857143e-06,
+      "loss": 0.5101,
+      "step": 38610
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.730188571428571e-06,
+      "loss": 0.8034,
+      "step": 38611
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.72992e-06,
+      "loss": 1.6339,
+      "step": 38612
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7296514285714284e-06,
+      "loss": 1.332,
+      "step": 38613
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7293828571428574e-06,
+      "loss": 1.1105,
+      "step": 38614
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.729114285714286e-06,
+      "loss": 1.2013,
+      "step": 38615
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.728845714285714e-06,
+      "loss": 1.252,
+      "step": 38616
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.728577142857143e-06,
+      "loss": 1.2832,
+      "step": 38617
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7283085714285714e-06,
+      "loss": 0.8558,
+      "step": 38618
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7280400000000004e-06,
+      "loss": 1.5357,
+      "step": 38619
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7277714285714286e-06,
+      "loss": 1.1608,
+      "step": 38620
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.727502857142857e-06,
+      "loss": 1.2383,
+      "step": 38621
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.727234285714286e-06,
+      "loss": 1.4934,
+      "step": 38622
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7269657142857144e-06,
+      "loss": 1.1815,
+      "step": 38623
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7266971428571434e-06,
+      "loss": 1.3667,
+      "step": 38624
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7264285714285716e-06,
+      "loss": 1.6023,
+      "step": 38625
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7261599999999997e-06,
+      "loss": 0.9802,
+      "step": 38626
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7258914285714288e-06,
+      "loss": 1.0278,
+      "step": 38627
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7256228571428574e-06,
+      "loss": 0.9901,
+      "step": 38628
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7253542857142855e-06,
+      "loss": 1.1031,
+      "step": 38629
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7250857142857146e-06,
+      "loss": 1.1332,
+      "step": 38630
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7248171428571427e-06,
+      "loss": 1.4824,
+      "step": 38631
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7245485714285718e-06,
+      "loss": 1.1433,
+      "step": 38632
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.72428e-06,
+      "loss": 1.0403,
+      "step": 38633
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7240114285714285e-06,
+      "loss": 1.4844,
+      "step": 38634
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7237428571428576e-06,
+      "loss": 1.4567,
+      "step": 38635
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7234742857142857e-06,
+      "loss": 1.2983,
+      "step": 38636
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7232057142857148e-06,
+      "loss": 1.2055,
+      "step": 38637
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.722937142857143e-06,
+      "loss": 0.7449,
+      "step": 38638
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.722668571428571e-06,
+      "loss": 0.8879,
+      "step": 38639
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7224e-06,
+      "loss": 1.1228,
+      "step": 38640
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7221314285714287e-06,
+      "loss": 0.762,
+      "step": 38641
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7218628571428578e-06,
+      "loss": 1.3181,
+      "step": 38642
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.721594285714286e-06,
+      "loss": 1.3516,
+      "step": 38643
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.721325714285714e-06,
+      "loss": 1.1107,
+      "step": 38644
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.721057142857143e-06,
+      "loss": 1.3726,
+      "step": 38645
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7207885714285713e-06,
+      "loss": 0.8498,
+      "step": 38646
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7205200000000003e-06,
+      "loss": 1.0663,
+      "step": 38647
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.720251428571429e-06,
+      "loss": 1.3303,
+      "step": 38648
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.719982857142857e-06,
+      "loss": 1.3599,
+      "step": 38649
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.719714285714286e-06,
+      "loss": 0.9034,
+      "step": 38650
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7194457142857143e-06,
+      "loss": 0.9827,
+      "step": 38651
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7191771428571425e-06,
+      "loss": 1.2393,
+      "step": 38652
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7189085714285715e-06,
+      "loss": 1.2391,
+      "step": 38653
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.71864e-06,
+      "loss": 1.2001,
+      "step": 38654
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.718371428571429e-06,
+      "loss": 1.0242,
+      "step": 38655
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7181028571428573e-06,
+      "loss": 1.1086,
+      "step": 38656
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7178342857142855e-06,
+      "loss": 1.2607,
+      "step": 38657
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7175657142857145e-06,
+      "loss": 1.1392,
+      "step": 38658
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7172971428571427e-06,
+      "loss": 1.0958,
+      "step": 38659
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7170285714285717e-06,
+      "loss": 0.9266,
+      "step": 38660
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7167600000000003e-06,
+      "loss": 1.2177,
+      "step": 38661
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7164914285714285e-06,
+      "loss": 1.1214,
+      "step": 38662
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7162228571428575e-06,
+      "loss": 1.1548,
+      "step": 38663
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7159542857142857e-06,
+      "loss": 1.8442,
+      "step": 38664
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7156857142857147e-06,
+      "loss": 1.3658,
+      "step": 38665
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.715417142857143e-06,
+      "loss": 1.5428,
+      "step": 38666
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7151485714285715e-06,
+      "loss": 0.7352,
+      "step": 38667
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7148800000000005e-06,
+      "loss": 1.215,
+      "step": 38668
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7146114285714287e-06,
+      "loss": 1.3245,
+      "step": 38669
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.714342857142857e-06,
+      "loss": 1.0733,
+      "step": 38670
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.714074285714286e-06,
+      "loss": 1.1553,
+      "step": 38671
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.713805714285714e-06,
+      "loss": 1.048,
+      "step": 38672
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.713537142857143e-06,
+      "loss": 0.7691,
+      "step": 38673
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7132685714285717e-06,
+      "loss": 1.2573,
+      "step": 38674
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.713e-06,
+      "loss": 0.612,
+      "step": 38675
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.712731428571429e-06,
+      "loss": 1.3386,
+      "step": 38676
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.712462857142857e-06,
+      "loss": 1.4312,
+      "step": 38677
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.712194285714286e-06,
+      "loss": 0.8081,
+      "step": 38678
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7119257142857143e-06,
+      "loss": 1.0318,
+      "step": 38679
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.711657142857143e-06,
+      "loss": 0.9221,
+      "step": 38680
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.711388571428572e-06,
+      "loss": 1.1407,
+      "step": 38681
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.71112e-06,
+      "loss": 1.3016,
+      "step": 38682
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.710851428571429e-06,
+      "loss": 1.4815,
+      "step": 38683
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7105828571428573e-06,
+      "loss": 0.8475,
+      "step": 38684
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7103142857142854e-06,
+      "loss": 1.2969,
+      "step": 38685
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7100457142857145e-06,
+      "loss": 1.0182,
+      "step": 38686
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.709777142857143e-06,
+      "loss": 1.2906,
+      "step": 38687
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.709508571428572e-06,
+      "loss": 1.1595,
+      "step": 38688
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7092400000000003e-06,
+      "loss": 1.2507,
+      "step": 38689
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7089714285714284e-06,
+      "loss": 0.9858,
+      "step": 38690
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7087028571428575e-06,
+      "loss": 1.4889,
+      "step": 38691
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7084342857142856e-06,
+      "loss": 0.9218,
+      "step": 38692
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7081657142857142e-06,
+      "loss": 1.0454,
+      "step": 38693
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7078971428571432e-06,
+      "loss": 1.4719,
+      "step": 38694
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7076285714285714e-06,
+      "loss": 1.2288,
+      "step": 38695
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7073600000000004e-06,
+      "loss": 1.1336,
+      "step": 38696
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7070914285714286e-06,
+      "loss": 0.7697,
+      "step": 38697
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.706822857142857e-06,
+      "loss": 0.9874,
+      "step": 38698
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.706554285714286e-06,
+      "loss": 0.9135,
+      "step": 38699
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7062857142857144e-06,
+      "loss": 1.0597,
+      "step": 38700
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7060171428571434e-06,
+      "loss": 1.2556,
+      "step": 38701
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7057485714285716e-06,
+      "loss": 1.0067,
+      "step": 38702
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.70548e-06,
+      "loss": 1.186,
+      "step": 38703
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.705211428571429e-06,
+      "loss": 0.9088,
+      "step": 38704
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.704942857142857e-06,
+      "loss": 0.7745,
+      "step": 38705
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.704674285714286e-06,
+      "loss": 0.8821,
+      "step": 38706
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7044057142857146e-06,
+      "loss": 1.2849,
+      "step": 38707
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.704137142857143e-06,
+      "loss": 1.3752,
+      "step": 38708
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.703868571428572e-06,
+      "loss": 1.2364,
+      "step": 38709
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7036e-06,
+      "loss": 1.3311,
+      "step": 38710
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.703331428571429e-06,
+      "loss": 1.2823,
+      "step": 38711
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.703062857142857e-06,
+      "loss": 1.4243,
+      "step": 38712
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.702794285714286e-06,
+      "loss": 1.2342,
+      "step": 38713
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.702525714285715e-06,
+      "loss": 0.846,
+      "step": 38714
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.702257142857143e-06,
+      "loss": 1.0317,
+      "step": 38715
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.701988571428571e-06,
+      "loss": 0.9662,
+      "step": 38716
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.70172e-06,
+      "loss": 0.8273,
+      "step": 38717
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7014514285714284e-06,
+      "loss": 0.8271,
+      "step": 38718
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7011828571428574e-06,
+      "loss": 1.2894,
+      "step": 38719
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.700914285714286e-06,
+      "loss": 1.2386,
+      "step": 38720
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.700645714285714e-06,
+      "loss": 1.1518,
+      "step": 38721
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.700377142857143e-06,
+      "loss": 1.2989,
+      "step": 38722
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7001085714285714e-06,
+      "loss": 1.087,
+      "step": 38723
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6998400000000004e-06,
+      "loss": 1.1615,
+      "step": 38724
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6995714285714286e-06,
+      "loss": 1.1118,
+      "step": 38725
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.699302857142857e-06,
+      "loss": 1.2908,
+      "step": 38726
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.699034285714286e-06,
+      "loss": 1.2051,
+      "step": 38727
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6987657142857144e-06,
+      "loss": 1.0662,
+      "step": 38728
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6984971428571434e-06,
+      "loss": 0.7639,
+      "step": 38729
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6982285714285716e-06,
+      "loss": 1.2243,
+      "step": 38730
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6979599999999997e-06,
+      "loss": 1.2696,
+      "step": 38731
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6976914285714288e-06,
+      "loss": 1.075,
+      "step": 38732
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6974228571428574e-06,
+      "loss": 1.2355,
+      "step": 38733
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6971542857142864e-06,
+      "loss": 0.9609,
+      "step": 38734
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6968857142857146e-06,
+      "loss": 1.1897,
+      "step": 38735
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6966171428571427e-06,
+      "loss": 0.6491,
+      "step": 38736
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6963485714285718e-06,
+      "loss": 1.0967,
+      "step": 38737
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.69608e-06,
+      "loss": 1.2489,
+      "step": 38738
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6958114285714285e-06,
+      "loss": 0.9931,
+      "step": 38739
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6955428571428576e-06,
+      "loss": 1.1887,
+      "step": 38740
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6952742857142857e-06,
+      "loss": 1.1308,
+      "step": 38741
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6950057142857148e-06,
+      "loss": 0.7562,
+      "step": 38742
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.694737142857143e-06,
+      "loss": 0.9974,
+      "step": 38743
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.694468571428571e-06,
+      "loss": 1.1151,
+      "step": 38744
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6942e-06,
+      "loss": 0.8831,
+      "step": 38745
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6939314285714287e-06,
+      "loss": 1.5873,
+      "step": 38746
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6936628571428578e-06,
+      "loss": 1.1508,
+      "step": 38747
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.693394285714286e-06,
+      "loss": 1.1476,
+      "step": 38748
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.693125714285714e-06,
+      "loss": 1.1375,
+      "step": 38749
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.692857142857143e-06,
+      "loss": 0.9858,
+      "step": 38750
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6925885714285713e-06,
+      "loss": 1.0473,
+      "step": 38751
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6923200000000003e-06,
+      "loss": 1.1061,
+      "step": 38752
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.692051428571429e-06,
+      "loss": 1.3326,
+      "step": 38753
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.691782857142857e-06,
+      "loss": 0.8397,
+      "step": 38754
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.691514285714286e-06,
+      "loss": 1.3653,
+      "step": 38755
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6912457142857143e-06,
+      "loss": 1.0664,
+      "step": 38756
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6909771428571425e-06,
+      "loss": 1.2699,
+      "step": 38757
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6907085714285715e-06,
+      "loss": 1.2675,
+      "step": 38758
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.69044e-06,
+      "loss": 0.7442,
+      "step": 38759
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.690171428571429e-06,
+      "loss": 1.641,
+      "step": 38760
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6899028571428573e-06,
+      "loss": 0.8546,
+      "step": 38761
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6896342857142855e-06,
+      "loss": 1.3451,
+      "step": 38762
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6893657142857145e-06,
+      "loss": 0.9572,
+      "step": 38763
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6890971428571427e-06,
+      "loss": 1.1031,
+      "step": 38764
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6888285714285717e-06,
+      "loss": 0.6697,
+      "step": 38765
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6885600000000003e-06,
+      "loss": 1.0594,
+      "step": 38766
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6882914285714285e-06,
+      "loss": 1.0737,
+      "step": 38767
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6880228571428575e-06,
+      "loss": 1.1297,
+      "step": 38768
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6877542857142857e-06,
+      "loss": 1.0938,
+      "step": 38769
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6874857142857147e-06,
+      "loss": 0.8992,
+      "step": 38770
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.687217142857143e-06,
+      "loss": 1.5722,
+      "step": 38771
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6869485714285715e-06,
+      "loss": 0.9037,
+      "step": 38772
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6866800000000005e-06,
+      "loss": 0.9696,
+      "step": 38773
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6864114285714287e-06,
+      "loss": 1.3809,
+      "step": 38774
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6861428571428577e-06,
+      "loss": 0.6934,
+      "step": 38775
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.685874285714286e-06,
+      "loss": 1.4088,
+      "step": 38776
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.685605714285714e-06,
+      "loss": 0.5867,
+      "step": 38777
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.685337142857143e-06,
+      "loss": 1.0714,
+      "step": 38778
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6850685714285717e-06,
+      "loss": 1.4575,
+      "step": 38779
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6848e-06,
+      "loss": 0.9124,
+      "step": 38780
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.684531428571429e-06,
+      "loss": 0.9644,
+      "step": 38781
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.684262857142857e-06,
+      "loss": 0.8783,
+      "step": 38782
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.683994285714286e-06,
+      "loss": 0.7513,
+      "step": 38783
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6837257142857143e-06,
+      "loss": 1.1497,
+      "step": 38784
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.683457142857143e-06,
+      "loss": 0.9376,
+      "step": 38785
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.683188571428572e-06,
+      "loss": 1.5355,
+      "step": 38786
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.68292e-06,
+      "loss": 1.4653,
+      "step": 38787
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.682651428571429e-06,
+      "loss": 1.1565,
+      "step": 38788
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6823828571428573e-06,
+      "loss": 1.2118,
+      "step": 38789
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6821142857142854e-06,
+      "loss": 0.9042,
+      "step": 38790
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6818457142857145e-06,
+      "loss": 1.1984,
+      "step": 38791
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.681577142857143e-06,
+      "loss": 1.0028,
+      "step": 38792
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.681308571428572e-06,
+      "loss": 1.0197,
+      "step": 38793
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6810400000000003e-06,
+      "loss": 1.3405,
+      "step": 38794
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6807714285714284e-06,
+      "loss": 1.0052,
+      "step": 38795
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6805028571428574e-06,
+      "loss": 1.1601,
+      "step": 38796
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6802342857142856e-06,
+      "loss": 0.8252,
+      "step": 38797
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6799657142857146e-06,
+      "loss": 1.2712,
+      "step": 38798
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6796971428571432e-06,
+      "loss": 1.4181,
+      "step": 38799
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6794285714285714e-06,
+      "loss": 0.9885,
+      "step": 38800
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6791600000000004e-06,
+      "loss": 0.9896,
+      "step": 38801
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6788914285714286e-06,
+      "loss": 1.0305,
+      "step": 38802
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.678622857142857e-06,
+      "loss": 0.5952,
+      "step": 38803
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.678354285714286e-06,
+      "loss": 0.9633,
+      "step": 38804
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6780857142857144e-06,
+      "loss": 1.0425,
+      "step": 38805
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6778171428571434e-06,
+      "loss": 1.2502,
+      "step": 38806
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6775485714285716e-06,
+      "loss": 0.875,
+      "step": 38807
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.67728e-06,
+      "loss": 0.8101,
+      "step": 38808
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.677011428571429e-06,
+      "loss": 1.5882,
+      "step": 38809
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.676742857142857e-06,
+      "loss": 1.0487,
+      "step": 38810
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.676474285714286e-06,
+      "loss": 1.0578,
+      "step": 38811
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6762057142857146e-06,
+      "loss": 0.9461,
+      "step": 38812
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.675937142857143e-06,
+      "loss": 1.2085,
+      "step": 38813
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.675668571428572e-06,
+      "loss": 1.3771,
+      "step": 38814
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6754e-06,
+      "loss": 0.819,
+      "step": 38815
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.675131428571429e-06,
+      "loss": 1.4985,
+      "step": 38816
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.674862857142857e-06,
+      "loss": 1.4669,
+      "step": 38817
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.674594285714286e-06,
+      "loss": 1.0031,
+      "step": 38818
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.674325714285715e-06,
+      "loss": 0.8773,
+      "step": 38819
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.674057142857143e-06,
+      "loss": 0.9978,
+      "step": 38820
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.673788571428572e-06,
+      "loss": 0.8647,
+      "step": 38821
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.67352e-06,
+      "loss": 0.6888,
+      "step": 38822
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6732514285714284e-06,
+      "loss": 1.2677,
+      "step": 38823
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6729828571428574e-06,
+      "loss": 1.0739,
+      "step": 38824
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.672714285714286e-06,
+      "loss": 1.2719,
+      "step": 38825
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.672445714285714e-06,
+      "loss": 1.0557,
+      "step": 38826
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.672177142857143e-06,
+      "loss": 0.9764,
+      "step": 38827
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6719085714285714e-06,
+      "loss": 0.7584,
+      "step": 38828
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6716400000000004e-06,
+      "loss": 1.2346,
+      "step": 38829
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6713714285714286e-06,
+      "loss": 1.5983,
+      "step": 38830
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6711028571428567e-06,
+      "loss": 1.0371,
+      "step": 38831
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.670834285714286e-06,
+      "loss": 1.2846,
+      "step": 38832
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6705657142857144e-06,
+      "loss": 1.1162,
+      "step": 38833
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6702971428571434e-06,
+      "loss": 0.8698,
+      "step": 38834
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6700285714285716e-06,
+      "loss": 1.7017,
+      "step": 38835
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6697599999999997e-06,
+      "loss": 1.1111,
+      "step": 38836
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6694914285714288e-06,
+      "loss": 0.9316,
+      "step": 38837
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6692228571428574e-06,
+      "loss": 1.1093,
+      "step": 38838
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6689542857142864e-06,
+      "loss": 0.9038,
+      "step": 38839
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6686857142857146e-06,
+      "loss": 1.1454,
+      "step": 38840
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6684171428571427e-06,
+      "loss": 0.8038,
+      "step": 38841
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6681485714285718e-06,
+      "loss": 1.1241,
+      "step": 38842
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.66788e-06,
+      "loss": 1.2623,
+      "step": 38843
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.667611428571428e-06,
+      "loss": 1.1951,
+      "step": 38844
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6673428571428576e-06,
+      "loss": 0.9011,
+      "step": 38845
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6670742857142857e-06,
+      "loss": 1.5885,
+      "step": 38846
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6668057142857148e-06,
+      "loss": 0.6449,
+      "step": 38847
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.666537142857143e-06,
+      "loss": 1.0795,
+      "step": 38848
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.666268571428571e-06,
+      "loss": 1.2269,
+      "step": 38849
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.666e-06,
+      "loss": 0.9868,
+      "step": 38850
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6657314285714287e-06,
+      "loss": 0.9761,
+      "step": 38851
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6654628571428578e-06,
+      "loss": 0.8885,
+      "step": 38852
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.665194285714286e-06,
+      "loss": 1.3517,
+      "step": 38853
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.664925714285714e-06,
+      "loss": 1.2071,
+      "step": 38854
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.664657142857143e-06,
+      "loss": 0.8203,
+      "step": 38855
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6643885714285713e-06,
+      "loss": 0.9945,
+      "step": 38856
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6641200000000003e-06,
+      "loss": 1.1373,
+      "step": 38857
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.663851428571429e-06,
+      "loss": 0.7191,
+      "step": 38858
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.663582857142857e-06,
+      "loss": 0.9631,
+      "step": 38859
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.663314285714286e-06,
+      "loss": 1.0387,
+      "step": 38860
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6630457142857143e-06,
+      "loss": 1.0385,
+      "step": 38861
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6627771428571433e-06,
+      "loss": 1.0248,
+      "step": 38862
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6625085714285715e-06,
+      "loss": 1.3215,
+      "step": 38863
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6622399999999997e-06,
+      "loss": 1.1474,
+      "step": 38864
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.661971428571429e-06,
+      "loss": 1.3316,
+      "step": 38865
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6617028571428573e-06,
+      "loss": 1.0675,
+      "step": 38866
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6614342857142855e-06,
+      "loss": 1.4527,
+      "step": 38867
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6611657142857145e-06,
+      "loss": 1.4798,
+      "step": 38868
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6608971428571427e-06,
+      "loss": 1.313,
+      "step": 38869
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6606285714285717e-06,
+      "loss": 1.0425,
+      "step": 38870
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6603600000000003e-06,
+      "loss": 0.9901,
+      "step": 38871
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6600914285714285e-06,
+      "loss": 0.8397,
+      "step": 38872
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6598228571428575e-06,
+      "loss": 0.9636,
+      "step": 38873
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6595542857142857e-06,
+      "loss": 1.5706,
+      "step": 38874
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6592857142857147e-06,
+      "loss": 1.0239,
+      "step": 38875
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.659017142857143e-06,
+      "loss": 1.2276,
+      "step": 38876
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.658748571428571e-06,
+      "loss": 1.1477,
+      "step": 38877
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6584800000000005e-06,
+      "loss": 0.9542,
+      "step": 38878
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6582114285714287e-06,
+      "loss": 0.6866,
+      "step": 38879
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6579428571428577e-06,
+      "loss": 1.4195,
+      "step": 38880
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.657674285714286e-06,
+      "loss": 1.3118,
+      "step": 38881
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.657405714285714e-06,
+      "loss": 1.2729,
+      "step": 38882
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.657137142857143e-06,
+      "loss": 1.4641,
+      "step": 38883
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6568685714285713e-06,
+      "loss": 1.047,
+      "step": 38884
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6566000000000007e-06,
+      "loss": 1.1209,
+      "step": 38885
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.656331428571429e-06,
+      "loss": 0.9908,
+      "step": 38886
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.656062857142857e-06,
+      "loss": 1.1381,
+      "step": 38887
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.655794285714286e-06,
+      "loss": 1.2228,
+      "step": 38888
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6555257142857143e-06,
+      "loss": 1.3103,
+      "step": 38889
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6552571428571424e-06,
+      "loss": 1.4664,
+      "step": 38890
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.654988571428572e-06,
+      "loss": 1.1422,
+      "step": 38891
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.65472e-06,
+      "loss": 1.3115,
+      "step": 38892
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.654451428571429e-06,
+      "loss": 1.0919,
+      "step": 38893
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6541828571428573e-06,
+      "loss": 1.1672,
+      "step": 38894
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6539142857142854e-06,
+      "loss": 1.1532,
+      "step": 38895
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6536457142857145e-06,
+      "loss": 0.8292,
+      "step": 38896
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6533771428571426e-06,
+      "loss": 1.2176,
+      "step": 38897
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.653108571428572e-06,
+      "loss": 1.0274,
+      "step": 38898
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6528400000000002e-06,
+      "loss": 0.8026,
+      "step": 38899
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6525714285714284e-06,
+      "loss": 0.9341,
+      "step": 38900
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6523028571428574e-06,
+      "loss": 1.1126,
+      "step": 38901
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6520342857142856e-06,
+      "loss": 1.0585,
+      "step": 38902
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6517657142857146e-06,
+      "loss": 1.3382,
+      "step": 38903
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6514971428571432e-06,
+      "loss": 1.0433,
+      "step": 38904
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6512285714285714e-06,
+      "loss": 1.2923,
+      "step": 38905
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6509600000000004e-06,
+      "loss": 1.0955,
+      "step": 38906
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6506914285714286e-06,
+      "loss": 1.1183,
+      "step": 38907
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6504228571428576e-06,
+      "loss": 1.0046,
+      "step": 38908
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.650154285714286e-06,
+      "loss": 0.8145,
+      "step": 38909
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.649885714285714e-06,
+      "loss": 1.3363,
+      "step": 38910
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6496171428571434e-06,
+      "loss": 1.322,
+      "step": 38911
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6493485714285716e-06,
+      "loss": 0.9842,
+      "step": 38912
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.64908e-06,
+      "loss": 1.1237,
+      "step": 38913
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.648811428571429e-06,
+      "loss": 1.1766,
+      "step": 38914
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.648542857142857e-06,
+      "loss": 1.2342,
+      "step": 38915
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.648274285714286e-06,
+      "loss": 0.9253,
+      "step": 38916
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.648005714285714e-06,
+      "loss": 1.0381,
+      "step": 38917
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.647737142857143e-06,
+      "loss": 1.1337,
+      "step": 38918
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.647468571428572e-06,
+      "loss": 1.0433,
+      "step": 38919
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6472e-06,
+      "loss": 1.6469,
+      "step": 38920
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.646931428571429e-06,
+      "loss": 1.4038,
+      "step": 38921
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.646662857142857e-06,
+      "loss": 1.3333,
+      "step": 38922
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6463942857142854e-06,
+      "loss": 0.9095,
+      "step": 38923
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.646125714285715e-06,
+      "loss": 1.1019,
+      "step": 38924
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.645857142857143e-06,
+      "loss": 1.0194,
+      "step": 38925
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.645588571428572e-06,
+      "loss": 1.2199,
+      "step": 38926
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.64532e-06,
+      "loss": 0.7482,
+      "step": 38927
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6450514285714284e-06,
+      "loss": 1.4502,
+      "step": 38928
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6447828571428574e-06,
+      "loss": 1.2855,
+      "step": 38929
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6445142857142856e-06,
+      "loss": 1.0992,
+      "step": 38930
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.644245714285714e-06,
+      "loss": 1.1723,
+      "step": 38931
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.643977142857143e-06,
+      "loss": 1.248,
+      "step": 38932
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6437085714285714e-06,
+      "loss": 1.4663,
+      "step": 38933
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6434400000000004e-06,
+      "loss": 0.8828,
+      "step": 38934
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6431714285714286e-06,
+      "loss": 1.215,
+      "step": 38935
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6429028571428567e-06,
+      "loss": 0.7082,
+      "step": 38936
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.6426342857142858e-06,
+      "loss": 0.9147,
+      "step": 38937
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6423657142857144e-06,
+      "loss": 1.2874,
+      "step": 38938
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6420971428571434e-06,
+      "loss": 1.3874,
+      "step": 38939
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6418285714285716e-06,
+      "loss": 0.9265,
+      "step": 38940
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6415599999999997e-06,
+      "loss": 0.9281,
+      "step": 38941
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6412914285714288e-06,
+      "loss": 1.4766,
+      "step": 38942
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.641022857142857e-06,
+      "loss": 1.4992,
+      "step": 38943
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6407542857142864e-06,
+      "loss": 0.9223,
+      "step": 38944
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6404857142857146e-06,
+      "loss": 1.1134,
+      "step": 38945
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6402171428571427e-06,
+      "loss": 1.3774,
+      "step": 38946
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6399485714285718e-06,
+      "loss": 1.2457,
+      "step": 38947
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.63968e-06,
+      "loss": 1.3626,
+      "step": 38948
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.639411428571429e-06,
+      "loss": 1.0708,
+      "step": 38949
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.639142857142857e-06,
+      "loss": 1.1963,
+      "step": 38950
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6388742857142857e-06,
+      "loss": 1.0243,
+      "step": 38951
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6386057142857148e-06,
+      "loss": 1.3784,
+      "step": 38952
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.638337142857143e-06,
+      "loss": 0.8692,
+      "step": 38953
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.638068571428571e-06,
+      "loss": 1.6012,
+      "step": 38954
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6378e-06,
+      "loss": 0.8675,
+      "step": 38955
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6375314285714283e-06,
+      "loss": 1.0186,
+      "step": 38956
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6372628571428578e-06,
+      "loss": 1.608,
+      "step": 38957
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.636994285714286e-06,
+      "loss": 1.1934,
+      "step": 38958
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.636725714285714e-06,
+      "loss": 1.3654,
+      "step": 38959
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.636457142857143e-06,
+      "loss": 1.4069,
+      "step": 38960
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6361885714285713e-06,
+      "loss": 1.0136,
+      "step": 38961
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6359200000000003e-06,
+      "loss": 1.1548,
+      "step": 38962
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6356514285714285e-06,
+      "loss": 0.8867,
+      "step": 38963
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.635382857142857e-06,
+      "loss": 1.078,
+      "step": 38964
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.635114285714286e-06,
+      "loss": 0.9805,
+      "step": 38965
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6348457142857143e-06,
+      "loss": 1.2435,
+      "step": 38966
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6345771428571433e-06,
+      "loss": 1.0165,
+      "step": 38967
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6343085714285715e-06,
+      "loss": 1.0231,
+      "step": 38968
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6340399999999997e-06,
+      "loss": 0.9941,
+      "step": 38969
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6337714285714287e-06,
+      "loss": 1.2704,
+      "step": 38970
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6335028571428573e-06,
+      "loss": 1.4367,
+      "step": 38971
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6332342857142863e-06,
+      "loss": 1.0347,
+      "step": 38972
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6329657142857145e-06,
+      "loss": 1.3117,
+      "step": 38973
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6326971428571427e-06,
+      "loss": 1.5179,
+      "step": 38974
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6324285714285717e-06,
+      "loss": 0.8604,
+      "step": 38975
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.63216e-06,
+      "loss": 1.0203,
+      "step": 38976
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6318914285714285e-06,
+      "loss": 0.8445,
+      "step": 38977
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6316228571428575e-06,
+      "loss": 0.8366,
+      "step": 38978
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6313542857142857e-06,
+      "loss": 1.4018,
+      "step": 38979
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6310857142857147e-06,
+      "loss": 1.0502,
+      "step": 38980
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.630817142857143e-06,
+      "loss": 1.0989,
+      "step": 38981
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.630548571428571e-06,
+      "loss": 1.1086,
+      "step": 38982
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.63028e-06,
+      "loss": 0.5641,
+      "step": 38983
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6300114285714287e-06,
+      "loss": 1.0318,
+      "step": 38984
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6297428571428577e-06,
+      "loss": 0.7449,
+      "step": 38985
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.629474285714286e-06,
+      "loss": 0.8126,
+      "step": 38986
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.629205714285714e-06,
+      "loss": 0.838,
+      "step": 38987
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.628937142857143e-06,
+      "loss": 1.2689,
+      "step": 38988
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6286685714285713e-06,
+      "loss": 1.0253,
+      "step": 38989
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6284000000000003e-06,
+      "loss": 1.4329,
+      "step": 38990
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.628131428571429e-06,
+      "loss": 1.0065,
+      "step": 38991
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.627862857142857e-06,
+      "loss": 1.2363,
+      "step": 38992
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.627594285714286e-06,
+      "loss": 1.2892,
+      "step": 38993
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6273257142857143e-06,
+      "loss": 0.6876,
+      "step": 38994
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6270571428571433e-06,
+      "loss": 0.9734,
+      "step": 38995
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6267885714285715e-06,
+      "loss": 1.1798,
+      "step": 38996
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.62652e-06,
+      "loss": 1.3089,
+      "step": 38997
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.626251428571429e-06,
+      "loss": 1.1709,
+      "step": 38998
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6259828571428573e-06,
+      "loss": 1.2219,
+      "step": 38999
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6257142857142854e-06,
+      "loss": 0.9381,
+      "step": 39000
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6254457142857144e-06,
+      "loss": 0.9467,
+      "step": 39001
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6251771428571426e-06,
+      "loss": 1.1732,
+      "step": 39002
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6249085714285716e-06,
+      "loss": 1.126,
+      "step": 39003
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6246400000000002e-06,
+      "loss": 1.1496,
+      "step": 39004
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6243714285714284e-06,
+      "loss": 0.9302,
+      "step": 39005
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6241028571428574e-06,
+      "loss": 1.4774,
+      "step": 39006
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6238342857142856e-06,
+      "loss": 1.273,
+      "step": 39007
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6235657142857146e-06,
+      "loss": 0.8636,
+      "step": 39008
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.623297142857143e-06,
+      "loss": 1.2493,
+      "step": 39009
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6230285714285714e-06,
+      "loss": 1.1036,
+      "step": 39010
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6227600000000004e-06,
+      "loss": 1.3703,
+      "step": 39011
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6224914285714286e-06,
+      "loss": 1.1514,
+      "step": 39012
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6222228571428576e-06,
+      "loss": 1.6497,
+      "step": 39013
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.621954285714286e-06,
+      "loss": 0.8855,
+      "step": 39014
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.621685714285714e-06,
+      "loss": 0.691,
+      "step": 39015
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.621417142857143e-06,
+      "loss": 1.3701,
+      "step": 39016
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6211485714285716e-06,
+      "loss": 1.0586,
+      "step": 39017
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.62088e-06,
+      "loss": 1.0568,
+      "step": 39018
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.620611428571429e-06,
+      "loss": 1.3723,
+      "step": 39019
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.620342857142857e-06,
+      "loss": 1.0081,
+      "step": 39020
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.620074285714286e-06,
+      "loss": 1.3461,
+      "step": 39021
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.619805714285714e-06,
+      "loss": 1.3502,
+      "step": 39022
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.619537142857143e-06,
+      "loss": 1.1801,
+      "step": 39023
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.619268571428572e-06,
+      "loss": 1.1079,
+      "step": 39024
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.619e-06,
+      "loss": 1.1705,
+      "step": 39025
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.618731428571429e-06,
+      "loss": 1.1789,
+      "step": 39026
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.618462857142857e-06,
+      "loss": 0.86,
+      "step": 39027
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6181942857142854e-06,
+      "loss": 0.8751,
+      "step": 39028
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6179257142857144e-06,
+      "loss": 1.374,
+      "step": 39029
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.617657142857143e-06,
+      "loss": 0.7463,
+      "step": 39030
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.617388571428572e-06,
+      "loss": 0.8749,
+      "step": 39031
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.61712e-06,
+      "loss": 1.2171,
+      "step": 39032
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6168514285714284e-06,
+      "loss": 1.7001,
+      "step": 39033
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6165828571428574e-06,
+      "loss": 0.9266,
+      "step": 39034
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6163142857142856e-06,
+      "loss": 0.9603,
+      "step": 39035
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6160457142857146e-06,
+      "loss": 1.67,
+      "step": 39036
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.615777142857143e-06,
+      "loss": 0.9491,
+      "step": 39037
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6155085714285714e-06,
+      "loss": 1.0683,
+      "step": 39038
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6152400000000004e-06,
+      "loss": 1.4653,
+      "step": 39039
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6149714285714286e-06,
+      "loss": 1.0182,
+      "step": 39040
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6147028571428567e-06,
+      "loss": 1.2645,
+      "step": 39041
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6144342857142858e-06,
+      "loss": 1.1114,
+      "step": 39042
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6141657142857144e-06,
+      "loss": 1.2166,
+      "step": 39043
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6138971428571434e-06,
+      "loss": 1.5615,
+      "step": 39044
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6136285714285716e-06,
+      "loss": 1.1367,
+      "step": 39045
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6133599999999997e-06,
+      "loss": 0.8652,
+      "step": 39046
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6130914285714288e-06,
+      "loss": 1.1362,
+      "step": 39047
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.612822857142857e-06,
+      "loss": 1.1019,
+      "step": 39048
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.612554285714286e-06,
+      "loss": 1.457,
+      "step": 39049
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6122857142857146e-06,
+      "loss": 1.2131,
+      "step": 39050
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6120171428571427e-06,
+      "loss": 1.2808,
+      "step": 39051
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6117485714285718e-06,
+      "loss": 0.9749,
+      "step": 39052
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.61148e-06,
+      "loss": 0.9198,
+      "step": 39053
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.611211428571429e-06,
+      "loss": 1.0346,
+      "step": 39054
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.610942857142857e-06,
+      "loss": 1.2241,
+      "step": 39055
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6106742857142857e-06,
+      "loss": 1.6302,
+      "step": 39056
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6104057142857148e-06,
+      "loss": 0.5675,
+      "step": 39057
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.610137142857143e-06,
+      "loss": 0.7943,
+      "step": 39058
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.609868571428572e-06,
+      "loss": 1.0765,
+      "step": 39059
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6096e-06,
+      "loss": 1.1926,
+      "step": 39060
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6093314285714283e-06,
+      "loss": 1.0685,
+      "step": 39061
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6090628571428573e-06,
+      "loss": 1.3915,
+      "step": 39062
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.608794285714286e-06,
+      "loss": 1.1712,
+      "step": 39063
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.608525714285714e-06,
+      "loss": 0.7331,
+      "step": 39064
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.608257142857143e-06,
+      "loss": 1.1794,
+      "step": 39065
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6079885714285713e-06,
+      "loss": 0.8561,
+      "step": 39066
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6077200000000003e-06,
+      "loss": 0.9969,
+      "step": 39067
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6074514285714285e-06,
+      "loss": 0.9653,
+      "step": 39068
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.607182857142857e-06,
+      "loss": 1.2559,
+      "step": 39069
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.606914285714286e-06,
+      "loss": 1.3451,
+      "step": 39070
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6066457142857143e-06,
+      "loss": 1.5607,
+      "step": 39071
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6063771428571433e-06,
+      "loss": 0.8655,
+      "step": 39072
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6061085714285715e-06,
+      "loss": 1.5416,
+      "step": 39073
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6058399999999997e-06,
+      "loss": 0.9871,
+      "step": 39074
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6055714285714287e-06,
+      "loss": 0.9567,
+      "step": 39075
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6053028571428573e-06,
+      "loss": 1.1183,
+      "step": 39076
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6050342857142863e-06,
+      "loss": 1.2296,
+      "step": 39077
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6047657142857145e-06,
+      "loss": 0.9003,
+      "step": 39078
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6044971428571427e-06,
+      "loss": 0.8152,
+      "step": 39079
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6042285714285717e-06,
+      "loss": 0.7754,
+      "step": 39080
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.60396e-06,
+      "loss": 1.2047,
+      "step": 39081
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.603691428571429e-06,
+      "loss": 0.9074,
+      "step": 39082
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6034228571428575e-06,
+      "loss": 1.3299,
+      "step": 39083
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6031542857142857e-06,
+      "loss": 1.1069,
+      "step": 39084
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6028857142857147e-06,
+      "loss": 1.1597,
+      "step": 39085
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.602617142857143e-06,
+      "loss": 0.9439,
+      "step": 39086
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.602348571428571e-06,
+      "loss": 1.2096,
+      "step": 39087
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.60208e-06,
+      "loss": 1.3976,
+      "step": 39088
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6018114285714287e-06,
+      "loss": 1.0775,
+      "step": 39089
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6015428571428577e-06,
+      "loss": 1.1944,
+      "step": 39090
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.601274285714286e-06,
+      "loss": 0.9921,
+      "step": 39091
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.601005714285714e-06,
+      "loss": 0.9216,
+      "step": 39092
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.600737142857143e-06,
+      "loss": 1.1922,
+      "step": 39093
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6004685714285713e-06,
+      "loss": 1.1557,
+      "step": 39094
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.6002000000000003e-06,
+      "loss": 0.7181,
+      "step": 39095
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.599931428571429e-06,
+      "loss": 1.5431,
+      "step": 39096
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.599662857142857e-06,
+      "loss": 1.1402,
+      "step": 39097
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.599394285714286e-06,
+      "loss": 1.2349,
+      "step": 39098
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5991257142857143e-06,
+      "loss": 1.0038,
+      "step": 39099
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5988571428571433e-06,
+      "loss": 1.2766,
+      "step": 39100
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5985885714285715e-06,
+      "loss": 0.9851,
+      "step": 39101
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.59832e-06,
+      "loss": 1.0546,
+      "step": 39102
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.598051428571429e-06,
+      "loss": 1.0279,
+      "step": 39103
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5977828571428572e-06,
+      "loss": 1.4162,
+      "step": 39104
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5975142857142854e-06,
+      "loss": 1.396,
+      "step": 39105
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5972457142857144e-06,
+      "loss": 1.0254,
+      "step": 39106
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5969771428571426e-06,
+      "loss": 0.9124,
+      "step": 39107
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5967085714285716e-06,
+      "loss": 1.4029,
+      "step": 39108
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5964400000000002e-06,
+      "loss": 1.3186,
+      "step": 39109
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5961714285714284e-06,
+      "loss": 1.4624,
+      "step": 39110
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5959028571428574e-06,
+      "loss": 1.079,
+      "step": 39111
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.5956342857142856e-06,
+      "loss": 0.947,
+      "step": 39112
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5953657142857146e-06,
+      "loss": 1.2953,
+      "step": 39113
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.595097142857143e-06,
+      "loss": 0.452,
+      "step": 39114
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5948285714285714e-06,
+      "loss": 0.9346,
+      "step": 39115
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5945600000000004e-06,
+      "loss": 1.2772,
+      "step": 39116
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5942914285714286e-06,
+      "loss": 1.1962,
+      "step": 39117
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5940228571428576e-06,
+      "loss": 0.9033,
+      "step": 39118
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.593754285714286e-06,
+      "loss": 1.1772,
+      "step": 39119
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.593485714285714e-06,
+      "loss": 1.6623,
+      "step": 39120
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.593217142857143e-06,
+      "loss": 1.3105,
+      "step": 39121
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5929485714285716e-06,
+      "loss": 1.4583,
+      "step": 39122
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5926800000000006e-06,
+      "loss": 0.9875,
+      "step": 39123
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.592411428571429e-06,
+      "loss": 0.6968,
+      "step": 39124
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.592142857142857e-06,
+      "loss": 1.1412,
+      "step": 39125
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.591874285714286e-06,
+      "loss": 0.9597,
+      "step": 39126
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.591605714285714e-06,
+      "loss": 1.4876,
+      "step": 39127
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.591337142857143e-06,
+      "loss": 0.7916,
+      "step": 39128
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.591068571428572e-06,
+      "loss": 1.0489,
+      "step": 39129
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5908e-06,
+      "loss": 1.0778,
+      "step": 39130
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.590531428571429e-06,
+      "loss": 0.9357,
+      "step": 39131
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.590262857142857e-06,
+      "loss": 0.8524,
+      "step": 39132
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5899942857142854e-06,
+      "loss": 1.3627,
+      "step": 39133
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5897257142857144e-06,
+      "loss": 1.2724,
+      "step": 39134
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.589457142857143e-06,
+      "loss": 1.1454,
+      "step": 39135
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.589188571428572e-06,
+      "loss": 0.8843,
+      "step": 39136
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.58892e-06,
+      "loss": 1.0304,
+      "step": 39137
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5886514285714284e-06,
+      "loss": 0.7469,
+      "step": 39138
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5883828571428574e-06,
+      "loss": 1.402,
+      "step": 39139
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5881142857142856e-06,
+      "loss": 0.9591,
+      "step": 39140
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5878457142857146e-06,
+      "loss": 1.2572,
+      "step": 39141
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.587577142857143e-06,
+      "loss": 1.0484,
+      "step": 39142
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5873085714285714e-06,
+      "loss": 0.9734,
+      "step": 39143
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5870400000000004e-06,
+      "loss": 0.7955,
+      "step": 39144
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5867714285714286e-06,
+      "loss": 1.5865,
+      "step": 39145
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5865028571428576e-06,
+      "loss": 1.4392,
+      "step": 39146
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5862342857142858e-06,
+      "loss": 1.1607,
+      "step": 39147
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5859657142857144e-06,
+      "loss": 1.2811,
+      "step": 39148
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5856971428571434e-06,
+      "loss": 1.0798,
+      "step": 39149
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5854285714285716e-06,
+      "loss": 1.2802,
+      "step": 39150
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5851599999999997e-06,
+      "loss": 0.8468,
+      "step": 39151
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5848914285714288e-06,
+      "loss": 0.8148,
+      "step": 39152
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.584622857142857e-06,
+      "loss": 1.3578,
+      "step": 39153
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.584354285714286e-06,
+      "loss": 1.2311,
+      "step": 39154
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5840857142857146e-06,
+      "loss": 1.2828,
+      "step": 39155
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5838171428571427e-06,
+      "loss": 0.949,
+      "step": 39156
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5835485714285718e-06,
+      "loss": 1.6584,
+      "step": 39157
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.58328e-06,
+      "loss": 1.0344,
+      "step": 39158
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.583011428571429e-06,
+      "loss": 1.105,
+      "step": 39159
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.582742857142857e-06,
+      "loss": 1.0587,
+      "step": 39160
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5824742857142857e-06,
+      "loss": 0.7574,
+      "step": 39161
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5822057142857148e-06,
+      "loss": 1.1406,
+      "step": 39162
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.581937142857143e-06,
+      "loss": 0.9061,
+      "step": 39163
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.581668571428572e-06,
+      "loss": 1.0284,
+      "step": 39164
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5814e-06,
+      "loss": 0.8505,
+      "step": 39165
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5811314285714283e-06,
+      "loss": 0.9593,
+      "step": 39166
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5808628571428573e-06,
+      "loss": 1.1211,
+      "step": 39167
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.580594285714286e-06,
+      "loss": 0.9599,
+      "step": 39168
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.580325714285714e-06,
+      "loss": 1.4588,
+      "step": 39169
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.580057142857143e-06,
+      "loss": 1.425,
+      "step": 39170
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5797885714285713e-06,
+      "loss": 0.9676,
+      "step": 39171
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5795200000000003e-06,
+      "loss": 1.1962,
+      "step": 39172
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5792514285714285e-06,
+      "loss": 1.1513,
+      "step": 39173
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.578982857142857e-06,
+      "loss": 1.2588,
+      "step": 39174
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.578714285714286e-06,
+      "loss": 1.5613,
+      "step": 39175
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5784457142857143e-06,
+      "loss": 1.4458,
+      "step": 39176
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5781771428571433e-06,
+      "loss": 0.7943,
+      "step": 39177
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5779085714285715e-06,
+      "loss": 0.9888,
+      "step": 39178
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5776399999999997e-06,
+      "loss": 1.3223,
+      "step": 39179
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5773714285714287e-06,
+      "loss": 0.7134,
+      "step": 39180
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5771028571428573e-06,
+      "loss": 1.0135,
+      "step": 39181
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5768342857142863e-06,
+      "loss": 1.0497,
+      "step": 39182
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5765657142857145e-06,
+      "loss": 0.9387,
+      "step": 39183
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5762971428571427e-06,
+      "loss": 1.3468,
+      "step": 39184
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5760285714285717e-06,
+      "loss": 1.1388,
+      "step": 39185
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.57576e-06,
+      "loss": 1.0682,
+      "step": 39186
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.575491428571429e-06,
+      "loss": 0.965,
+      "step": 39187
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5752228571428575e-06,
+      "loss": 0.9824,
+      "step": 39188
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5749542857142857e-06,
+      "loss": 1.6652,
+      "step": 39189
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5746857142857147e-06,
+      "loss": 0.7697,
+      "step": 39190
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.574417142857143e-06,
+      "loss": 1.1751,
+      "step": 39191
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.574148571428571e-06,
+      "loss": 1.1813,
+      "step": 39192
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.57388e-06,
+      "loss": 1.341,
+      "step": 39193
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5736114285714287e-06,
+      "loss": 0.9829,
+      "step": 39194
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5733428571428577e-06,
+      "loss": 1.1133,
+      "step": 39195
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.573074285714286e-06,
+      "loss": 1.4509,
+      "step": 39196
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.572805714285714e-06,
+      "loss": 0.8273,
+      "step": 39197
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.572537142857143e-06,
+      "loss": 1.0988,
+      "step": 39198
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5722685714285713e-06,
+      "loss": 1.0158,
+      "step": 39199
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5720000000000003e-06,
+      "loss": 0.8301,
+      "step": 39200
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.571731428571429e-06,
+      "loss": 0.8859,
+      "step": 39201
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.571462857142857e-06,
+      "loss": 1.0384,
+      "step": 39202
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.571194285714286e-06,
+      "loss": 1.1814,
+      "step": 39203
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5709257142857142e-06,
+      "loss": 1.2998,
+      "step": 39204
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5706571428571433e-06,
+      "loss": 1.0687,
+      "step": 39205
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5703885714285714e-06,
+      "loss": 0.9222,
+      "step": 39206
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.57012e-06,
+      "loss": 0.9059,
+      "step": 39207
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.569851428571429e-06,
+      "loss": 0.9854,
+      "step": 39208
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5695828571428572e-06,
+      "loss": 1.274,
+      "step": 39209
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5693142857142863e-06,
+      "loss": 1.4145,
+      "step": 39210
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5690457142857144e-06,
+      "loss": 1.0339,
+      "step": 39211
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5687771428571426e-06,
+      "loss": 1.6005,
+      "step": 39212
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5685085714285716e-06,
+      "loss": 1.2468,
+      "step": 39213
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5682400000000002e-06,
+      "loss": 1.1346,
+      "step": 39214
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5679714285714284e-06,
+      "loss": 0.9577,
+      "step": 39215
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5677028571428574e-06,
+      "loss": 1.0558,
+      "step": 39216
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5674342857142856e-06,
+      "loss": 0.8574,
+      "step": 39217
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5671657142857146e-06,
+      "loss": 1.0318,
+      "step": 39218
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.566897142857143e-06,
+      "loss": 0.9601,
+      "step": 39219
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5666285714285714e-06,
+      "loss": 1.1384,
+      "step": 39220
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5663600000000004e-06,
+      "loss": 1.3132,
+      "step": 39221
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5660914285714286e-06,
+      "loss": 0.9986,
+      "step": 39222
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5658228571428576e-06,
+      "loss": 1.6755,
+      "step": 39223
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.565554285714286e-06,
+      "loss": 0.9919,
+      "step": 39224
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.565285714285714e-06,
+      "loss": 0.6741,
+      "step": 39225
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.565017142857143e-06,
+      "loss": 1.1761,
+      "step": 39226
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5647485714285716e-06,
+      "loss": 1.1991,
+      "step": 39227
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5644800000000006e-06,
+      "loss": 1.0012,
+      "step": 39228
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.564211428571429e-06,
+      "loss": 1.3206,
+      "step": 39229
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.563942857142857e-06,
+      "loss": 0.756,
+      "step": 39230
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.563674285714286e-06,
+      "loss": 1.3553,
+      "step": 39231
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.563405714285714e-06,
+      "loss": 1.0401,
+      "step": 39232
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5631371428571432e-06,
+      "loss": 0.7162,
+      "step": 39233
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.562868571428572e-06,
+      "loss": 0.9735,
+      "step": 39234
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5626e-06,
+      "loss": 0.8862,
+      "step": 39235
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.562331428571429e-06,
+      "loss": 0.9723,
+      "step": 39236
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.562062857142857e-06,
+      "loss": 1.3339,
+      "step": 39237
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5617942857142854e-06,
+      "loss": 1.1047,
+      "step": 39238
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5615257142857144e-06,
+      "loss": 1.0635,
+      "step": 39239
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.561257142857143e-06,
+      "loss": 0.9887,
+      "step": 39240
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.560988571428572e-06,
+      "loss": 1.3481,
+      "step": 39241
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.56072e-06,
+      "loss": 1.0743,
+      "step": 39242
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5604514285714284e-06,
+      "loss": 1.1833,
+      "step": 39243
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5601828571428574e-06,
+      "loss": 1.0569,
+      "step": 39244
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5599142857142856e-06,
+      "loss": 1.0161,
+      "step": 39245
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5596457142857146e-06,
+      "loss": 0.7978,
+      "step": 39246
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.559377142857143e-06,
+      "loss": 1.1622,
+      "step": 39247
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5591085714285714e-06,
+      "loss": 0.8848,
+      "step": 39248
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5588400000000004e-06,
+      "loss": 1.0913,
+      "step": 39249
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5585714285714286e-06,
+      "loss": 1.0856,
+      "step": 39250
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5583028571428576e-06,
+      "loss": 1.4076,
+      "step": 39251
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5580342857142858e-06,
+      "loss": 1.1524,
+      "step": 39252
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5577657142857144e-06,
+      "loss": 0.9331,
+      "step": 39253
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5574971428571434e-06,
+      "loss": 0.9259,
+      "step": 39254
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5572285714285716e-06,
+      "loss": 1.0762,
+      "step": 39255
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5569599999999997e-06,
+      "loss": 1.2202,
+      "step": 39256
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5566914285714288e-06,
+      "loss": 1.1231,
+      "step": 39257
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.556422857142857e-06,
+      "loss": 1.1289,
+      "step": 39258
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.556154285714286e-06,
+      "loss": 0.8766,
+      "step": 39259
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5558857142857146e-06,
+      "loss": 1.0855,
+      "step": 39260
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5556171428571427e-06,
+      "loss": 1.3685,
+      "step": 39261
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5553485714285718e-06,
+      "loss": 0.9955,
+      "step": 39262
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.55508e-06,
+      "loss": 0.9667,
+      "step": 39263
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.554811428571429e-06,
+      "loss": 0.8755,
+      "step": 39264
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.554542857142857e-06,
+      "loss": 1.1352,
+      "step": 39265
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5542742857142857e-06,
+      "loss": 0.8557,
+      "step": 39266
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5540057142857148e-06,
+      "loss": 1.6172,
+      "step": 39267
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.553737142857143e-06,
+      "loss": 0.9886,
+      "step": 39268
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.553468571428572e-06,
+      "loss": 0.8896,
+      "step": 39269
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5532e-06,
+      "loss": 1.2133,
+      "step": 39270
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5529314285714283e-06,
+      "loss": 0.9729,
+      "step": 39271
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5526628571428573e-06,
+      "loss": 1.1463,
+      "step": 39272
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.552394285714286e-06,
+      "loss": 1.4765,
+      "step": 39273
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.552125714285715e-06,
+      "loss": 0.7113,
+      "step": 39274
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.551857142857143e-06,
+      "loss": 1.2522,
+      "step": 39275
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5515885714285713e-06,
+      "loss": 1.1742,
+      "step": 39276
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5513200000000003e-06,
+      "loss": 0.6337,
+      "step": 39277
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5510514285714285e-06,
+      "loss": 0.9856,
+      "step": 39278
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.550782857142857e-06,
+      "loss": 1.1956,
+      "step": 39279
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.550514285714286e-06,
+      "loss": 0.9458,
+      "step": 39280
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5502457142857143e-06,
+      "loss": 1.1455,
+      "step": 39281
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5499771428571433e-06,
+      "loss": 1.206,
+      "step": 39282
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5497085714285715e-06,
+      "loss": 0.8675,
+      "step": 39283
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5494399999999997e-06,
+      "loss": 1.8306,
+      "step": 39284
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5491714285714287e-06,
+      "loss": 1.4048,
+      "step": 39285
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5489028571428573e-06,
+      "loss": 1.3766,
+      "step": 39286
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.5486342857142863e-06,
+      "loss": 1.1464,
+      "step": 39287
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5483657142857145e-06,
+      "loss": 0.6388,
+      "step": 39288
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5480971428571427e-06,
+      "loss": 1.2294,
+      "step": 39289
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5478285714285717e-06,
+      "loss": 1.2129,
+      "step": 39290
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.54756e-06,
+      "loss": 1.4258,
+      "step": 39291
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.547291428571429e-06,
+      "loss": 1.0983,
+      "step": 39292
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5470228571428575e-06,
+      "loss": 1.2844,
+      "step": 39293
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5467542857142857e-06,
+      "loss": 0.9929,
+      "step": 39294
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5464857142857147e-06,
+      "loss": 1.4461,
+      "step": 39295
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.546217142857143e-06,
+      "loss": 1.1194,
+      "step": 39296
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.545948571428572e-06,
+      "loss": 0.7239,
+      "step": 39297
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.54568e-06,
+      "loss": 1.2718,
+      "step": 39298
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5454114285714287e-06,
+      "loss": 1.038,
+      "step": 39299
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5451428571428577e-06,
+      "loss": 0.7706,
+      "step": 39300
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.544874285714286e-06,
+      "loss": 0.5411,
+      "step": 39301
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.544605714285714e-06,
+      "loss": 0.983,
+      "step": 39302
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.544337142857143e-06,
+      "loss": 1.3844,
+      "step": 39303
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5440685714285713e-06,
+      "loss": 1.1901,
+      "step": 39304
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5438000000000003e-06,
+      "loss": 1.043,
+      "step": 39305
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.543531428571429e-06,
+      "loss": 1.177,
+      "step": 39306
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.543262857142857e-06,
+      "loss": 0.7898,
+      "step": 39307
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.542994285714286e-06,
+      "loss": 1.488,
+      "step": 39308
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5427257142857142e-06,
+      "loss": 1.2327,
+      "step": 39309
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5424571428571433e-06,
+      "loss": 1.2029,
+      "step": 39310
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5421885714285714e-06,
+      "loss": 1.1475,
+      "step": 39311
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5419199999999996e-06,
+      "loss": 0.7852,
+      "step": 39312
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.541651428571429e-06,
+      "loss": 1.4314,
+      "step": 39313
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5413828571428572e-06,
+      "loss": 1.3695,
+      "step": 39314
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5411142857142863e-06,
+      "loss": 1.2201,
+      "step": 39315
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5408457142857144e-06,
+      "loss": 1.4771,
+      "step": 39316
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5405771428571426e-06,
+      "loss": 1.1004,
+      "step": 39317
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5403085714285716e-06,
+      "loss": 1.1169,
+      "step": 39318
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5400400000000002e-06,
+      "loss": 1.0249,
+      "step": 39319
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5397714285714293e-06,
+      "loss": 1.1411,
+      "step": 39320
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5395028571428574e-06,
+      "loss": 1.2694,
+      "step": 39321
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5392342857142856e-06,
+      "loss": 0.8943,
+      "step": 39322
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5389657142857146e-06,
+      "loss": 1.2767,
+      "step": 39323
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.538697142857143e-06,
+      "loss": 1.1655,
+      "step": 39324
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.538428571428571e-06,
+      "loss": 1.1937,
+      "step": 39325
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5381600000000004e-06,
+      "loss": 1.4168,
+      "step": 39326
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5378914285714286e-06,
+      "loss": 1.2625,
+      "step": 39327
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5376228571428576e-06,
+      "loss": 1.123,
+      "step": 39328
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.537354285714286e-06,
+      "loss": 1.1782,
+      "step": 39329
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.537085714285714e-06,
+      "loss": 0.808,
+      "step": 39330
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.536817142857143e-06,
+      "loss": 0.7971,
+      "step": 39331
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5365485714285716e-06,
+      "loss": 1.1035,
+      "step": 39332
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5362800000000006e-06,
+      "loss": 1.0754,
+      "step": 39333
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.536011428571429e-06,
+      "loss": 1.0813,
+      "step": 39334
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.535742857142857e-06,
+      "loss": 1.0502,
+      "step": 39335
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.535474285714286e-06,
+      "loss": 1.3992,
+      "step": 39336
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.535205714285714e-06,
+      "loss": 1.021,
+      "step": 39337
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5349371428571432e-06,
+      "loss": 1.3909,
+      "step": 39338
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.534668571428572e-06,
+      "loss": 1.3463,
+      "step": 39339
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5344e-06,
+      "loss": 1.0005,
+      "step": 39340
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.534131428571429e-06,
+      "loss": 1.0326,
+      "step": 39341
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.533862857142857e-06,
+      "loss": 1.1692,
+      "step": 39342
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5335942857142854e-06,
+      "loss": 0.7382,
+      "step": 39343
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5333257142857144e-06,
+      "loss": 1.3352,
+      "step": 39344
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5330571428571426e-06,
+      "loss": 1.1003,
+      "step": 39345
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.532788571428572e-06,
+      "loss": 1.2297,
+      "step": 39346
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.53252e-06,
+      "loss": 1.1446,
+      "step": 39347
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5322514285714284e-06,
+      "loss": 0.7992,
+      "step": 39348
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5319828571428574e-06,
+      "loss": 1.3341,
+      "step": 39349
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5317142857142856e-06,
+      "loss": 1.1284,
+      "step": 39350
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5314457142857146e-06,
+      "loss": 1.2813,
+      "step": 39351
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.531177142857143e-06,
+      "loss": 1.147,
+      "step": 39352
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5309085714285714e-06,
+      "loss": 1.5086,
+      "step": 39353
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5306400000000004e-06,
+      "loss": 1.2838,
+      "step": 39354
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5303714285714286e-06,
+      "loss": 0.4972,
+      "step": 39355
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5301028571428576e-06,
+      "loss": 0.8337,
+      "step": 39356
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5298342857142858e-06,
+      "loss": 0.5519,
+      "step": 39357
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.529565714285714e-06,
+      "loss": 1.5729,
+      "step": 39358
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5292971428571434e-06,
+      "loss": 1.2321,
+      "step": 39359
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5290285714285716e-06,
+      "loss": 0.9704,
+      "step": 39360
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5287600000000006e-06,
+      "loss": 1.1505,
+      "step": 39361
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5284914285714288e-06,
+      "loss": 1.1599,
+      "step": 39362
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.528222857142857e-06,
+      "loss": 1.0252,
+      "step": 39363
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.527954285714286e-06,
+      "loss": 0.8473,
+      "step": 39364
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.527685714285714e-06,
+      "loss": 1.2809,
+      "step": 39365
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5274171428571427e-06,
+      "loss": 0.6662,
+      "step": 39366
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5271485714285718e-06,
+      "loss": 1.4996,
+      "step": 39367
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.52688e-06,
+      "loss": 1.4433,
+      "step": 39368
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.526611428571429e-06,
+      "loss": 1.4871,
+      "step": 39369
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.526342857142857e-06,
+      "loss": 1.3144,
+      "step": 39370
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5260742857142853e-06,
+      "loss": 0.7492,
+      "step": 39371
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5258057142857148e-06,
+      "loss": 1.2945,
+      "step": 39372
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.525537142857143e-06,
+      "loss": 1.0724,
+      "step": 39373
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.525268571428572e-06,
+      "loss": 0.7202,
+      "step": 39374
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.525e-06,
+      "loss": 1.1748,
+      "step": 39375
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5247314285714287e-06,
+      "loss": 1.0508,
+      "step": 39376
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5244628571428573e-06,
+      "loss": 0.8242,
+      "step": 39377
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5241942857142855e-06,
+      "loss": 0.6559,
+      "step": 39378
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5239257142857145e-06,
+      "loss": 0.8875,
+      "step": 39379
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.523657142857143e-06,
+      "loss": 1.2866,
+      "step": 39380
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5233885714285717e-06,
+      "loss": 1.3421,
+      "step": 39381
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.52312e-06,
+      "loss": 0.8991,
+      "step": 39382
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5228514285714285e-06,
+      "loss": 1.1269,
+      "step": 39383
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.522582857142857e-06,
+      "loss": 1.4316,
+      "step": 39384
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.522314285714286e-06,
+      "loss": 1.4682,
+      "step": 39385
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5220457142857147e-06,
+      "loss": 1.2619,
+      "step": 39386
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.521777142857143e-06,
+      "loss": 1.4332,
+      "step": 39387
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5215085714285715e-06,
+      "loss": 0.6517,
+      "step": 39388
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.52124e-06,
+      "loss": 1.2567,
+      "step": 39389
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5209714285714287e-06,
+      "loss": 1.3713,
+      "step": 39390
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5207028571428573e-06,
+      "loss": 1.0222,
+      "step": 39391
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.520434285714286e-06,
+      "loss": 0.9948,
+      "step": 39392
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5201657142857145e-06,
+      "loss": 1.1942,
+      "step": 39393
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.519897142857143e-06,
+      "loss": 1.0974,
+      "step": 39394
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5196285714285717e-06,
+      "loss": 0.8682,
+      "step": 39395
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.51936e-06,
+      "loss": 1.2903,
+      "step": 39396
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5190914285714285e-06,
+      "loss": 1.1041,
+      "step": 39397
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.518822857142857e-06,
+      "loss": 1.1159,
+      "step": 39398
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.518554285714286e-06,
+      "loss": 1.4617,
+      "step": 39399
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5182857142857147e-06,
+      "loss": 0.9809,
+      "step": 39400
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.518017142857143e-06,
+      "loss": 0.9579,
+      "step": 39401
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5177485714285715e-06,
+      "loss": 1.0876,
+      "step": 39402
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.51748e-06,
+      "loss": 1.3764,
+      "step": 39403
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5172114285714287e-06,
+      "loss": 1.366,
+      "step": 39404
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5169428571428573e-06,
+      "loss": 1.183,
+      "step": 39405
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.516674285714286e-06,
+      "loss": 1.2782,
+      "step": 39406
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5164057142857145e-06,
+      "loss": 1.0138,
+      "step": 39407
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.516137142857143e-06,
+      "loss": 1.2414,
+      "step": 39408
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5158685714285717e-06,
+      "loss": 1.5075,
+      "step": 39409
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5156e-06,
+      "loss": 1.3581,
+      "step": 39410
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5153314285714284e-06,
+      "loss": 1.3393,
+      "step": 39411
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5150628571428575e-06,
+      "loss": 0.6704,
+      "step": 39412
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.514794285714286e-06,
+      "loss": 0.9492,
+      "step": 39413
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5145257142857142e-06,
+      "loss": 1.1498,
+      "step": 39414
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.514257142857143e-06,
+      "loss": 0.6917,
+      "step": 39415
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5139885714285714e-06,
+      "loss": 0.7231,
+      "step": 39416
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.51372e-06,
+      "loss": 1.2103,
+      "step": 39417
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5134514285714286e-06,
+      "loss": 1.1688,
+      "step": 39418
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5131828571428572e-06,
+      "loss": 1.2105,
+      "step": 39419
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.512914285714286e-06,
+      "loss": 1.0764,
+      "step": 39420
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5126457142857144e-06,
+      "loss": 0.6361,
+      "step": 39421
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.512377142857143e-06,
+      "loss": 1.1877,
+      "step": 39422
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5121085714285716e-06,
+      "loss": 1.0539,
+      "step": 39423
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.51184e-06,
+      "loss": 0.889,
+      "step": 39424
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.511571428571429e-06,
+      "loss": 1.3037,
+      "step": 39425
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5113028571428574e-06,
+      "loss": 0.8782,
+      "step": 39426
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.511034285714286e-06,
+      "loss": 1.0903,
+      "step": 39427
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5107657142857142e-06,
+      "loss": 1.2869,
+      "step": 39428
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.510497142857143e-06,
+      "loss": 0.7776,
+      "step": 39429
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5102285714285714e-06,
+      "loss": 1.416,
+      "step": 39430
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.50996e-06,
+      "loss": 1.1553,
+      "step": 39431
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.509691428571429e-06,
+      "loss": 1.7072,
+      "step": 39432
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5094228571428572e-06,
+      "loss": 0.9593,
+      "step": 39433
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.509154285714286e-06,
+      "loss": 1.1159,
+      "step": 39434
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5088857142857144e-06,
+      "loss": 1.164,
+      "step": 39435
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.508617142857143e-06,
+      "loss": 1.0004,
+      "step": 39436
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.508348571428571e-06,
+      "loss": 1.299,
+      "step": 39437
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5080800000000002e-06,
+      "loss": 1.3617,
+      "step": 39438
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.507811428571429e-06,
+      "loss": 1.1384,
+      "step": 39439
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5075428571428574e-06,
+      "loss": 1.3272,
+      "step": 39440
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.507274285714286e-06,
+      "loss": 1.0416,
+      "step": 39441
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.507005714285714e-06,
+      "loss": 1.0613,
+      "step": 39442
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.506737142857143e-06,
+      "loss": 1.1662,
+      "step": 39443
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5064685714285714e-06,
+      "loss": 1.1672,
+      "step": 39444
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5062000000000004e-06,
+      "loss": 1.0113,
+      "step": 39445
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5059314285714286e-06,
+      "loss": 1.2108,
+      "step": 39446
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.505662857142857e-06,
+      "loss": 1.1965,
+      "step": 39447
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.505394285714286e-06,
+      "loss": 1.1229,
+      "step": 39448
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5051257142857144e-06,
+      "loss": 1.4094,
+      "step": 39449
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.504857142857143e-06,
+      "loss": 1.4862,
+      "step": 39450
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5045885714285716e-06,
+      "loss": 0.8634,
+      "step": 39451
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.50432e-06,
+      "loss": 1.1016,
+      "step": 39452
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5040514285714288e-06,
+      "loss": 1.2128,
+      "step": 39453
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5037828571428574e-06,
+      "loss": 0.8369,
+      "step": 39454
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.503514285714286e-06,
+      "loss": 0.9639,
+      "step": 39455
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.503245714285714e-06,
+      "loss": 1.089,
+      "step": 39456
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5029771428571428e-06,
+      "loss": 1.6084,
+      "step": 39457
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5027085714285718e-06,
+      "loss": 0.7767,
+      "step": 39458
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5024400000000004e-06,
+      "loss": 1.0386,
+      "step": 39459
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5021714285714286e-06,
+      "loss": 1.4404,
+      "step": 39460
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.501902857142857e-06,
+      "loss": 1.1425,
+      "step": 39461
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.5016342857142858e-06,
+      "loss": 1.1369,
+      "step": 39462
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.5013657142857144e-06,
+      "loss": 1.5443,
+      "step": 39463
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.501097142857143e-06,
+      "loss": 1.7286,
+      "step": 39464
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.5008285714285716e-06,
+      "loss": 0.7791,
+      "step": 39465
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.50056e-06,
+      "loss": 1.1399,
+      "step": 39466
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.5002914285714288e-06,
+      "loss": 1.6916,
+      "step": 39467
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.5000228571428574e-06,
+      "loss": 1.2647,
+      "step": 39468
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4997542857142855e-06,
+      "loss": 1.1863,
+      "step": 39469
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.499485714285714e-06,
+      "loss": 1.0974,
+      "step": 39470
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.499217142857143e-06,
+      "loss": 1.3462,
+      "step": 39471
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4989485714285718e-06,
+      "loss": 1.257,
+      "step": 39472
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4986800000000004e-06,
+      "loss": 1.3544,
+      "step": 39473
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4984114285714285e-06,
+      "loss": 1.1347,
+      "step": 39474
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.498142857142857e-06,
+      "loss": 1.1945,
+      "step": 39475
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4978742857142857e-06,
+      "loss": 1.4467,
+      "step": 39476
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4976057142857143e-06,
+      "loss": 1.0126,
+      "step": 39477
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4973371428571434e-06,
+      "loss": 1.5,
+      "step": 39478
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4970685714285715e-06,
+      "loss": 1.5093,
+      "step": 39479
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4968e-06,
+      "loss": 1.4641,
+      "step": 39480
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4965314285714287e-06,
+      "loss": 1.0367,
+      "step": 39481
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4962628571428573e-06,
+      "loss": 1.5884,
+      "step": 39482
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4959942857142855e-06,
+      "loss": 1.4804,
+      "step": 39483
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4957257142857145e-06,
+      "loss": 1.0575,
+      "step": 39484
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.495457142857143e-06,
+      "loss": 1.3205,
+      "step": 39485
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4951885714285717e-06,
+      "loss": 0.7186,
+      "step": 39486
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4949200000000003e-06,
+      "loss": 1.3404,
+      "step": 39487
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4946514285714285e-06,
+      "loss": 1.0057,
+      "step": 39488
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.494382857142857e-06,
+      "loss": 1.4421,
+      "step": 39489
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4941142857142857e-06,
+      "loss": 1.1737,
+      "step": 39490
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4938457142857147e-06,
+      "loss": 1.374,
+      "step": 39491
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.493577142857143e-06,
+      "loss": 1.3954,
+      "step": 39492
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4933085714285715e-06,
+      "loss": 1.2704,
+      "step": 39493
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.49304e-06,
+      "loss": 1.2664,
+      "step": 39494
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4927714285714287e-06,
+      "loss": 1.3961,
+      "step": 39495
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4925028571428573e-06,
+      "loss": 1.2932,
+      "step": 39496
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.492234285714286e-06,
+      "loss": 1.0448,
+      "step": 39497
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4919657142857145e-06,
+      "loss": 1.1311,
+      "step": 39498
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.491697142857143e-06,
+      "loss": 1.0986,
+      "step": 39499
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4914285714285717e-06,
+      "loss": 0.9542,
+      "step": 39500
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.49116e-06,
+      "loss": 0.7596,
+      "step": 39501
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4908914285714285e-06,
+      "loss": 1.0716,
+      "step": 39502
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.490622857142857e-06,
+      "loss": 1.3432,
+      "step": 39503
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.490354285714286e-06,
+      "loss": 1.7661,
+      "step": 39504
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4900857142857147e-06,
+      "loss": 1.4319,
+      "step": 39505
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.489817142857143e-06,
+      "loss": 1.2142,
+      "step": 39506
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4895485714285715e-06,
+      "loss": 1.0092,
+      "step": 39507
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.48928e-06,
+      "loss": 1.3928,
+      "step": 39508
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4890114285714287e-06,
+      "loss": 0.9984,
+      "step": 39509
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4887428571428573e-06,
+      "loss": 1.2366,
+      "step": 39510
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.488474285714286e-06,
+      "loss": 0.9923,
+      "step": 39511
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4882057142857145e-06,
+      "loss": 1.2349,
+      "step": 39512
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.487937142857143e-06,
+      "loss": 1.1484,
+      "step": 39513
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4876685714285717e-06,
+      "loss": 0.7747,
+      "step": 39514
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4874e-06,
+      "loss": 1.1065,
+      "step": 39515
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4871314285714284e-06,
+      "loss": 1.5146,
+      "step": 39516
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4868628571428575e-06,
+      "loss": 1.2173,
+      "step": 39517
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.486594285714286e-06,
+      "loss": 1.0649,
+      "step": 39518
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4863257142857147e-06,
+      "loss": 0.9604,
+      "step": 39519
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.486057142857143e-06,
+      "loss": 1.1337,
+      "step": 39520
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4857885714285714e-06,
+      "loss": 1.4513,
+      "step": 39521
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.48552e-06,
+      "loss": 0.7917,
+      "step": 39522
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4852514285714286e-06,
+      "loss": 0.9566,
+      "step": 39523
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4849828571428572e-06,
+      "loss": 0.9135,
+      "step": 39524
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.484714285714286e-06,
+      "loss": 1.1676,
+      "step": 39525
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4844457142857144e-06,
+      "loss": 1.2894,
+      "step": 39526
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.484177142857143e-06,
+      "loss": 1.2018,
+      "step": 39527
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4839085714285716e-06,
+      "loss": 1.2301,
+      "step": 39528
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.48364e-06,
+      "loss": 1.0653,
+      "step": 39529
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.483371428571429e-06,
+      "loss": 1.4107,
+      "step": 39530
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4831028571428574e-06,
+      "loss": 1.3506,
+      "step": 39531
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.482834285714286e-06,
+      "loss": 1.1353,
+      "step": 39532
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4825657142857142e-06,
+      "loss": 1.1075,
+      "step": 39533
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.482297142857143e-06,
+      "loss": 1.1382,
+      "step": 39534
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4820285714285714e-06,
+      "loss": 1.3348,
+      "step": 39535
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.48176e-06,
+      "loss": 1.1784,
+      "step": 39536
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.481491428571429e-06,
+      "loss": 1.3428,
+      "step": 39537
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4812228571428572e-06,
+      "loss": 0.9166,
+      "step": 39538
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.480954285714286e-06,
+      "loss": 0.9659,
+      "step": 39539
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4806857142857144e-06,
+      "loss": 1.0719,
+      "step": 39540
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.480417142857143e-06,
+      "loss": 0.5891,
+      "step": 39541
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4801485714285716e-06,
+      "loss": 1.0613,
+      "step": 39542
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4798800000000002e-06,
+      "loss": 0.8127,
+      "step": 39543
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.479611428571429e-06,
+      "loss": 1.446,
+      "step": 39544
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4793428571428574e-06,
+      "loss": 1.1896,
+      "step": 39545
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.479074285714286e-06,
+      "loss": 1.2738,
+      "step": 39546
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.478805714285714e-06,
+      "loss": 0.9719,
+      "step": 39547
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.478537142857143e-06,
+      "loss": 1.0587,
+      "step": 39548
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4782685714285714e-06,
+      "loss": 1.2364,
+      "step": 39549
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4780000000000004e-06,
+      "loss": 1.1804,
+      "step": 39550
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.477731428571429e-06,
+      "loss": 0.9761,
+      "step": 39551
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.477462857142857e-06,
+      "loss": 1.0793,
+      "step": 39552
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4771942857142858e-06,
+      "loss": 1.006,
+      "step": 39553
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4769257142857144e-06,
+      "loss": 1.509,
+      "step": 39554
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.476657142857143e-06,
+      "loss": 1.0171,
+      "step": 39555
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4763885714285716e-06,
+      "loss": 0.9614,
+      "step": 39556
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.47612e-06,
+      "loss": 1.2204,
+      "step": 39557
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4758514285714288e-06,
+      "loss": 0.9801,
+      "step": 39558
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4755828571428574e-06,
+      "loss": 1.1985,
+      "step": 39559
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.475314285714286e-06,
+      "loss": 1.2859,
+      "step": 39560
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.475045714285714e-06,
+      "loss": 0.8688,
+      "step": 39561
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4747771428571428e-06,
+      "loss": 0.8314,
+      "step": 39562
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4745085714285718e-06,
+      "loss": 1.5911,
+      "step": 39563
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4742400000000004e-06,
+      "loss": 1.0223,
+      "step": 39564
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4739714285714286e-06,
+      "loss": 1.3252,
+      "step": 39565
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.473702857142857e-06,
+      "loss": 1.3894,
+      "step": 39566
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4734342857142858e-06,
+      "loss": 1.2044,
+      "step": 39567
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4731657142857144e-06,
+      "loss": 1.5061,
+      "step": 39568
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.472897142857143e-06,
+      "loss": 1.5634,
+      "step": 39569
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4726285714285716e-06,
+      "loss": 0.973,
+      "step": 39570
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.47236e-06,
+      "loss": 1.1912,
+      "step": 39571
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4720914285714288e-06,
+      "loss": 1.2462,
+      "step": 39572
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4718228571428574e-06,
+      "loss": 1.2357,
+      "step": 39573
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.471554285714286e-06,
+      "loss": 1.4878,
+      "step": 39574
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.471285714285714e-06,
+      "loss": 0.8141,
+      "step": 39575
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.471017142857143e-06,
+      "loss": 1.1177,
+      "step": 39576
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4707485714285718e-06,
+      "loss": 0.9874,
+      "step": 39577
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4704800000000004e-06,
+      "loss": 1.1829,
+      "step": 39578
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4702114285714285e-06,
+      "loss": 1.0731,
+      "step": 39579
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.469942857142857e-06,
+      "loss": 0.8235,
+      "step": 39580
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4696742857142857e-06,
+      "loss": 0.8935,
+      "step": 39581
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4694057142857143e-06,
+      "loss": 0.7316,
+      "step": 39582
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4691371428571434e-06,
+      "loss": 1.1923,
+      "step": 39583
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4688685714285715e-06,
+      "loss": 1.1179,
+      "step": 39584
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4686e-06,
+      "loss": 1.1796,
+      "step": 39585
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4683314285714287e-06,
+      "loss": 1.1287,
+      "step": 39586
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4680628571428573e-06,
+      "loss": 1.2444,
+      "step": 39587
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4677942857142855e-06,
+      "loss": 1.3663,
+      "step": 39588
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.467525714285714e-06,
+      "loss": 0.9804,
+      "step": 39589
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.467257142857143e-06,
+      "loss": 1.0608,
+      "step": 39590
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4669885714285717e-06,
+      "loss": 0.8421,
+      "step": 39591
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4667200000000003e-06,
+      "loss": 0.7134,
+      "step": 39592
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4664514285714285e-06,
+      "loss": 1.0063,
+      "step": 39593
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.466182857142857e-06,
+      "loss": 1.3794,
+      "step": 39594
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4659142857142857e-06,
+      "loss": 1.4296,
+      "step": 39595
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4656457142857147e-06,
+      "loss": 1.6619,
+      "step": 39596
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4653771428571433e-06,
+      "loss": 1.234,
+      "step": 39597
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4651085714285715e-06,
+      "loss": 0.8288,
+      "step": 39598
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.46484e-06,
+      "loss": 1.1873,
+      "step": 39599
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4645714285714287e-06,
+      "loss": 1.0297,
+      "step": 39600
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4643028571428573e-06,
+      "loss": 1.1175,
+      "step": 39601
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4640342857142855e-06,
+      "loss": 1.5628,
+      "step": 39602
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4637657142857145e-06,
+      "loss": 1.1971,
+      "step": 39603
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.463497142857143e-06,
+      "loss": 1.1038,
+      "step": 39604
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4632285714285717e-06,
+      "loss": 1.2916,
+      "step": 39605
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4629600000000003e-06,
+      "loss": 1.0876,
+      "step": 39606
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4626914285714285e-06,
+      "loss": 0.8714,
+      "step": 39607
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.462422857142857e-06,
+      "loss": 1.328,
+      "step": 39608
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.462154285714286e-06,
+      "loss": 0.8833,
+      "step": 39609
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4618857142857147e-06,
+      "loss": 1.0321,
+      "step": 39610
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.461617142857143e-06,
+      "loss": 1.4631,
+      "step": 39611
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4613485714285715e-06,
+      "loss": 1.0166,
+      "step": 39612
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.46108e-06,
+      "loss": 1.2027,
+      "step": 39613
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4608114285714287e-06,
+      "loss": 1.0977,
+      "step": 39614
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4605428571428573e-06,
+      "loss": 1.4387,
+      "step": 39615
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.460274285714286e-06,
+      "loss": 1.1756,
+      "step": 39616
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4600057142857145e-06,
+      "loss": 1.0823,
+      "step": 39617
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.459737142857143e-06,
+      "loss": 1.1217,
+      "step": 39618
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4594685714285717e-06,
+      "loss": 1.1685,
+      "step": 39619
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4592e-06,
+      "loss": 1.6114,
+      "step": 39620
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4589314285714284e-06,
+      "loss": 1.1234,
+      "step": 39621
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.458662857142857e-06,
+      "loss": 0.8533,
+      "step": 39622
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.458394285714286e-06,
+      "loss": 1.4136,
+      "step": 39623
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4581257142857147e-06,
+      "loss": 1.1931,
+      "step": 39624
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.457857142857143e-06,
+      "loss": 1.6374,
+      "step": 39625
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4575885714285714e-06,
+      "loss": 1.0515,
+      "step": 39626
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.45732e-06,
+      "loss": 1.1808,
+      "step": 39627
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4570514285714286e-06,
+      "loss": 1.293,
+      "step": 39628
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4567828571428577e-06,
+      "loss": 1.0641,
+      "step": 39629
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.456514285714286e-06,
+      "loss": 1.4255,
+      "step": 39630
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4562457142857144e-06,
+      "loss": 1.3008,
+      "step": 39631
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.455977142857143e-06,
+      "loss": 0.8818,
+      "step": 39632
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4557085714285716e-06,
+      "loss": 1.2122,
+      "step": 39633
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.45544e-06,
+      "loss": 0.9552,
+      "step": 39634
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4551714285714284e-06,
+      "loss": 0.5807,
+      "step": 39635
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.4549028571428574e-06,
+      "loss": 0.9553,
+      "step": 39636
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.454634285714286e-06,
+      "loss": 1.1893,
+      "step": 39637
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4543657142857146e-06,
+      "loss": 1.3128,
+      "step": 39638
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.454097142857143e-06,
+      "loss": 0.9281,
+      "step": 39639
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4538285714285714e-06,
+      "loss": 1.0318,
+      "step": 39640
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.45356e-06,
+      "loss": 1.476,
+      "step": 39641
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4532914285714286e-06,
+      "loss": 1.2529,
+      "step": 39642
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4530228571428572e-06,
+      "loss": 0.9267,
+      "step": 39643
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.452754285714286e-06,
+      "loss": 1.3693,
+      "step": 39644
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4524857142857144e-06,
+      "loss": 1.2589,
+      "step": 39645
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.452217142857143e-06,
+      "loss": 1.8126,
+      "step": 39646
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4519485714285716e-06,
+      "loss": 1.2191,
+      "step": 39647
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.45168e-06,
+      "loss": 1.1021,
+      "step": 39648
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.451411428571429e-06,
+      "loss": 1.3304,
+      "step": 39649
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4511428571428574e-06,
+      "loss": 1.3281,
+      "step": 39650
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.450874285714286e-06,
+      "loss": 1.3332,
+      "step": 39651
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.450605714285714e-06,
+      "loss": 1.0864,
+      "step": 39652
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.450337142857143e-06,
+      "loss": 1.3201,
+      "step": 39653
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4500685714285714e-06,
+      "loss": 1.3634,
+      "step": 39654
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4498e-06,
+      "loss": 1.3312,
+      "step": 39655
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.449531428571429e-06,
+      "loss": 1.2714,
+      "step": 39656
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.449262857142857e-06,
+      "loss": 0.9959,
+      "step": 39657
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4489942857142858e-06,
+      "loss": 1.2872,
+      "step": 39658
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4487257142857144e-06,
+      "loss": 0.9847,
+      "step": 39659
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.448457142857143e-06,
+      "loss": 0.9743,
+      "step": 39660
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4481885714285716e-06,
+      "loss": 1.1599,
+      "step": 39661
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.44792e-06,
+      "loss": 0.909,
+      "step": 39662
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4476514285714288e-06,
+      "loss": 1.3926,
+      "step": 39663
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4473828571428574e-06,
+      "loss": 1.099,
+      "step": 39664
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.447114285714286e-06,
+      "loss": 1.4259,
+      "step": 39665
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.446845714285714e-06,
+      "loss": 1.1053,
+      "step": 39666
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4465771428571428e-06,
+      "loss": 1.1795,
+      "step": 39667
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4463085714285714e-06,
+      "loss": 1.1666,
+      "step": 39668
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4460400000000004e-06,
+      "loss": 1.3922,
+      "step": 39669
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.445771428571429e-06,
+      "loss": 1.3849,
+      "step": 39670
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.445502857142857e-06,
+      "loss": 0.9857,
+      "step": 39671
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4452342857142858e-06,
+      "loss": 1.1785,
+      "step": 39672
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4449657142857144e-06,
+      "loss": 1.572,
+      "step": 39673
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.444697142857143e-06,
+      "loss": 1.7897,
+      "step": 39674
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4444285714285716e-06,
+      "loss": 1.3968,
+      "step": 39675
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.44416e-06,
+      "loss": 1.3888,
+      "step": 39676
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4438914285714288e-06,
+      "loss": 1.068,
+      "step": 39677
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4436228571428574e-06,
+      "loss": 1.1446,
+      "step": 39678
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.443354285714286e-06,
+      "loss": 1.0019,
+      "step": 39679
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.443085714285714e-06,
+      "loss": 1.5059,
+      "step": 39680
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4428171428571427e-06,
+      "loss": 0.9678,
+      "step": 39681
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4425485714285718e-06,
+      "loss": 1.4873,
+      "step": 39682
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4422800000000004e-06,
+      "loss": 1.0789,
+      "step": 39683
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.442011428571429e-06,
+      "loss": 1.1342,
+      "step": 39684
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.441742857142857e-06,
+      "loss": 1.2833,
+      "step": 39685
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4414742857142857e-06,
+      "loss": 1.354,
+      "step": 39686
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4412057142857143e-06,
+      "loss": 0.9618,
+      "step": 39687
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.440937142857143e-06,
+      "loss": 0.6511,
+      "step": 39688
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4406685714285715e-06,
+      "loss": 0.8798,
+      "step": 39689
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4404e-06,
+      "loss": 1.6643,
+      "step": 39690
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4401314285714287e-06,
+      "loss": 0.887,
+      "step": 39691
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4398628571428573e-06,
+      "loss": 1.232,
+      "step": 39692
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.439594285714286e-06,
+      "loss": 1.2292,
+      "step": 39693
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.439325714285714e-06,
+      "loss": 1.1389,
+      "step": 39694
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.439057142857143e-06,
+      "loss": 0.6093,
+      "step": 39695
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4387885714285717e-06,
+      "loss": 1.1898,
+      "step": 39696
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4385200000000003e-06,
+      "loss": 1.6365,
+      "step": 39697
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4382514285714285e-06,
+      "loss": 1.4361,
+      "step": 39698
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.437982857142857e-06,
+      "loss": 1.2272,
+      "step": 39699
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4377142857142857e-06,
+      "loss": 1.225,
+      "step": 39700
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4374457142857143e-06,
+      "loss": 0.8258,
+      "step": 39701
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4371771428571433e-06,
+      "loss": 0.9284,
+      "step": 39702
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4369085714285715e-06,
+      "loss": 1.2501,
+      "step": 39703
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.43664e-06,
+      "loss": 1.1376,
+      "step": 39704
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4363714285714287e-06,
+      "loss": 1.1442,
+      "step": 39705
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4361028571428573e-06,
+      "loss": 1.0388,
+      "step": 39706
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4358342857142855e-06,
+      "loss": 1.0949,
+      "step": 39707
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4355657142857145e-06,
+      "loss": 1.2993,
+      "step": 39708
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.435297142857143e-06,
+      "loss": 0.9804,
+      "step": 39709
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4350285714285717e-06,
+      "loss": 1.0248,
+      "step": 39710
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4347600000000003e-06,
+      "loss": 1.4321,
+      "step": 39711
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4344914285714285e-06,
+      "loss": 1.0821,
+      "step": 39712
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.434222857142857e-06,
+      "loss": 1.2872,
+      "step": 39713
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4339542857142857e-06,
+      "loss": 0.7369,
+      "step": 39714
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4336857142857147e-06,
+      "loss": 1.0382,
+      "step": 39715
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4334171428571433e-06,
+      "loss": 1.3297,
+      "step": 39716
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4331485714285715e-06,
+      "loss": 1.3508,
+      "step": 39717
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.43288e-06,
+      "loss": 1.242,
+      "step": 39718
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4326114285714287e-06,
+      "loss": 1.1863,
+      "step": 39719
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4323428571428573e-06,
+      "loss": 1.2032,
+      "step": 39720
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.432074285714286e-06,
+      "loss": 1.0674,
+      "step": 39721
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4318057142857145e-06,
+      "loss": 1.4239,
+      "step": 39722
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.431537142857143e-06,
+      "loss": 0.9941,
+      "step": 39723
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4312685714285717e-06,
+      "loss": 1.157,
+      "step": 39724
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4310000000000003e-06,
+      "loss": 0.9959,
+      "step": 39725
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4307314285714284e-06,
+      "loss": 1.4943,
+      "step": 39726
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.430462857142857e-06,
+      "loss": 0.9148,
+      "step": 39727
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.430194285714286e-06,
+      "loss": 1.3591,
+      "step": 39728
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4299257142857147e-06,
+      "loss": 1.2985,
+      "step": 39729
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.429657142857143e-06,
+      "loss": 0.6978,
+      "step": 39730
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4293885714285714e-06,
+      "loss": 1.1819,
+      "step": 39731
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.42912e-06,
+      "loss": 1.4513,
+      "step": 39732
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4288514285714286e-06,
+      "loss": 1.1651,
+      "step": 39733
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4285828571428572e-06,
+      "loss": 1.1825,
+      "step": 39734
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.428314285714286e-06,
+      "loss": 0.7327,
+      "step": 39735
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4280457142857144e-06,
+      "loss": 0.7559,
+      "step": 39736
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.427777142857143e-06,
+      "loss": 1.0236,
+      "step": 39737
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4275085714285716e-06,
+      "loss": 1.0588,
+      "step": 39738
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.42724e-06,
+      "loss": 1.376,
+      "step": 39739
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4269714285714284e-06,
+      "loss": 1.2541,
+      "step": 39740
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4267028571428574e-06,
+      "loss": 1.2699,
+      "step": 39741
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.426434285714286e-06,
+      "loss": 1.2966,
+      "step": 39742
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4261657142857146e-06,
+      "loss": 1.1761,
+      "step": 39743
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.425897142857143e-06,
+      "loss": 1.0693,
+      "step": 39744
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4256285714285714e-06,
+      "loss": 0.7703,
+      "step": 39745
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.42536e-06,
+      "loss": 1.0958,
+      "step": 39746
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4250914285714286e-06,
+      "loss": 1.0725,
+      "step": 39747
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4248228571428576e-06,
+      "loss": 0.8756,
+      "step": 39748
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.424554285714286e-06,
+      "loss": 1.3376,
+      "step": 39749
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4242857142857144e-06,
+      "loss": 1.2871,
+      "step": 39750
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.424017142857143e-06,
+      "loss": 0.9005,
+      "step": 39751
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4237485714285716e-06,
+      "loss": 0.9835,
+      "step": 39752
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.42348e-06,
+      "loss": 1.0339,
+      "step": 39753
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.423211428571429e-06,
+      "loss": 1.2906,
+      "step": 39754
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4229428571428574e-06,
+      "loss": 1.443,
+      "step": 39755
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.422674285714286e-06,
+      "loss": 0.8436,
+      "step": 39756
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4224057142857146e-06,
+      "loss": 0.8311,
+      "step": 39757
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4221371428571428e-06,
+      "loss": 0.9702,
+      "step": 39758
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4218685714285714e-06,
+      "loss": 1.1038,
+      "step": 39759
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4216e-06,
+      "loss": 1.4659,
+      "step": 39760
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.421331428571429e-06,
+      "loss": 0.7928,
+      "step": 39761
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.421062857142857e-06,
+      "loss": 1.0884,
+      "step": 39762
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4207942857142858e-06,
+      "loss": 1.3225,
+      "step": 39763
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4205257142857144e-06,
+      "loss": 0.9559,
+      "step": 39764
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.420257142857143e-06,
+      "loss": 0.7965,
+      "step": 39765
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4199885714285716e-06,
+      "loss": 1.0343,
+      "step": 39766
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.41972e-06,
+      "loss": 0.9417,
+      "step": 39767
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4194514285714288e-06,
+      "loss": 1.3095,
+      "step": 39768
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4191828571428574e-06,
+      "loss": 1.2352,
+      "step": 39769
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.418914285714286e-06,
+      "loss": 0.7703,
+      "step": 39770
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.418645714285714e-06,
+      "loss": 0.9832,
+      "step": 39771
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4183771428571428e-06,
+      "loss": 1.0496,
+      "step": 39772
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4181085714285714e-06,
+      "loss": 1.5055,
+      "step": 39773
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4178400000000004e-06,
+      "loss": 1.0221,
+      "step": 39774
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.417571428571429e-06,
+      "loss": 0.9119,
+      "step": 39775
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.417302857142857e-06,
+      "loss": 1.2137,
+      "step": 39776
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4170342857142858e-06,
+      "loss": 0.88,
+      "step": 39777
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4167657142857144e-06,
+      "loss": 0.8756,
+      "step": 39778
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.416497142857143e-06,
+      "loss": 1.0835,
+      "step": 39779
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4162285714285716e-06,
+      "loss": 0.8042,
+      "step": 39780
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.41596e-06,
+      "loss": 1.3078,
+      "step": 39781
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4156914285714288e-06,
+      "loss": 1.2652,
+      "step": 39782
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4154228571428574e-06,
+      "loss": 1.4807,
+      "step": 39783
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.415154285714286e-06,
+      "loss": 1.4901,
+      "step": 39784
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.414885714285714e-06,
+      "loss": 1.1711,
+      "step": 39785
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4146171428571427e-06,
+      "loss": 1.0444,
+      "step": 39786
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4143485714285718e-06,
+      "loss": 1.0213,
+      "step": 39787
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4140800000000004e-06,
+      "loss": 0.9523,
+      "step": 39788
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.413811428571429e-06,
+      "loss": 0.9772,
+      "step": 39789
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.413542857142857e-06,
+      "loss": 1.1425,
+      "step": 39790
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4132742857142857e-06,
+      "loss": 1.2321,
+      "step": 39791
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4130057142857143e-06,
+      "loss": 1.1677,
+      "step": 39792
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.412737142857143e-06,
+      "loss": 1.2181,
+      "step": 39793
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4124685714285715e-06,
+      "loss": 1.3362,
+      "step": 39794
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4122e-06,
+      "loss": 0.5126,
+      "step": 39795
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4119314285714287e-06,
+      "loss": 0.8816,
+      "step": 39796
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4116628571428573e-06,
+      "loss": 0.9943,
+      "step": 39797
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.411394285714286e-06,
+      "loss": 1.054,
+      "step": 39798
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.411125714285714e-06,
+      "loss": 1.1252,
+      "step": 39799
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.410857142857143e-06,
+      "loss": 1.2794,
+      "step": 39800
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4105885714285717e-06,
+      "loss": 1.1507,
+      "step": 39801
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4103200000000003e-06,
+      "loss": 1.1262,
+      "step": 39802
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.410051428571429e-06,
+      "loss": 0.996,
+      "step": 39803
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.409782857142857e-06,
+      "loss": 0.7865,
+      "step": 39804
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4095142857142857e-06,
+      "loss": 1.4782,
+      "step": 39805
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4092457142857143e-06,
+      "loss": 1.0032,
+      "step": 39806
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4089771428571433e-06,
+      "loss": 0.8019,
+      "step": 39807
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4087085714285715e-06,
+      "loss": 0.948,
+      "step": 39808
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.40844e-06,
+      "loss": 1.4274,
+      "step": 39809
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4081714285714287e-06,
+      "loss": 0.7459,
+      "step": 39810
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.4079028571428573e-06,
+      "loss": 1.1756,
+      "step": 39811
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.407634285714286e-06,
+      "loss": 1.08,
+      "step": 39812
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4073657142857145e-06,
+      "loss": 1.4939,
+      "step": 39813
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.407097142857143e-06,
+      "loss": 1.2476,
+      "step": 39814
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4068285714285717e-06,
+      "loss": 1.4437,
+      "step": 39815
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4065600000000003e-06,
+      "loss": 1.2192,
+      "step": 39816
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4062914285714285e-06,
+      "loss": 0.8617,
+      "step": 39817
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.406022857142857e-06,
+      "loss": 0.9002,
+      "step": 39818
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4057542857142857e-06,
+      "loss": 1.1754,
+      "step": 39819
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4054857142857147e-06,
+      "loss": 1.3003,
+      "step": 39820
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4052171428571433e-06,
+      "loss": 1.22,
+      "step": 39821
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4049485714285715e-06,
+      "loss": 1.0314,
+      "step": 39822
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.40468e-06,
+      "loss": 1.1256,
+      "step": 39823
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4044114285714287e-06,
+      "loss": 1.3688,
+      "step": 39824
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4041428571428573e-06,
+      "loss": 1.7464,
+      "step": 39825
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4038742857142854e-06,
+      "loss": 1.2343,
+      "step": 39826
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4036057142857145e-06,
+      "loss": 0.9516,
+      "step": 39827
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.403337142857143e-06,
+      "loss": 0.9604,
+      "step": 39828
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4030685714285717e-06,
+      "loss": 0.8841,
+      "step": 39829
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4028000000000003e-06,
+      "loss": 0.9251,
+      "step": 39830
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4025314285714284e-06,
+      "loss": 1.0942,
+      "step": 39831
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.402262857142857e-06,
+      "loss": 0.9166,
+      "step": 39832
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.401994285714286e-06,
+      "loss": 0.9099,
+      "step": 39833
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4017257142857147e-06,
+      "loss": 1.4549,
+      "step": 39834
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4014571428571433e-06,
+      "loss": 0.9645,
+      "step": 39835
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4011885714285714e-06,
+      "loss": 0.6962,
+      "step": 39836
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.40092e-06,
+      "loss": 1.0766,
+      "step": 39837
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4006514285714286e-06,
+      "loss": 1.216,
+      "step": 39838
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.4003828571428572e-06,
+      "loss": 1.1507,
+      "step": 39839
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.400114285714286e-06,
+      "loss": 1.2076,
+      "step": 39840
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3998457142857144e-06,
+      "loss": 1.299,
+      "step": 39841
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.399577142857143e-06,
+      "loss": 1.3777,
+      "step": 39842
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3993085714285716e-06,
+      "loss": 1.0165,
+      "step": 39843
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3990400000000002e-06,
+      "loss": 1.2606,
+      "step": 39844
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3987714285714284e-06,
+      "loss": 1.1577,
+      "step": 39845
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.398502857142857e-06,
+      "loss": 0.9019,
+      "step": 39846
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.398234285714286e-06,
+      "loss": 0.9481,
+      "step": 39847
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3979657142857146e-06,
+      "loss": 1.1537,
+      "step": 39848
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.397697142857143e-06,
+      "loss": 1.0046,
+      "step": 39849
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3974285714285714e-06,
+      "loss": 0.7784,
+      "step": 39850
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.39716e-06,
+      "loss": 1.0534,
+      "step": 39851
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3968914285714286e-06,
+      "loss": 1.3023,
+      "step": 39852
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3966228571428576e-06,
+      "loss": 0.9592,
+      "step": 39853
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.396354285714286e-06,
+      "loss": 0.9762,
+      "step": 39854
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3960857142857144e-06,
+      "loss": 0.9952,
+      "step": 39855
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.395817142857143e-06,
+      "loss": 0.8431,
+      "step": 39856
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3955485714285716e-06,
+      "loss": 1.3438,
+      "step": 39857
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3952799999999998e-06,
+      "loss": 1.2887,
+      "step": 39858
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3950114285714284e-06,
+      "loss": 0.8279,
+      "step": 39859
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3947428571428574e-06,
+      "loss": 0.9907,
+      "step": 39860
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.394474285714286e-06,
+      "loss": 0.9337,
+      "step": 39861
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3942057142857146e-06,
+      "loss": 1.171,
+      "step": 39862
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3939371428571428e-06,
+      "loss": 1.2212,
+      "step": 39863
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3936685714285714e-06,
+      "loss": 1.015,
+      "step": 39864
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3934e-06,
+      "loss": 0.9539,
+      "step": 39865
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.393131428571429e-06,
+      "loss": 0.9028,
+      "step": 39866
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3928628571428576e-06,
+      "loss": 1.0838,
+      "step": 39867
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3925942857142858e-06,
+      "loss": 1.1487,
+      "step": 39868
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3923257142857144e-06,
+      "loss": 1.0585,
+      "step": 39869
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.392057142857143e-06,
+      "loss": 1.0435,
+      "step": 39870
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3917885714285716e-06,
+      "loss": 1.5619,
+      "step": 39871
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3915199999999998e-06,
+      "loss": 1.1387,
+      "step": 39872
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3912514285714288e-06,
+      "loss": 1.1971,
+      "step": 39873
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3909828571428574e-06,
+      "loss": 1.1868,
+      "step": 39874
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.390714285714286e-06,
+      "loss": 0.7979,
+      "step": 39875
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3904457142857146e-06,
+      "loss": 1.1794,
+      "step": 39876
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3901771428571428e-06,
+      "loss": 1.6085,
+      "step": 39877
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3899085714285714e-06,
+      "loss": 1.7043,
+      "step": 39878
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.38964e-06,
+      "loss": 1.4088,
+      "step": 39879
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.389371428571429e-06,
+      "loss": 1.0599,
+      "step": 39880
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.389102857142857e-06,
+      "loss": 1.6953,
+      "step": 39881
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3888342857142858e-06,
+      "loss": 1.0533,
+      "step": 39882
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3885657142857144e-06,
+      "loss": 1.1782,
+      "step": 39883
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.388297142857143e-06,
+      "loss": 1.167,
+      "step": 39884
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3880285714285716e-06,
+      "loss": 1.0395,
+      "step": 39885
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.38776e-06,
+      "loss": 0.8575,
+      "step": 39886
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3874914285714288e-06,
+      "loss": 1.7559,
+      "step": 39887
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3872228571428574e-06,
+      "loss": 1.2157,
+      "step": 39888
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.386954285714286e-06,
+      "loss": 0.6263,
+      "step": 39889
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3866857142857146e-06,
+      "loss": 1.0033,
+      "step": 39890
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3864171428571427e-06,
+      "loss": 0.9129,
+      "step": 39891
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3861485714285713e-06,
+      "loss": 1.0464,
+      "step": 39892
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3858800000000004e-06,
+      "loss": 1.4492,
+      "step": 39893
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.385611428571429e-06,
+      "loss": 0.9996,
+      "step": 39894
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.385342857142857e-06,
+      "loss": 1.1468,
+      "step": 39895
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3850742857142857e-06,
+      "loss": 1.0946,
+      "step": 39896
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3848057142857143e-06,
+      "loss": 1.3783,
+      "step": 39897
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.384537142857143e-06,
+      "loss": 1.7001,
+      "step": 39898
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3842685714285715e-06,
+      "loss": 1.1706,
+      "step": 39899
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.384e-06,
+      "loss": 1.4307,
+      "step": 39900
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3837314285714287e-06,
+      "loss": 1.4968,
+      "step": 39901
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3834628571428573e-06,
+      "loss": 1.0651,
+      "step": 39902
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.383194285714286e-06,
+      "loss": 1.188,
+      "step": 39903
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.382925714285714e-06,
+      "loss": 0.9736,
+      "step": 39904
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3826571428571427e-06,
+      "loss": 1.4119,
+      "step": 39905
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3823885714285717e-06,
+      "loss": 0.7552,
+      "step": 39906
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3821200000000003e-06,
+      "loss": 1.1306,
+      "step": 39907
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.381851428571429e-06,
+      "loss": 1.036,
+      "step": 39908
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.381582857142857e-06,
+      "loss": 0.7845,
+      "step": 39909
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3813142857142857e-06,
+      "loss": 0.7257,
+      "step": 39910
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3810457142857143e-06,
+      "loss": 1.0809,
+      "step": 39911
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.380777142857143e-06,
+      "loss": 0.7131,
+      "step": 39912
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3805085714285715e-06,
+      "loss": 0.9025,
+      "step": 39913
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.38024e-06,
+      "loss": 1.1177,
+      "step": 39914
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3799714285714287e-06,
+      "loss": 1.3662,
+      "step": 39915
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3797028571428573e-06,
+      "loss": 0.912,
+      "step": 39916
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.379434285714286e-06,
+      "loss": 1.3224,
+      "step": 39917
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.379165714285714e-06,
+      "loss": 1.2102,
+      "step": 39918
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.378897142857143e-06,
+      "loss": 1.045,
+      "step": 39919
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3786285714285717e-06,
+      "loss": 1.5352,
+      "step": 39920
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3783600000000003e-06,
+      "loss": 1.1715,
+      "step": 39921
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.378091428571429e-06,
+      "loss": 1.2398,
+      "step": 39922
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.377822857142857e-06,
+      "loss": 1.6698,
+      "step": 39923
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3775542857142857e-06,
+      "loss": 1.0432,
+      "step": 39924
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3772857142857143e-06,
+      "loss": 0.9416,
+      "step": 39925
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3770171428571433e-06,
+      "loss": 1.4638,
+      "step": 39926
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3767485714285715e-06,
+      "loss": 1.1937,
+      "step": 39927
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.37648e-06,
+      "loss": 1.5424,
+      "step": 39928
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3762114285714287e-06,
+      "loss": 1.1561,
+      "step": 39929
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3759428571428573e-06,
+      "loss": 1.3629,
+      "step": 39930
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.375674285714286e-06,
+      "loss": 0.9774,
+      "step": 39931
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3754057142857145e-06,
+      "loss": 1.2741,
+      "step": 39932
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.375137142857143e-06,
+      "loss": 1.0499,
+      "step": 39933
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3748685714285717e-06,
+      "loss": 0.9851,
+      "step": 39934
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3746000000000003e-06,
+      "loss": 1.2158,
+      "step": 39935
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3743314285714284e-06,
+      "loss": 0.9517,
+      "step": 39936
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.374062857142857e-06,
+      "loss": 0.9815,
+      "step": 39937
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3737942857142856e-06,
+      "loss": 1.1213,
+      "step": 39938
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3735257142857147e-06,
+      "loss": 0.8723,
+      "step": 39939
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3732571428571433e-06,
+      "loss": 1.3223,
+      "step": 39940
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3729885714285714e-06,
+      "loss": 0.6657,
+      "step": 39941
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.37272e-06,
+      "loss": 1.0784,
+      "step": 39942
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3724514285714286e-06,
+      "loss": 1.021,
+      "step": 39943
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3721828571428572e-06,
+      "loss": 1.2315,
+      "step": 39944
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.371914285714286e-06,
+      "loss": 0.8697,
+      "step": 39945
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3716457142857144e-06,
+      "loss": 1.0287,
+      "step": 39946
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.371377142857143e-06,
+      "loss": 1.2401,
+      "step": 39947
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3711085714285716e-06,
+      "loss": 1.1794,
+      "step": 39948
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3708400000000002e-06,
+      "loss": 1.2988,
+      "step": 39949
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3705714285714284e-06,
+      "loss": 1.2203,
+      "step": 39950
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.370302857142857e-06,
+      "loss": 1.2735,
+      "step": 39951
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.370034285714286e-06,
+      "loss": 1.1808,
+      "step": 39952
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3697657142857146e-06,
+      "loss": 1.0126,
+      "step": 39953
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3694971428571432e-06,
+      "loss": 1.2962,
+      "step": 39954
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3692285714285714e-06,
+      "loss": 1.2916,
+      "step": 39955
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.36896e-06,
+      "loss": 1.5641,
+      "step": 39956
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3686914285714286e-06,
+      "loss": 1.2468,
+      "step": 39957
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3684228571428572e-06,
+      "loss": 1.296,
+      "step": 39958
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.368154285714286e-06,
+      "loss": 0.8982,
+      "step": 39959
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3678857142857144e-06,
+      "loss": 0.9305,
+      "step": 39960
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.367617142857143e-06,
+      "loss": 1.2295,
+      "step": 39961
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3673485714285716e-06,
+      "loss": 1.5088,
+      "step": 39962
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.36708e-06,
+      "loss": 1.0047,
+      "step": 39963
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3668114285714284e-06,
+      "loss": 1.7268,
+      "step": 39964
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3665428571428574e-06,
+      "loss": 1.6033,
+      "step": 39965
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.366274285714286e-06,
+      "loss": 1.2438,
+      "step": 39966
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3660057142857146e-06,
+      "loss": 1.3849,
+      "step": 39967
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3657371428571428e-06,
+      "loss": 1.0652,
+      "step": 39968
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3654685714285714e-06,
+      "loss": 0.8556,
+      "step": 39969
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3652e-06,
+      "loss": 0.9929,
+      "step": 39970
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3649314285714286e-06,
+      "loss": 0.9567,
+      "step": 39971
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3646628571428576e-06,
+      "loss": 1.1958,
+      "step": 39972
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3643942857142858e-06,
+      "loss": 1.539,
+      "step": 39973
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3641257142857144e-06,
+      "loss": 1.1686,
+      "step": 39974
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.363857142857143e-06,
+      "loss": 1.4503,
+      "step": 39975
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3635885714285716e-06,
+      "loss": 1.1851,
+      "step": 39976
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.36332e-06,
+      "loss": 1.2582,
+      "step": 39977
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3630514285714288e-06,
+      "loss": 1.1189,
+      "step": 39978
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3627828571428574e-06,
+      "loss": 0.9563,
+      "step": 39979
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.362514285714286e-06,
+      "loss": 1.1913,
+      "step": 39980
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3622457142857146e-06,
+      "loss": 1.0266,
+      "step": 39981
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3619771428571428e-06,
+      "loss": 1.5441,
+      "step": 39982
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3617085714285714e-06,
+      "loss": 1.2038,
+      "step": 39983
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.36144e-06,
+      "loss": 1.0197,
+      "step": 39984
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.361171428571429e-06,
+      "loss": 0.8227,
+      "step": 39985
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3609028571428576e-06,
+      "loss": 0.9624,
+      "step": 39986
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.3606342857142858e-06,
+      "loss": 1.0355,
+      "step": 39987
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3603657142857144e-06,
+      "loss": 1.0402,
+      "step": 39988
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.360097142857143e-06,
+      "loss": 1.8314,
+      "step": 39989
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3598285714285716e-06,
+      "loss": 1.5651,
+      "step": 39990
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.35956e-06,
+      "loss": 0.955,
+      "step": 39991
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3592914285714288e-06,
+      "loss": 1.8472,
+      "step": 39992
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3590228571428574e-06,
+      "loss": 1.0051,
+      "step": 39993
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.358754285714286e-06,
+      "loss": 1.0513,
+      "step": 39994
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3584857142857146e-06,
+      "loss": 1.0084,
+      "step": 39995
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3582171428571427e-06,
+      "loss": 1.0071,
+      "step": 39996
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3579485714285713e-06,
+      "loss": 1.2965,
+      "step": 39997
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3576800000000004e-06,
+      "loss": 0.888,
+      "step": 39998
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.357411428571429e-06,
+      "loss": 1.1146,
+      "step": 39999
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.357142857142857e-06,
+      "loss": 0.884,
+      "step": 40000
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3568742857142857e-06,
+      "loss": 0.7772,
+      "step": 40001
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3566057142857143e-06,
+      "loss": 0.76,
+      "step": 40002
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.356337142857143e-06,
+      "loss": 1.1569,
+      "step": 40003
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3560685714285715e-06,
+      "loss": 0.8325,
+      "step": 40004
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3558e-06,
+      "loss": 0.9306,
+      "step": 40005
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3555314285714287e-06,
+      "loss": 1.0996,
+      "step": 40006
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3552628571428573e-06,
+      "loss": 0.7873,
+      "step": 40007
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.354994285714286e-06,
+      "loss": 0.7255,
+      "step": 40008
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3547257142857145e-06,
+      "loss": 1.0082,
+      "step": 40009
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3544571428571427e-06,
+      "loss": 1.3524,
+      "step": 40010
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3541885714285717e-06,
+      "loss": 1.3925,
+      "step": 40011
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3539200000000003e-06,
+      "loss": 0.9481,
+      "step": 40012
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.353651428571429e-06,
+      "loss": 1.1316,
+      "step": 40013
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.353382857142857e-06,
+      "loss": 1.0404,
+      "step": 40014
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3531142857142857e-06,
+      "loss": 0.7134,
+      "step": 40015
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3528457142857143e-06,
+      "loss": 0.8496,
+      "step": 40016
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.352577142857143e-06,
+      "loss": 1.2662,
+      "step": 40017
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.352308571428572e-06,
+      "loss": 0.5669,
+      "step": 40018
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.35204e-06,
+      "loss": 1.495,
+      "step": 40019
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3517714285714287e-06,
+      "loss": 1.2461,
+      "step": 40020
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3515028571428573e-06,
+      "loss": 1.22,
+      "step": 40021
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.351234285714286e-06,
+      "loss": 1.1819,
+      "step": 40022
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.350965714285714e-06,
+      "loss": 0.8705,
+      "step": 40023
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.350697142857143e-06,
+      "loss": 0.9571,
+      "step": 40024
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3504285714285717e-06,
+      "loss": 0.7739,
+      "step": 40025
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3501600000000003e-06,
+      "loss": 1.0595,
+      "step": 40026
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.349891428571429e-06,
+      "loss": 1.2858,
+      "step": 40027
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.349622857142857e-06,
+      "loss": 1.3233,
+      "step": 40028
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3493542857142857e-06,
+      "loss": 1.422,
+      "step": 40029
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3490857142857143e-06,
+      "loss": 0.9559,
+      "step": 40030
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3488171428571433e-06,
+      "loss": 0.9885,
+      "step": 40031
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3485485714285715e-06,
+      "loss": 1.7328,
+      "step": 40032
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.34828e-06,
+      "loss": 1.5187,
+      "step": 40033
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3480114285714287e-06,
+      "loss": 1.314,
+      "step": 40034
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3477428571428573e-06,
+      "loss": 0.9285,
+      "step": 40035
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.347474285714286e-06,
+      "loss": 1.3728,
+      "step": 40036
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3472057142857145e-06,
+      "loss": 1.4335,
+      "step": 40037
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.346937142857143e-06,
+      "loss": 1.2369,
+      "step": 40038
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3466685714285717e-06,
+      "loss": 1.1476,
+      "step": 40039
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3464000000000003e-06,
+      "loss": 0.9366,
+      "step": 40040
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.346131428571429e-06,
+      "loss": 0.8926,
+      "step": 40041
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.345862857142857e-06,
+      "loss": 1.102,
+      "step": 40042
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3455942857142856e-06,
+      "loss": 1.0088,
+      "step": 40043
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3453257142857147e-06,
+      "loss": 1.1671,
+      "step": 40044
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3450571428571433e-06,
+      "loss": 1.2336,
+      "step": 40045
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3447885714285714e-06,
+      "loss": 1.1055,
+      "step": 40046
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.34452e-06,
+      "loss": 1.4566,
+      "step": 40047
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3442514285714286e-06,
+      "loss": 1.0755,
+      "step": 40048
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3439828571428572e-06,
+      "loss": 1.3331,
+      "step": 40049
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.343714285714286e-06,
+      "loss": 0.8911,
+      "step": 40050
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3434457142857144e-06,
+      "loss": 0.9288,
+      "step": 40051
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.343177142857143e-06,
+      "loss": 1.1515,
+      "step": 40052
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3429085714285716e-06,
+      "loss": 1.035,
+      "step": 40053
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3426400000000002e-06,
+      "loss": 0.9205,
+      "step": 40054
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3423714285714284e-06,
+      "loss": 1.1718,
+      "step": 40055
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.342102857142857e-06,
+      "loss": 1.1224,
+      "step": 40056
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.341834285714286e-06,
+      "loss": 0.871,
+      "step": 40057
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3415657142857146e-06,
+      "loss": 1.3574,
+      "step": 40058
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3412971428571432e-06,
+      "loss": 1.2862,
+      "step": 40059
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3410285714285714e-06,
+      "loss": 0.8039,
+      "step": 40060
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.34076e-06,
+      "loss": 1.2698,
+      "step": 40061
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3404914285714286e-06,
+      "loss": 1.2398,
+      "step": 40062
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.340222857142857e-06,
+      "loss": 1.0173,
+      "step": 40063
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.339954285714286e-06,
+      "loss": 1.1202,
+      "step": 40064
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3396857142857144e-06,
+      "loss": 1.1753,
+      "step": 40065
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.339417142857143e-06,
+      "loss": 1.2067,
+      "step": 40066
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3391485714285716e-06,
+      "loss": 0.7476,
+      "step": 40067
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.33888e-06,
+      "loss": 0.7811,
+      "step": 40068
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3386114285714284e-06,
+      "loss": 1.1813,
+      "step": 40069
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.338342857142857e-06,
+      "loss": 1.428,
+      "step": 40070
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.338074285714286e-06,
+      "loss": 1.3206,
+      "step": 40071
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3378057142857146e-06,
+      "loss": 1.3654,
+      "step": 40072
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.337537142857143e-06,
+      "loss": 1.0706,
+      "step": 40073
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3372685714285714e-06,
+      "loss": 1.2174,
+      "step": 40074
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.337e-06,
+      "loss": 1.3893,
+      "step": 40075
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3367314285714286e-06,
+      "loss": 0.6591,
+      "step": 40076
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3364628571428576e-06,
+      "loss": 0.8938,
+      "step": 40077
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3361942857142858e-06,
+      "loss": 1.1094,
+      "step": 40078
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3359257142857144e-06,
+      "loss": 1.472,
+      "step": 40079
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.335657142857143e-06,
+      "loss": 1.0899,
+      "step": 40080
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3353885714285716e-06,
+      "loss": 0.859,
+      "step": 40081
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.33512e-06,
+      "loss": 1.317,
+      "step": 40082
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3348514285714284e-06,
+      "loss": 0.7074,
+      "step": 40083
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3345828571428574e-06,
+      "loss": 0.674,
+      "step": 40084
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.334314285714286e-06,
+      "loss": 0.953,
+      "step": 40085
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3340457142857146e-06,
+      "loss": 1.455,
+      "step": 40086
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3337771428571428e-06,
+      "loss": 0.6943,
+      "step": 40087
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3335085714285714e-06,
+      "loss": 1.1912,
+      "step": 40088
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.33324e-06,
+      "loss": 0.8528,
+      "step": 40089
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.332971428571429e-06,
+      "loss": 1.5587,
+      "step": 40090
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3327028571428576e-06,
+      "loss": 0.9425,
+      "step": 40091
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3324342857142858e-06,
+      "loss": 0.9193,
+      "step": 40092
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3321657142857144e-06,
+      "loss": 1.16,
+      "step": 40093
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.331897142857143e-06,
+      "loss": 1.1685,
+      "step": 40094
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3316285714285716e-06,
+      "loss": 0.8632,
+      "step": 40095
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.33136e-06,
+      "loss": 0.9623,
+      "step": 40096
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3310914285714288e-06,
+      "loss": 0.7322,
+      "step": 40097
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3308228571428574e-06,
+      "loss": 0.9584,
+      "step": 40098
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.330554285714286e-06,
+      "loss": 0.8437,
+      "step": 40099
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3302857142857146e-06,
+      "loss": 1.2059,
+      "step": 40100
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3300171428571427e-06,
+      "loss": 1.2955,
+      "step": 40101
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3297485714285713e-06,
+      "loss": 1.1145,
+      "step": 40102
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.32948e-06,
+      "loss": 1.0983,
+      "step": 40103
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.329211428571429e-06,
+      "loss": 1.5914,
+      "step": 40104
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3289428571428575e-06,
+      "loss": 0.9862,
+      "step": 40105
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3286742857142857e-06,
+      "loss": 1.1013,
+      "step": 40106
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3284057142857143e-06,
+      "loss": 1.1328,
+      "step": 40107
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.328137142857143e-06,
+      "loss": 1.2182,
+      "step": 40108
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3278685714285715e-06,
+      "loss": 1.1633,
+      "step": 40109
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3276e-06,
+      "loss": 1.0495,
+      "step": 40110
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3273314285714287e-06,
+      "loss": 0.9092,
+      "step": 40111
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3270628571428573e-06,
+      "loss": 1.1102,
+      "step": 40112
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.326794285714286e-06,
+      "loss": 0.8505,
+      "step": 40113
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3265257142857145e-06,
+      "loss": 1.1142,
+      "step": 40114
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3262571428571427e-06,
+      "loss": 1.2036,
+      "step": 40115
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3259885714285713e-06,
+      "loss": 0.9786,
+      "step": 40116
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3257200000000003e-06,
+      "loss": 0.8422,
+      "step": 40117
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.325451428571429e-06,
+      "loss": 1.2917,
+      "step": 40118
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.325182857142857e-06,
+      "loss": 1.2281,
+      "step": 40119
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3249142857142857e-06,
+      "loss": 1.1642,
+      "step": 40120
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3246457142857143e-06,
+      "loss": 0.8624,
+      "step": 40121
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.324377142857143e-06,
+      "loss": 1.1685,
+      "step": 40122
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3241085714285715e-06,
+      "loss": 1.4621,
+      "step": 40123
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.32384e-06,
+      "loss": 1.2967,
+      "step": 40124
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3235714285714287e-06,
+      "loss": 0.6604,
+      "step": 40125
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3233028571428573e-06,
+      "loss": 1.0081,
+      "step": 40126
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.323034285714286e-06,
+      "loss": 1.3616,
+      "step": 40127
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3227657142857145e-06,
+      "loss": 1.4311,
+      "step": 40128
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3224971428571427e-06,
+      "loss": 1.5569,
+      "step": 40129
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3222285714285717e-06,
+      "loss": 1.0762,
+      "step": 40130
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3219600000000003e-06,
+      "loss": 1.2143,
+      "step": 40131
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.321691428571429e-06,
+      "loss": 1.6345,
+      "step": 40132
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.321422857142857e-06,
+      "loss": 1.047,
+      "step": 40133
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3211542857142857e-06,
+      "loss": 1.3097,
+      "step": 40134
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3208857142857143e-06,
+      "loss": 1.4168,
+      "step": 40135
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.320617142857143e-06,
+      "loss": 0.9231,
+      "step": 40136
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.320348571428572e-06,
+      "loss": 1.2569,
+      "step": 40137
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.32008e-06,
+      "loss": 1.2069,
+      "step": 40138
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3198114285714287e-06,
+      "loss": 1.6153,
+      "step": 40139
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3195428571428573e-06,
+      "loss": 1.1222,
+      "step": 40140
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.319274285714286e-06,
+      "loss": 1.0872,
+      "step": 40141
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.319005714285714e-06,
+      "loss": 1.358,
+      "step": 40142
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.318737142857143e-06,
+      "loss": 1.1026,
+      "step": 40143
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3184685714285717e-06,
+      "loss": 0.933,
+      "step": 40144
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3182000000000003e-06,
+      "loss": 1.292,
+      "step": 40145
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.317931428571429e-06,
+      "loss": 1.3722,
+      "step": 40146
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.317662857142857e-06,
+      "loss": 1.055,
+      "step": 40147
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3173942857142856e-06,
+      "loss": 1.5989,
+      "step": 40148
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3171257142857142e-06,
+      "loss": 1.0957,
+      "step": 40149
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3168571428571433e-06,
+      "loss": 1.0282,
+      "step": 40150
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3165885714285714e-06,
+      "loss": 1.3393,
+      "step": 40151
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.31632e-06,
+      "loss": 0.9902,
+      "step": 40152
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3160514285714286e-06,
+      "loss": 1.1598,
+      "step": 40153
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3157828571428572e-06,
+      "loss": 0.8629,
+      "step": 40154
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.315514285714286e-06,
+      "loss": 1.1776,
+      "step": 40155
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3152457142857144e-06,
+      "loss": 1.0934,
+      "step": 40156
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.314977142857143e-06,
+      "loss": 1.2205,
+      "step": 40157
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3147085714285716e-06,
+      "loss": 1.1342,
+      "step": 40158
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3144400000000002e-06,
+      "loss": 1.2804,
+      "step": 40159
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.314171428571429e-06,
+      "loss": 1.2059,
+      "step": 40160
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.313902857142857e-06,
+      "loss": 0.8975,
+      "step": 40161
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.3136342857142856e-06,
+      "loss": 0.8861,
+      "step": 40162
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3133657142857146e-06,
+      "loss": 1.1383,
+      "step": 40163
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3130971428571432e-06,
+      "loss": 0.7858,
+      "step": 40164
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3128285714285714e-06,
+      "loss": 1.0426,
+      "step": 40165
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.31256e-06,
+      "loss": 1.1868,
+      "step": 40166
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3122914285714286e-06,
+      "loss": 1.02,
+      "step": 40167
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.312022857142857e-06,
+      "loss": 1.0257,
+      "step": 40168
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.311754285714286e-06,
+      "loss": 1.0828,
+      "step": 40169
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3114857142857144e-06,
+      "loss": 0.8145,
+      "step": 40170
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.311217142857143e-06,
+      "loss": 1.3499,
+      "step": 40171
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3109485714285716e-06,
+      "loss": 1.3205,
+      "step": 40172
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.31068e-06,
+      "loss": 1.3537,
+      "step": 40173
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3104114285714284e-06,
+      "loss": 1.0027,
+      "step": 40174
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.310142857142857e-06,
+      "loss": 0.8356,
+      "step": 40175
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.309874285714286e-06,
+      "loss": 1.5583,
+      "step": 40176
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3096057142857146e-06,
+      "loss": 1.2264,
+      "step": 40177
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.309337142857143e-06,
+      "loss": 0.9453,
+      "step": 40178
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3090685714285714e-06,
+      "loss": 0.8241,
+      "step": 40179
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3088e-06,
+      "loss": 0.9645,
+      "step": 40180
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3085314285714286e-06,
+      "loss": 1.2607,
+      "step": 40181
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.308262857142857e-06,
+      "loss": 1.0188,
+      "step": 40182
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.307994285714286e-06,
+      "loss": 1.0317,
+      "step": 40183
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3077257142857144e-06,
+      "loss": 0.9582,
+      "step": 40184
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.307457142857143e-06,
+      "loss": 1.1944,
+      "step": 40185
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3071885714285716e-06,
+      "loss": 1.1122,
+      "step": 40186
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.30692e-06,
+      "loss": 1.3919,
+      "step": 40187
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3066514285714284e-06,
+      "loss": 1.1657,
+      "step": 40188
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3063828571428574e-06,
+      "loss": 1.3859,
+      "step": 40189
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.306114285714286e-06,
+      "loss": 1.6015,
+      "step": 40190
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3058457142857146e-06,
+      "loss": 1.1743,
+      "step": 40191
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.305577142857143e-06,
+      "loss": 1.6252,
+      "step": 40192
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3053085714285714e-06,
+      "loss": 1.1639,
+      "step": 40193
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.30504e-06,
+      "loss": 1.0673,
+      "step": 40194
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3047714285714286e-06,
+      "loss": 0.8183,
+      "step": 40195
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3045028571428576e-06,
+      "loss": 1.3108,
+      "step": 40196
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3042342857142858e-06,
+      "loss": 1.0561,
+      "step": 40197
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3039657142857144e-06,
+      "loss": 1.6129,
+      "step": 40198
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.303697142857143e-06,
+      "loss": 1.1408,
+      "step": 40199
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3034285714285716e-06,
+      "loss": 0.9189,
+      "step": 40200
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.30316e-06,
+      "loss": 0.7736,
+      "step": 40201
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3028914285714288e-06,
+      "loss": 1.1127,
+      "step": 40202
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3026228571428574e-06,
+      "loss": 0.9174,
+      "step": 40203
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.302354285714286e-06,
+      "loss": 1.2287,
+      "step": 40204
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3020857142857146e-06,
+      "loss": 1.2496,
+      "step": 40205
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3018171428571427e-06,
+      "loss": 1.3733,
+      "step": 40206
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3015485714285713e-06,
+      "loss": 1.2502,
+      "step": 40207
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.30128e-06,
+      "loss": 1.0094,
+      "step": 40208
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.301011428571429e-06,
+      "loss": 1.236,
+      "step": 40209
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3007428571428575e-06,
+      "loss": 1.435,
+      "step": 40210
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3004742857142857e-06,
+      "loss": 1.1062,
+      "step": 40211
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.3002057142857143e-06,
+      "loss": 0.9197,
+      "step": 40212
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.299937142857143e-06,
+      "loss": 1.0597,
+      "step": 40213
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2996685714285715e-06,
+      "loss": 1.379,
+      "step": 40214
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2994e-06,
+      "loss": 1.0891,
+      "step": 40215
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2991314285714287e-06,
+      "loss": 0.9225,
+      "step": 40216
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2988628571428573e-06,
+      "loss": 1.5674,
+      "step": 40217
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.298594285714286e-06,
+      "loss": 1.3853,
+      "step": 40218
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2983257142857145e-06,
+      "loss": 0.927,
+      "step": 40219
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2980571428571427e-06,
+      "loss": 1.3595,
+      "step": 40220
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2977885714285713e-06,
+      "loss": 0.8398,
+      "step": 40221
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2975200000000003e-06,
+      "loss": 1.2801,
+      "step": 40222
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.297251428571429e-06,
+      "loss": 1.8423,
+      "step": 40223
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2969828571428575e-06,
+      "loss": 1.1116,
+      "step": 40224
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2967142857142857e-06,
+      "loss": 0.6743,
+      "step": 40225
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2964457142857143e-06,
+      "loss": 1.1615,
+      "step": 40226
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.296177142857143e-06,
+      "loss": 1.1398,
+      "step": 40227
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2959085714285715e-06,
+      "loss": 0.8226,
+      "step": 40228
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.29564e-06,
+      "loss": 1.2292,
+      "step": 40229
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2953714285714287e-06,
+      "loss": 1.4736,
+      "step": 40230
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2951028571428573e-06,
+      "loss": 1.0588,
+      "step": 40231
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.294834285714286e-06,
+      "loss": 0.8768,
+      "step": 40232
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2945657142857145e-06,
+      "loss": 1.5133,
+      "step": 40233
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2942971428571427e-06,
+      "loss": 1.2243,
+      "step": 40234
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2940285714285717e-06,
+      "loss": 0.9921,
+      "step": 40235
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2937600000000003e-06,
+      "loss": 1.3237,
+      "step": 40236
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.293491428571429e-06,
+      "loss": 0.7767,
+      "step": 40237
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.293222857142857e-06,
+      "loss": 1.1296,
+      "step": 40238
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2929542857142857e-06,
+      "loss": 1.5808,
+      "step": 40239
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2926857142857143e-06,
+      "loss": 1.4627,
+      "step": 40240
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.292417142857143e-06,
+      "loss": 1.3704,
+      "step": 40241
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.292148571428572e-06,
+      "loss": 1.0115,
+      "step": 40242
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.29188e-06,
+      "loss": 1.1676,
+      "step": 40243
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2916114285714287e-06,
+      "loss": 1.1779,
+      "step": 40244
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2913428571428573e-06,
+      "loss": 1.0294,
+      "step": 40245
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.291074285714286e-06,
+      "loss": 1.0386,
+      "step": 40246
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2908057142857145e-06,
+      "loss": 0.6467,
+      "step": 40247
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.290537142857143e-06,
+      "loss": 1.0306,
+      "step": 40248
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2902685714285717e-06,
+      "loss": 0.8268,
+      "step": 40249
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2900000000000003e-06,
+      "loss": 1.1376,
+      "step": 40250
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.289731428571429e-06,
+      "loss": 0.9122,
+      "step": 40251
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.289462857142857e-06,
+      "loss": 0.8267,
+      "step": 40252
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2891942857142856e-06,
+      "loss": 1.014,
+      "step": 40253
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2889257142857142e-06,
+      "loss": 1.2363,
+      "step": 40254
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2886571428571433e-06,
+      "loss": 1.6005,
+      "step": 40255
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.288388571428572e-06,
+      "loss": 1.2777,
+      "step": 40256
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.28812e-06,
+      "loss": 1.1056,
+      "step": 40257
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2878514285714286e-06,
+      "loss": 1.0886,
+      "step": 40258
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2875828571428572e-06,
+      "loss": 1.2917,
+      "step": 40259
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.287314285714286e-06,
+      "loss": 1.1786,
+      "step": 40260
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2870457142857144e-06,
+      "loss": 1.1291,
+      "step": 40261
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.286777142857143e-06,
+      "loss": 1.4627,
+      "step": 40262
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2865085714285716e-06,
+      "loss": 1.3131,
+      "step": 40263
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2862400000000002e-06,
+      "loss": 0.9974,
+      "step": 40264
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.285971428571429e-06,
+      "loss": 1.2091,
+      "step": 40265
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.285702857142857e-06,
+      "loss": 1.387,
+      "step": 40266
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2854342857142856e-06,
+      "loss": 1.073,
+      "step": 40267
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2851657142857146e-06,
+      "loss": 0.7786,
+      "step": 40268
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2848971428571432e-06,
+      "loss": 1.1576,
+      "step": 40269
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2846285714285714e-06,
+      "loss": 1.3269,
+      "step": 40270
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.28436e-06,
+      "loss": 1.229,
+      "step": 40271
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2840914285714286e-06,
+      "loss": 1.1462,
+      "step": 40272
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.283822857142857e-06,
+      "loss": 0.9245,
+      "step": 40273
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.283554285714286e-06,
+      "loss": 1.0826,
+      "step": 40274
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2832857142857144e-06,
+      "loss": 1.217,
+      "step": 40275
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.283017142857143e-06,
+      "loss": 1.0556,
+      "step": 40276
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2827485714285716e-06,
+      "loss": 1.097,
+      "step": 40277
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.28248e-06,
+      "loss": 1.2957,
+      "step": 40278
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.282211428571429e-06,
+      "loss": 0.8948,
+      "step": 40279
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.281942857142857e-06,
+      "loss": 1.5934,
+      "step": 40280
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.281674285714286e-06,
+      "loss": 1.0063,
+      "step": 40281
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2814057142857146e-06,
+      "loss": 1.0243,
+      "step": 40282
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.281137142857143e-06,
+      "loss": 1.2953,
+      "step": 40283
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2808685714285714e-06,
+      "loss": 1.3787,
+      "step": 40284
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2806e-06,
+      "loss": 1.207,
+      "step": 40285
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2803314285714286e-06,
+      "loss": 1.0695,
+      "step": 40286
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.280062857142857e-06,
+      "loss": 1.1133,
+      "step": 40287
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.279794285714286e-06,
+      "loss": 1.0316,
+      "step": 40288
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2795257142857144e-06,
+      "loss": 1.1198,
+      "step": 40289
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.279257142857143e-06,
+      "loss": 0.9882,
+      "step": 40290
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2789885714285716e-06,
+      "loss": 1.443,
+      "step": 40291
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.27872e-06,
+      "loss": 1.3198,
+      "step": 40292
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2784514285714284e-06,
+      "loss": 0.9936,
+      "step": 40293
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.278182857142857e-06,
+      "loss": 1.3272,
+      "step": 40294
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.277914285714286e-06,
+      "loss": 1.0704,
+      "step": 40295
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2776457142857146e-06,
+      "loss": 1.1824,
+      "step": 40296
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.277377142857143e-06,
+      "loss": 0.6327,
+      "step": 40297
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2771085714285714e-06,
+      "loss": 1.2676,
+      "step": 40298
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.27684e-06,
+      "loss": 1.2175,
+      "step": 40299
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2765714285714286e-06,
+      "loss": 0.9698,
+      "step": 40300
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2763028571428576e-06,
+      "loss": 1.0326,
+      "step": 40301
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.276034285714286e-06,
+      "loss": 1.3326,
+      "step": 40302
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2757657142857144e-06,
+      "loss": 0.9802,
+      "step": 40303
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.275497142857143e-06,
+      "loss": 0.9874,
+      "step": 40304
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2752285714285716e-06,
+      "loss": 1.3791,
+      "step": 40305
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.27496e-06,
+      "loss": 1.0879,
+      "step": 40306
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2746914285714283e-06,
+      "loss": 1.4481,
+      "step": 40307
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2744228571428573e-06,
+      "loss": 1.4873,
+      "step": 40308
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.274154285714286e-06,
+      "loss": 0.9117,
+      "step": 40309
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2738857142857145e-06,
+      "loss": 1.2856,
+      "step": 40310
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.273617142857143e-06,
+      "loss": 0.8521,
+      "step": 40311
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2733485714285713e-06,
+      "loss": 1.279,
+      "step": 40312
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.27308e-06,
+      "loss": 1.2321,
+      "step": 40313
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.272811428571429e-06,
+      "loss": 1.3582,
+      "step": 40314
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2725428571428575e-06,
+      "loss": 1.2308,
+      "step": 40315
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2722742857142857e-06,
+      "loss": 1.1855,
+      "step": 40316
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2720057142857143e-06,
+      "loss": 0.8857,
+      "step": 40317
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.271737142857143e-06,
+      "loss": 1.0906,
+      "step": 40318
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2714685714285715e-06,
+      "loss": 1.1189,
+      "step": 40319
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2712e-06,
+      "loss": 1.0624,
+      "step": 40320
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2709314285714287e-06,
+      "loss": 1.1262,
+      "step": 40321
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2706628571428573e-06,
+      "loss": 0.904,
+      "step": 40322
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.270394285714286e-06,
+      "loss": 0.9184,
+      "step": 40323
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2701257142857145e-06,
+      "loss": 0.9437,
+      "step": 40324
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2698571428571427e-06,
+      "loss": 0.7357,
+      "step": 40325
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2695885714285713e-06,
+      "loss": 1.1478,
+      "step": 40326
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.26932e-06,
+      "loss": 0.8898,
+      "step": 40327
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.269051428571429e-06,
+      "loss": 0.9222,
+      "step": 40328
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2687828571428575e-06,
+      "loss": 0.8704,
+      "step": 40329
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2685142857142857e-06,
+      "loss": 0.8377,
+      "step": 40330
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2682457142857143e-06,
+      "loss": 1.5443,
+      "step": 40331
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.267977142857143e-06,
+      "loss": 0.8491,
+      "step": 40332
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2677085714285715e-06,
+      "loss": 1.4364,
+      "step": 40333
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2674400000000005e-06,
+      "loss": 0.9139,
+      "step": 40334
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2671714285714287e-06,
+      "loss": 1.0187,
+      "step": 40335
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.2669028571428573e-06,
+      "loss": 0.9857,
+      "step": 40336
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.266634285714286e-06,
+      "loss": 1.1979,
+      "step": 40337
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2663657142857145e-06,
+      "loss": 1.0074,
+      "step": 40338
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2660971428571427e-06,
+      "loss": 1.5,
+      "step": 40339
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2658285714285713e-06,
+      "loss": 1.2413,
+      "step": 40340
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2655600000000003e-06,
+      "loss": 0.9703,
+      "step": 40341
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.265291428571429e-06,
+      "loss": 1.072,
+      "step": 40342
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2650228571428575e-06,
+      "loss": 1.0624,
+      "step": 40343
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2647542857142857e-06,
+      "loss": 1.2138,
+      "step": 40344
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2644857142857143e-06,
+      "loss": 1.2042,
+      "step": 40345
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.264217142857143e-06,
+      "loss": 1.2816,
+      "step": 40346
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2639485714285715e-06,
+      "loss": 1.1607,
+      "step": 40347
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.26368e-06,
+      "loss": 1.071,
+      "step": 40348
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2634114285714287e-06,
+      "loss": 1.034,
+      "step": 40349
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2631428571428573e-06,
+      "loss": 0.9419,
+      "step": 40350
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.262874285714286e-06,
+      "loss": 0.9868,
+      "step": 40351
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2626057142857145e-06,
+      "loss": 1.2862,
+      "step": 40352
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2623371428571426e-06,
+      "loss": 0.8677,
+      "step": 40353
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2620685714285717e-06,
+      "loss": 1.4585,
+      "step": 40354
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2618000000000003e-06,
+      "loss": 1.0205,
+      "step": 40355
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.261531428571429e-06,
+      "loss": 1.3849,
+      "step": 40356
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.261262857142857e-06,
+      "loss": 1.3576,
+      "step": 40357
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2609942857142856e-06,
+      "loss": 1.0167,
+      "step": 40358
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2607257142857142e-06,
+      "loss": 1.0846,
+      "step": 40359
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.260457142857143e-06,
+      "loss": 1.0452,
+      "step": 40360
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.260188571428572e-06,
+      "loss": 1.2752,
+      "step": 40361
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.25992e-06,
+      "loss": 1.171,
+      "step": 40362
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2596514285714286e-06,
+      "loss": 1.2443,
+      "step": 40363
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2593828571428572e-06,
+      "loss": 1.0614,
+      "step": 40364
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.259114285714286e-06,
+      "loss": 0.9831,
+      "step": 40365
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2588457142857144e-06,
+      "loss": 1.267,
+      "step": 40366
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.258577142857143e-06,
+      "loss": 1.2009,
+      "step": 40367
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2583085714285716e-06,
+      "loss": 1.2082,
+      "step": 40368
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2580400000000002e-06,
+      "loss": 1.0521,
+      "step": 40369
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.257771428571429e-06,
+      "loss": 0.9261,
+      "step": 40370
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.257502857142857e-06,
+      "loss": 0.8882,
+      "step": 40371
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2572342857142856e-06,
+      "loss": 0.8344,
+      "step": 40372
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.256965714285714e-06,
+      "loss": 1.4266,
+      "step": 40373
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2566971428571432e-06,
+      "loss": 1.2297,
+      "step": 40374
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.256428571428572e-06,
+      "loss": 1.4129,
+      "step": 40375
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.25616e-06,
+      "loss": 1.0796,
+      "step": 40376
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2558914285714286e-06,
+      "loss": 1.0763,
+      "step": 40377
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.255622857142857e-06,
+      "loss": 1.0384,
+      "step": 40378
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.255354285714286e-06,
+      "loss": 1.2651,
+      "step": 40379
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2550857142857144e-06,
+      "loss": 1.4405,
+      "step": 40380
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.254817142857143e-06,
+      "loss": 0.9944,
+      "step": 40381
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2545485714285716e-06,
+      "loss": 0.9558,
+      "step": 40382
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.25428e-06,
+      "loss": 1.2258,
+      "step": 40383
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.254011428571429e-06,
+      "loss": 1.3522,
+      "step": 40384
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.253742857142857e-06,
+      "loss": 0.8748,
+      "step": 40385
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2534742857142856e-06,
+      "loss": 1.5723,
+      "step": 40386
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2532057142857146e-06,
+      "loss": 1.305,
+      "step": 40387
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.252937142857143e-06,
+      "loss": 1.0966,
+      "step": 40388
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.252668571428572e-06,
+      "loss": 0.8557,
+      "step": 40389
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2524e-06,
+      "loss": 0.9553,
+      "step": 40390
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2521314285714286e-06,
+      "loss": 1.0561,
+      "step": 40391
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.251862857142857e-06,
+      "loss": 0.8154,
+      "step": 40392
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2515942857142858e-06,
+      "loss": 1.5523,
+      "step": 40393
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2513257142857144e-06,
+      "loss": 1.3105,
+      "step": 40394
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.251057142857143e-06,
+      "loss": 0.96,
+      "step": 40395
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2507885714285716e-06,
+      "loss": 1.0379,
+      "step": 40396
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.25052e-06,
+      "loss": 1.3504,
+      "step": 40397
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2502514285714288e-06,
+      "loss": 1.052,
+      "step": 40398
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.249982857142857e-06,
+      "loss": 1.0494,
+      "step": 40399
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.249714285714286e-06,
+      "loss": 1.2149,
+      "step": 40400
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2494457142857146e-06,
+      "loss": 1.2959,
+      "step": 40401
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.249177142857143e-06,
+      "loss": 1.2872,
+      "step": 40402
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2489085714285714e-06,
+      "loss": 1.2363,
+      "step": 40403
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.24864e-06,
+      "loss": 0.9835,
+      "step": 40404
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2483714285714286e-06,
+      "loss": 0.9532,
+      "step": 40405
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.248102857142857e-06,
+      "loss": 1.0785,
+      "step": 40406
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.247834285714286e-06,
+      "loss": 1.2954,
+      "step": 40407
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2475657142857144e-06,
+      "loss": 1.3124,
+      "step": 40408
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.247297142857143e-06,
+      "loss": 0.9408,
+      "step": 40409
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2470285714285715e-06,
+      "loss": 1.0612,
+      "step": 40410
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.24676e-06,
+      "loss": 1.2815,
+      "step": 40411
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2464914285714283e-06,
+      "loss": 1.1693,
+      "step": 40412
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2462228571428573e-06,
+      "loss": 1.0437,
+      "step": 40413
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.245954285714286e-06,
+      "loss": 1.0022,
+      "step": 40414
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2456857142857145e-06,
+      "loss": 1.0093,
+      "step": 40415
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.245417142857143e-06,
+      "loss": 1.1249,
+      "step": 40416
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2451485714285713e-06,
+      "loss": 0.5759,
+      "step": 40417
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.24488e-06,
+      "loss": 0.8115,
+      "step": 40418
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2446114285714285e-06,
+      "loss": 1.5256,
+      "step": 40419
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2443428571428575e-06,
+      "loss": 1.2422,
+      "step": 40420
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.244074285714286e-06,
+      "loss": 1.4426,
+      "step": 40421
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2438057142857143e-06,
+      "loss": 1.0015,
+      "step": 40422
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.243537142857143e-06,
+      "loss": 1.2157,
+      "step": 40423
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2432685714285715e-06,
+      "loss": 1.3494,
+      "step": 40424
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.243e-06,
+      "loss": 1.1751,
+      "step": 40425
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2427314285714287e-06,
+      "loss": 1.3878,
+      "step": 40426
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2424628571428573e-06,
+      "loss": 1.4022,
+      "step": 40427
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.242194285714286e-06,
+      "loss": 1.0746,
+      "step": 40428
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2419257142857145e-06,
+      "loss": 1.0828,
+      "step": 40429
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.241657142857143e-06,
+      "loss": 1.3073,
+      "step": 40430
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2413885714285713e-06,
+      "loss": 1.138,
+      "step": 40431
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.24112e-06,
+      "loss": 0.894,
+      "step": 40432
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.240851428571429e-06,
+      "loss": 0.8356,
+      "step": 40433
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2405828571428575e-06,
+      "loss": 1.1033,
+      "step": 40434
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2403142857142857e-06,
+      "loss": 1.6931,
+      "step": 40435
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2400457142857143e-06,
+      "loss": 0.9375,
+      "step": 40436
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.239777142857143e-06,
+      "loss": 1.3342,
+      "step": 40437
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2395085714285715e-06,
+      "loss": 1.4238,
+      "step": 40438
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.23924e-06,
+      "loss": 0.982,
+      "step": 40439
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2389714285714287e-06,
+      "loss": 0.6948,
+      "step": 40440
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2387028571428573e-06,
+      "loss": 1.1236,
+      "step": 40441
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.238434285714286e-06,
+      "loss": 1.1797,
+      "step": 40442
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2381657142857145e-06,
+      "loss": 1.2778,
+      "step": 40443
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2378971428571427e-06,
+      "loss": 1.2659,
+      "step": 40444
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2376285714285713e-06,
+      "loss": 0.9363,
+      "step": 40445
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2373600000000003e-06,
+      "loss": 0.7406,
+      "step": 40446
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.237091428571429e-06,
+      "loss": 1.2873,
+      "step": 40447
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2368228571428575e-06,
+      "loss": 1.3313,
+      "step": 40448
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2365542857142857e-06,
+      "loss": 1.3514,
+      "step": 40449
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2362857142857143e-06,
+      "loss": 0.933,
+      "step": 40450
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.236017142857143e-06,
+      "loss": 1.0957,
+      "step": 40451
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2357485714285715e-06,
+      "loss": 1.1691,
+      "step": 40452
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2354800000000005e-06,
+      "loss": 0.8455,
+      "step": 40453
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2352114285714287e-06,
+      "loss": 1.1674,
+      "step": 40454
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2349428571428573e-06,
+      "loss": 1.2562,
+      "step": 40455
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.234674285714286e-06,
+      "loss": 1.0811,
+      "step": 40456
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2344057142857145e-06,
+      "loss": 1.1678,
+      "step": 40457
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2341371428571426e-06,
+      "loss": 1.1022,
+      "step": 40458
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2338685714285717e-06,
+      "loss": 1.0845,
+      "step": 40459
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2336000000000003e-06,
+      "loss": 1.2143,
+      "step": 40460
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.233331428571429e-06,
+      "loss": 1.1587,
+      "step": 40461
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2330628571428575e-06,
+      "loss": 1.6316,
+      "step": 40462
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2327942857142856e-06,
+      "loss": 1.3627,
+      "step": 40463
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2325257142857142e-06,
+      "loss": 1.3956,
+      "step": 40464
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.232257142857143e-06,
+      "loss": 0.8185,
+      "step": 40465
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.231988571428572e-06,
+      "loss": 1.2439,
+      "step": 40466
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.23172e-06,
+      "loss": 1.4327,
+      "step": 40467
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2314514285714286e-06,
+      "loss": 1.19,
+      "step": 40468
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2311828571428572e-06,
+      "loss": 1.2822,
+      "step": 40469
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.230914285714286e-06,
+      "loss": 0.7821,
+      "step": 40470
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2306457142857144e-06,
+      "loss": 1.0948,
+      "step": 40471
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.230377142857143e-06,
+      "loss": 1.395,
+      "step": 40472
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2301085714285716e-06,
+      "loss": 0.9713,
+      "step": 40473
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2298400000000002e-06,
+      "loss": 1.2591,
+      "step": 40474
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.229571428571429e-06,
+      "loss": 1.1885,
+      "step": 40475
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.229302857142857e-06,
+      "loss": 1.3224,
+      "step": 40476
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2290342857142856e-06,
+      "loss": 0.9535,
+      "step": 40477
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.228765714285714e-06,
+      "loss": 0.9289,
+      "step": 40478
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2284971428571432e-06,
+      "loss": 1.1107,
+      "step": 40479
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.228228571428572e-06,
+      "loss": 0.8297,
+      "step": 40480
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.22796e-06,
+      "loss": 1.111,
+      "step": 40481
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2276914285714286e-06,
+      "loss": 1.3789,
+      "step": 40482
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.227422857142857e-06,
+      "loss": 1.3721,
+      "step": 40483
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.227154285714286e-06,
+      "loss": 1.1837,
+      "step": 40484
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2268857142857144e-06,
+      "loss": 1.0378,
+      "step": 40485
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.226617142857143e-06,
+      "loss": 1.4834,
+      "step": 40486
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2263485714285716e-06,
+      "loss": 1.144,
+      "step": 40487
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.22608e-06,
+      "loss": 1.3535,
+      "step": 40488
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.225811428571429e-06,
+      "loss": 1.0848,
+      "step": 40489
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.225542857142857e-06,
+      "loss": 0.914,
+      "step": 40490
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2252742857142856e-06,
+      "loss": 1.2925,
+      "step": 40491
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2250057142857146e-06,
+      "loss": 1.0586,
+      "step": 40492
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.224737142857143e-06,
+      "loss": 0.8875,
+      "step": 40493
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.224468571428572e-06,
+      "loss": 1.2766,
+      "step": 40494
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2242e-06,
+      "loss": 0.9412,
+      "step": 40495
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2239314285714286e-06,
+      "loss": 1.0138,
+      "step": 40496
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.223662857142857e-06,
+      "loss": 1.129,
+      "step": 40497
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2233942857142858e-06,
+      "loss": 0.9521,
+      "step": 40498
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2231257142857144e-06,
+      "loss": 1.2889,
+      "step": 40499
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.222857142857143e-06,
+      "loss": 1.31,
+      "step": 40500
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2225885714285716e-06,
+      "loss": 0.951,
+      "step": 40501
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.22232e-06,
+      "loss": 0.8724,
+      "step": 40502
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2220514285714288e-06,
+      "loss": 1.1867,
+      "step": 40503
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.221782857142857e-06,
+      "loss": 0.9532,
+      "step": 40504
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.221514285714286e-06,
+      "loss": 1.2629,
+      "step": 40505
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2212457142857146e-06,
+      "loss": 1.3826,
+      "step": 40506
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.220977142857143e-06,
+      "loss": 0.8653,
+      "step": 40507
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2207085714285718e-06,
+      "loss": 1.053,
+      "step": 40508
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.22044e-06,
+      "loss": 0.9152,
+      "step": 40509
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.2201714285714286e-06,
+      "loss": 1.236,
+      "step": 40510
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.219902857142857e-06,
+      "loss": 1.0995,
+      "step": 40511
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.219634285714286e-06,
+      "loss": 0.9569,
+      "step": 40512
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2193657142857143e-06,
+      "loss": 1.1472,
+      "step": 40513
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.219097142857143e-06,
+      "loss": 1.2236,
+      "step": 40514
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2188285714285715e-06,
+      "loss": 1.314,
+      "step": 40515
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.21856e-06,
+      "loss": 1.3595,
+      "step": 40516
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2182914285714287e-06,
+      "loss": 1.4326,
+      "step": 40517
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.218022857142857e-06,
+      "loss": 1.1626,
+      "step": 40518
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.217754285714286e-06,
+      "loss": 1.1147,
+      "step": 40519
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2174857142857145e-06,
+      "loss": 1.2386,
+      "step": 40520
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.217217142857143e-06,
+      "loss": 1.0326,
+      "step": 40521
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2169485714285713e-06,
+      "loss": 1.0221,
+      "step": 40522
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.21668e-06,
+      "loss": 1.1721,
+      "step": 40523
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2164114285714285e-06,
+      "loss": 1.1273,
+      "step": 40524
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2161428571428575e-06,
+      "loss": 1.5061,
+      "step": 40525
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.215874285714286e-06,
+      "loss": 1.2094,
+      "step": 40526
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2156057142857143e-06,
+      "loss": 0.8953,
+      "step": 40527
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.215337142857143e-06,
+      "loss": 1.0191,
+      "step": 40528
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2150685714285715e-06,
+      "loss": 1.0428,
+      "step": 40529
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2148e-06,
+      "loss": 1.2237,
+      "step": 40530
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2145314285714283e-06,
+      "loss": 1.3436,
+      "step": 40531
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2142628571428573e-06,
+      "loss": 0.751,
+      "step": 40532
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.213994285714286e-06,
+      "loss": 1.021,
+      "step": 40533
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2137257142857145e-06,
+      "loss": 1.4945,
+      "step": 40534
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.213457142857143e-06,
+      "loss": 1.0504,
+      "step": 40535
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2131885714285713e-06,
+      "loss": 0.9233,
+      "step": 40536
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.21292e-06,
+      "loss": 0.7965,
+      "step": 40537
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.212651428571429e-06,
+      "loss": 0.8878,
+      "step": 40538
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2123828571428575e-06,
+      "loss": 1.2435,
+      "step": 40539
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.212114285714286e-06,
+      "loss": 1.0398,
+      "step": 40540
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2118457142857143e-06,
+      "loss": 1.2086,
+      "step": 40541
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.211577142857143e-06,
+      "loss": 1.1721,
+      "step": 40542
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2113085714285715e-06,
+      "loss": 1.3554,
+      "step": 40543
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.21104e-06,
+      "loss": 1.266,
+      "step": 40544
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2107714285714287e-06,
+      "loss": 0.8167,
+      "step": 40545
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2105028571428573e-06,
+      "loss": 1.0396,
+      "step": 40546
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.210234285714286e-06,
+      "loss": 1.2828,
+      "step": 40547
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2099657142857145e-06,
+      "loss": 1.0335,
+      "step": 40548
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.209697142857143e-06,
+      "loss": 1.2204,
+      "step": 40549
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2094285714285713e-06,
+      "loss": 1.0155,
+      "step": 40550
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.20916e-06,
+      "loss": 1.7522,
+      "step": 40551
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.208891428571429e-06,
+      "loss": 1.0613,
+      "step": 40552
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2086228571428575e-06,
+      "loss": 1.1223,
+      "step": 40553
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2083542857142857e-06,
+      "loss": 0.7035,
+      "step": 40554
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2080857142857143e-06,
+      "loss": 1.0015,
+      "step": 40555
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.207817142857143e-06,
+      "loss": 1.3358,
+      "step": 40556
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2075485714285715e-06,
+      "loss": 0.8635,
+      "step": 40557
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2072800000000005e-06,
+      "loss": 0.8779,
+      "step": 40558
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2070114285714287e-06,
+      "loss": 0.9167,
+      "step": 40559
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2067428571428573e-06,
+      "loss": 1.0913,
+      "step": 40560
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.206474285714286e-06,
+      "loss": 1.2233,
+      "step": 40561
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2062057142857145e-06,
+      "loss": 1.0159,
+      "step": 40562
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2059371428571426e-06,
+      "loss": 0.6622,
+      "step": 40563
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2056685714285712e-06,
+      "loss": 1.0571,
+      "step": 40564
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2054000000000003e-06,
+      "loss": 1.141,
+      "step": 40565
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.205131428571429e-06,
+      "loss": 0.9234,
+      "step": 40566
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2048628571428575e-06,
+      "loss": 0.9733,
+      "step": 40567
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2045942857142856e-06,
+      "loss": 1.313,
+      "step": 40568
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2043257142857142e-06,
+      "loss": 1.202,
+      "step": 40569
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.204057142857143e-06,
+      "loss": 1.263,
+      "step": 40570
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2037885714285714e-06,
+      "loss": 1.3546,
+      "step": 40571
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2035200000000005e-06,
+      "loss": 0.9814,
+      "step": 40572
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2032514285714286e-06,
+      "loss": 1.0282,
+      "step": 40573
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2029828571428572e-06,
+      "loss": 1.4629,
+      "step": 40574
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.202714285714286e-06,
+      "loss": 0.9502,
+      "step": 40575
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2024457142857144e-06,
+      "loss": 1.3298,
+      "step": 40576
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2021771428571426e-06,
+      "loss": 1.2325,
+      "step": 40577
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2019085714285716e-06,
+      "loss": 1.0328,
+      "step": 40578
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2016400000000002e-06,
+      "loss": 0.8612,
+      "step": 40579
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.201371428571429e-06,
+      "loss": 1.2605,
+      "step": 40580
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2011028571428574e-06,
+      "loss": 1.0647,
+      "step": 40581
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.2008342857142856e-06,
+      "loss": 1.7917,
+      "step": 40582
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.200565714285714e-06,
+      "loss": 1.2065,
+      "step": 40583
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.200297142857143e-06,
+      "loss": 1.1131,
+      "step": 40584
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.200028571428572e-06,
+      "loss": 0.9804,
+      "step": 40585
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.19976e-06,
+      "loss": 0.7392,
+      "step": 40586
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1994914285714286e-06,
+      "loss": 1.3302,
+      "step": 40587
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.199222857142857e-06,
+      "loss": 1.0004,
+      "step": 40588
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.198954285714286e-06,
+      "loss": 0.9305,
+      "step": 40589
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1986857142857144e-06,
+      "loss": 0.9409,
+      "step": 40590
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.198417142857143e-06,
+      "loss": 1.3475,
+      "step": 40591
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1981485714285716e-06,
+      "loss": 0.5774,
+      "step": 40592
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.19788e-06,
+      "loss": 1.0812,
+      "step": 40593
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.197611428571429e-06,
+      "loss": 1.5746,
+      "step": 40594
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1973428571428574e-06,
+      "loss": 0.7172,
+      "step": 40595
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1970742857142856e-06,
+      "loss": 1.3616,
+      "step": 40596
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.196805714285714e-06,
+      "loss": 1.3143,
+      "step": 40597
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.196537142857143e-06,
+      "loss": 1.3554,
+      "step": 40598
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.196268571428572e-06,
+      "loss": 0.7166,
+      "step": 40599
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.196e-06,
+      "loss": 1.2782,
+      "step": 40600
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1957314285714286e-06,
+      "loss": 1.212,
+      "step": 40601
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.195462857142857e-06,
+      "loss": 1.1687,
+      "step": 40602
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1951942857142858e-06,
+      "loss": 0.6934,
+      "step": 40603
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1949257142857144e-06,
+      "loss": 1.0667,
+      "step": 40604
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.194657142857143e-06,
+      "loss": 1.2625,
+      "step": 40605
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1943885714285716e-06,
+      "loss": 1.3715,
+      "step": 40606
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.19412e-06,
+      "loss": 1.1896,
+      "step": 40607
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1938514285714288e-06,
+      "loss": 1.515,
+      "step": 40608
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.193582857142857e-06,
+      "loss": 1.1982,
+      "step": 40609
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1933142857142856e-06,
+      "loss": 0.9962,
+      "step": 40610
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1930457142857146e-06,
+      "loss": 0.7417,
+      "step": 40611
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.192777142857143e-06,
+      "loss": 0.7611,
+      "step": 40612
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1925085714285718e-06,
+      "loss": 0.7594,
+      "step": 40613
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.19224e-06,
+      "loss": 1.5034,
+      "step": 40614
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1919714285714285e-06,
+      "loss": 1.0751,
+      "step": 40615
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.191702857142857e-06,
+      "loss": 0.9666,
+      "step": 40616
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1914342857142857e-06,
+      "loss": 1.3003,
+      "step": 40617
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1911657142857143e-06,
+      "loss": 1.0457,
+      "step": 40618
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.190897142857143e-06,
+      "loss": 1.2657,
+      "step": 40619
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1906285714285715e-06,
+      "loss": 1.1387,
+      "step": 40620
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.19036e-06,
+      "loss": 1.0039,
+      "step": 40621
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1900914285714287e-06,
+      "loss": 1.4294,
+      "step": 40622
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.189822857142857e-06,
+      "loss": 1.3867,
+      "step": 40623
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.189554285714286e-06,
+      "loss": 1.6271,
+      "step": 40624
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1892857142857145e-06,
+      "loss": 1.3841,
+      "step": 40625
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.189017142857143e-06,
+      "loss": 1.3701,
+      "step": 40626
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1887485714285717e-06,
+      "loss": 1.2276,
+      "step": 40627
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.18848e-06,
+      "loss": 1.4172,
+      "step": 40628
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1882114285714285e-06,
+      "loss": 1.5813,
+      "step": 40629
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.187942857142857e-06,
+      "loss": 1.0658,
+      "step": 40630
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.187674285714286e-06,
+      "loss": 0.9338,
+      "step": 40631
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1874057142857143e-06,
+      "loss": 1.4521,
+      "step": 40632
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.187137142857143e-06,
+      "loss": 1.2508,
+      "step": 40633
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1868685714285715e-06,
+      "loss": 0.9732,
+      "step": 40634
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1866e-06,
+      "loss": 0.5649,
+      "step": 40635
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1863314285714287e-06,
+      "loss": 1.0288,
+      "step": 40636
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1860628571428573e-06,
+      "loss": 1.4652,
+      "step": 40637
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.185794285714286e-06,
+      "loss": 1.1538,
+      "step": 40638
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1855257142857145e-06,
+      "loss": 0.9989,
+      "step": 40639
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.185257142857143e-06,
+      "loss": 1.1611,
+      "step": 40640
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1849885714285713e-06,
+      "loss": 1.418,
+      "step": 40641
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.18472e-06,
+      "loss": 1.0075,
+      "step": 40642
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1844514285714285e-06,
+      "loss": 1.4459,
+      "step": 40643
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1841828571428575e-06,
+      "loss": 1.1316,
+      "step": 40644
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.183914285714286e-06,
+      "loss": 1.0439,
+      "step": 40645
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1836457142857143e-06,
+      "loss": 0.7279,
+      "step": 40646
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.183377142857143e-06,
+      "loss": 1.3499,
+      "step": 40647
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1831085714285715e-06,
+      "loss": 1.1621,
+      "step": 40648
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.18284e-06,
+      "loss": 1.4252,
+      "step": 40649
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1825714285714287e-06,
+      "loss": 0.9607,
+      "step": 40650
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1823028571428573e-06,
+      "loss": 0.9683,
+      "step": 40651
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.182034285714286e-06,
+      "loss": 0.667,
+      "step": 40652
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1817657142857145e-06,
+      "loss": 1.3103,
+      "step": 40653
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.181497142857143e-06,
+      "loss": 1.4027,
+      "step": 40654
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1812285714285713e-06,
+      "loss": 1.0489,
+      "step": 40655
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.18096e-06,
+      "loss": 0.7449,
+      "step": 40656
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.180691428571429e-06,
+      "loss": 1.4248,
+      "step": 40657
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1804228571428575e-06,
+      "loss": 0.8655,
+      "step": 40658
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.180154285714286e-06,
+      "loss": 1.0431,
+      "step": 40659
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1798857142857143e-06,
+      "loss": 1.2724,
+      "step": 40660
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.179617142857143e-06,
+      "loss": 1.2573,
+      "step": 40661
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1793485714285715e-06,
+      "loss": 0.9793,
+      "step": 40662
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.17908e-06,
+      "loss": 0.9158,
+      "step": 40663
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1788114285714287e-06,
+      "loss": 1.2833,
+      "step": 40664
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1785428571428573e-06,
+      "loss": 0.7933,
+      "step": 40665
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.178274285714286e-06,
+      "loss": 1.44,
+      "step": 40666
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1780057142857145e-06,
+      "loss": 0.9788,
+      "step": 40667
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.177737142857143e-06,
+      "loss": 1.1948,
+      "step": 40668
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1774685714285712e-06,
+      "loss": 1.0791,
+      "step": 40669
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1772000000000003e-06,
+      "loss": 1.3645,
+      "step": 40670
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.176931428571429e-06,
+      "loss": 1.357,
+      "step": 40671
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1766628571428575e-06,
+      "loss": 1.1298,
+      "step": 40672
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1763942857142856e-06,
+      "loss": 1.1653,
+      "step": 40673
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1761257142857142e-06,
+      "loss": 1.1769,
+      "step": 40674
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.175857142857143e-06,
+      "loss": 1.1857,
+      "step": 40675
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1755885714285714e-06,
+      "loss": 1.1394,
+      "step": 40676
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1753200000000005e-06,
+      "loss": 0.8101,
+      "step": 40677
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1750514285714286e-06,
+      "loss": 1.2362,
+      "step": 40678
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1747828571428572e-06,
+      "loss": 0.9206,
+      "step": 40679
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.174514285714286e-06,
+      "loss": 0.9638,
+      "step": 40680
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1742457142857144e-06,
+      "loss": 1.1632,
+      "step": 40681
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.173977142857143e-06,
+      "loss": 1.1913,
+      "step": 40682
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1737085714285716e-06,
+      "loss": 1.0868,
+      "step": 40683
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1734400000000002e-06,
+      "loss": 0.8114,
+      "step": 40684
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.173171428571429e-06,
+      "loss": 1.1618,
+      "step": 40685
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1729028571428574e-06,
+      "loss": 1.3326,
+      "step": 40686
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.1726342857142856e-06,
+      "loss": 1.3124,
+      "step": 40687
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.172365714285714e-06,
+      "loss": 1.0332,
+      "step": 40688
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.172097142857143e-06,
+      "loss": 1.1391,
+      "step": 40689
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.171828571428572e-06,
+      "loss": 1.4316,
+      "step": 40690
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1715600000000004e-06,
+      "loss": 1.2672,
+      "step": 40691
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1712914285714286e-06,
+      "loss": 1.6115,
+      "step": 40692
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.171022857142857e-06,
+      "loss": 1.0688,
+      "step": 40693
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.170754285714286e-06,
+      "loss": 0.7986,
+      "step": 40694
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1704857142857144e-06,
+      "loss": 1.648,
+      "step": 40695
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.170217142857143e-06,
+      "loss": 0.6747,
+      "step": 40696
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1699485714285716e-06,
+      "loss": 1.0637,
+      "step": 40697
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.16968e-06,
+      "loss": 0.8968,
+      "step": 40698
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.169411428571429e-06,
+      "loss": 1.3356,
+      "step": 40699
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1691428571428574e-06,
+      "loss": 1.1634,
+      "step": 40700
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1688742857142856e-06,
+      "loss": 1.3737,
+      "step": 40701
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.168605714285714e-06,
+      "loss": 1.1468,
+      "step": 40702
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.168337142857143e-06,
+      "loss": 1.6974,
+      "step": 40703
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.168068571428572e-06,
+      "loss": 1.0167,
+      "step": 40704
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1678e-06,
+      "loss": 1.2113,
+      "step": 40705
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1675314285714286e-06,
+      "loss": 1.4483,
+      "step": 40706
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.167262857142857e-06,
+      "loss": 0.9701,
+      "step": 40707
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1669942857142858e-06,
+      "loss": 0.7894,
+      "step": 40708
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1667257142857144e-06,
+      "loss": 0.8103,
+      "step": 40709
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.166457142857143e-06,
+      "loss": 0.8938,
+      "step": 40710
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1661885714285716e-06,
+      "loss": 1.0445,
+      "step": 40711
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.16592e-06,
+      "loss": 1.2097,
+      "step": 40712
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1656514285714288e-06,
+      "loss": 1.2641,
+      "step": 40713
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1653828571428574e-06,
+      "loss": 1.0795,
+      "step": 40714
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1651142857142856e-06,
+      "loss": 0.7696,
+      "step": 40715
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1648457142857146e-06,
+      "loss": 1.0381,
+      "step": 40716
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.164577142857143e-06,
+      "loss": 0.7738,
+      "step": 40717
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1643085714285718e-06,
+      "loss": 1.1496,
+      "step": 40718
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.16404e-06,
+      "loss": 1.084,
+      "step": 40719
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1637714285714285e-06,
+      "loss": 1.2626,
+      "step": 40720
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.163502857142857e-06,
+      "loss": 0.9296,
+      "step": 40721
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1632342857142857e-06,
+      "loss": 1.4624,
+      "step": 40722
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1629657142857148e-06,
+      "loss": 1.2479,
+      "step": 40723
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.162697142857143e-06,
+      "loss": 0.7867,
+      "step": 40724
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1624285714285715e-06,
+      "loss": 1.7956,
+      "step": 40725
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.16216e-06,
+      "loss": 1.2008,
+      "step": 40726
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1618914285714287e-06,
+      "loss": 1.1589,
+      "step": 40727
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.161622857142857e-06,
+      "loss": 0.9781,
+      "step": 40728
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.161354285714286e-06,
+      "loss": 1.0658,
+      "step": 40729
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1610857142857145e-06,
+      "loss": 1.4587,
+      "step": 40730
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.160817142857143e-06,
+      "loss": 1.3022,
+      "step": 40731
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1605485714285717e-06,
+      "loss": 1.0453,
+      "step": 40732
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.16028e-06,
+      "loss": 1.5195,
+      "step": 40733
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1600114285714285e-06,
+      "loss": 1.225,
+      "step": 40734
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.159742857142857e-06,
+      "loss": 1.0546,
+      "step": 40735
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.159474285714286e-06,
+      "loss": 1.1923,
+      "step": 40736
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1592057142857143e-06,
+      "loss": 1.3466,
+      "step": 40737
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.158937142857143e-06,
+      "loss": 1.4684,
+      "step": 40738
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1586685714285715e-06,
+      "loss": 1.0378,
+      "step": 40739
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1584e-06,
+      "loss": 1.6006,
+      "step": 40740
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1581314285714287e-06,
+      "loss": 0.9642,
+      "step": 40741
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1578628571428573e-06,
+      "loss": 1.2427,
+      "step": 40742
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.157594285714286e-06,
+      "loss": 1.0027,
+      "step": 40743
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1573257142857145e-06,
+      "loss": 1.0945,
+      "step": 40744
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.157057142857143e-06,
+      "loss": 0.809,
+      "step": 40745
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1567885714285717e-06,
+      "loss": 1.2451,
+      "step": 40746
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.15652e-06,
+      "loss": 1.3413,
+      "step": 40747
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1562514285714285e-06,
+      "loss": 0.9664,
+      "step": 40748
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1559828571428575e-06,
+      "loss": 1.2985,
+      "step": 40749
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.155714285714286e-06,
+      "loss": 1.0861,
+      "step": 40750
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1554457142857143e-06,
+      "loss": 1.0811,
+      "step": 40751
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.155177142857143e-06,
+      "loss": 1.207,
+      "step": 40752
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1549085714285715e-06,
+      "loss": 1.5141,
+      "step": 40753
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.15464e-06,
+      "loss": 0.9517,
+      "step": 40754
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1543714285714287e-06,
+      "loss": 1.1403,
+      "step": 40755
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1541028571428573e-06,
+      "loss": 1.1246,
+      "step": 40756
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.153834285714286e-06,
+      "loss": 1.2702,
+      "step": 40757
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1535657142857145e-06,
+      "loss": 0.9817,
+      "step": 40758
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.153297142857143e-06,
+      "loss": 1.4679,
+      "step": 40759
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1530285714285713e-06,
+      "loss": 1.4026,
+      "step": 40760
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.15276e-06,
+      "loss": 1.1057,
+      "step": 40761
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.152491428571429e-06,
+      "loss": 1.2191,
+      "step": 40762
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1522228571428575e-06,
+      "loss": 1.2089,
+      "step": 40763
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.151954285714286e-06,
+      "loss": 0.8168,
+      "step": 40764
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1516857142857143e-06,
+      "loss": 1.0971,
+      "step": 40765
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.151417142857143e-06,
+      "loss": 1.1918,
+      "step": 40766
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1511485714285715e-06,
+      "loss": 1.005,
+      "step": 40767
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.15088e-06,
+      "loss": 1.6241,
+      "step": 40768
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1506114285714287e-06,
+      "loss": 0.779,
+      "step": 40769
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1503428571428573e-06,
+      "loss": 1.069,
+      "step": 40770
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.150074285714286e-06,
+      "loss": 1.3701,
+      "step": 40771
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1498057142857145e-06,
+      "loss": 0.7665,
+      "step": 40772
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.149537142857143e-06,
+      "loss": 1.1409,
+      "step": 40773
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1492685714285712e-06,
+      "loss": 0.9026,
+      "step": 40774
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.149e-06,
+      "loss": 1.2942,
+      "step": 40775
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.148731428571429e-06,
+      "loss": 1.1603,
+      "step": 40776
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1484628571428575e-06,
+      "loss": 1.4803,
+      "step": 40777
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.148194285714286e-06,
+      "loss": 0.9331,
+      "step": 40778
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1479257142857142e-06,
+      "loss": 0.6864,
+      "step": 40779
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.147657142857143e-06,
+      "loss": 1.1651,
+      "step": 40780
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1473885714285714e-06,
+      "loss": 1.3389,
+      "step": 40781
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1471200000000005e-06,
+      "loss": 1.3175,
+      "step": 40782
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1468514285714286e-06,
+      "loss": 1.104,
+      "step": 40783
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1465828571428572e-06,
+      "loss": 1.3935,
+      "step": 40784
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.146314285714286e-06,
+      "loss": 1.3585,
+      "step": 40785
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1460457142857144e-06,
+      "loss": 1.4651,
+      "step": 40786
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.145777142857143e-06,
+      "loss": 1.0961,
+      "step": 40787
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.145508571428571e-06,
+      "loss": 1.6293,
+      "step": 40788
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1452400000000002e-06,
+      "loss": 1.1204,
+      "step": 40789
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.144971428571429e-06,
+      "loss": 0.9458,
+      "step": 40790
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1447028571428574e-06,
+      "loss": 1.1577,
+      "step": 40791
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1444342857142856e-06,
+      "loss": 1.177,
+      "step": 40792
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.144165714285714e-06,
+      "loss": 1.3584,
+      "step": 40793
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.143897142857143e-06,
+      "loss": 1.2939,
+      "step": 40794
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.143628571428572e-06,
+      "loss": 1.0811,
+      "step": 40795
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1433600000000004e-06,
+      "loss": 1.2299,
+      "step": 40796
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1430914285714286e-06,
+      "loss": 1.2226,
+      "step": 40797
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.142822857142857e-06,
+      "loss": 0.9275,
+      "step": 40798
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.142554285714286e-06,
+      "loss": 1.1271,
+      "step": 40799
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1422857142857144e-06,
+      "loss": 1.0211,
+      "step": 40800
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.142017142857143e-06,
+      "loss": 0.9541,
+      "step": 40801
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1417485714285716e-06,
+      "loss": 1.1174,
+      "step": 40802
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.14148e-06,
+      "loss": 1.1233,
+      "step": 40803
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.141211428571429e-06,
+      "loss": 0.9723,
+      "step": 40804
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1409428571428574e-06,
+      "loss": 1.0273,
+      "step": 40805
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1406742857142856e-06,
+      "loss": 1.3771,
+      "step": 40806
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.140405714285714e-06,
+      "loss": 1.3312,
+      "step": 40807
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1401371428571428e-06,
+      "loss": 1.1813,
+      "step": 40808
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.139868571428572e-06,
+      "loss": 1.1775,
+      "step": 40809
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1396000000000004e-06,
+      "loss": 1.15,
+      "step": 40810
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1393314285714286e-06,
+      "loss": 1.553,
+      "step": 40811
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.139062857142857e-06,
+      "loss": 1.3708,
+      "step": 40812
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1387942857142858e-06,
+      "loss": 0.8055,
+      "step": 40813
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1385257142857144e-06,
+      "loss": 0.8867,
+      "step": 40814
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.138257142857143e-06,
+      "loss": 1.5064,
+      "step": 40815
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1379885714285716e-06,
+      "loss": 1.0912,
+      "step": 40816
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.13772e-06,
+      "loss": 1.0206,
+      "step": 40817
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1374514285714288e-06,
+      "loss": 0.9925,
+      "step": 40818
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1371828571428574e-06,
+      "loss": 1.0417,
+      "step": 40819
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1369142857142855e-06,
+      "loss": 0.8631,
+      "step": 40820
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.136645714285714e-06,
+      "loss": 1.0944,
+      "step": 40821
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.136377142857143e-06,
+      "loss": 1.2541,
+      "step": 40822
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1361085714285718e-06,
+      "loss": 1.3215,
+      "step": 40823
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.13584e-06,
+      "loss": 1.0768,
+      "step": 40824
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1355714285714285e-06,
+      "loss": 1.0522,
+      "step": 40825
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.135302857142857e-06,
+      "loss": 0.6978,
+      "step": 40826
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1350342857142857e-06,
+      "loss": 1.099,
+      "step": 40827
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1347657142857143e-06,
+      "loss": 1.0318,
+      "step": 40828
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.134497142857143e-06,
+      "loss": 1.0299,
+      "step": 40829
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1342285714285715e-06,
+      "loss": 0.7682,
+      "step": 40830
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.13396e-06,
+      "loss": 1.3414,
+      "step": 40831
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1336914285714287e-06,
+      "loss": 1.0854,
+      "step": 40832
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1334228571428573e-06,
+      "loss": 0.8209,
+      "step": 40833
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1331542857142855e-06,
+      "loss": 1.2639,
+      "step": 40834
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1328857142857145e-06,
+      "loss": 1.3156,
+      "step": 40835
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.132617142857143e-06,
+      "loss": 1.2109,
+      "step": 40836
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1323485714285717e-06,
+      "loss": 0.9992,
+      "step": 40837
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.13208e-06,
+      "loss": 1.2238,
+      "step": 40838
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1318114285714285e-06,
+      "loss": 0.9474,
+      "step": 40839
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.131542857142857e-06,
+      "loss": 0.7578,
+      "step": 40840
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1312742857142857e-06,
+      "loss": 1.3331,
+      "step": 40841
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1310057142857147e-06,
+      "loss": 0.808,
+      "step": 40842
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.130737142857143e-06,
+      "loss": 1.1138,
+      "step": 40843
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1304685714285715e-06,
+      "loss": 1.5433,
+      "step": 40844
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1302e-06,
+      "loss": 1.0207,
+      "step": 40845
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1299314285714287e-06,
+      "loss": 1.2088,
+      "step": 40846
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.129662857142857e-06,
+      "loss": 1.1304,
+      "step": 40847
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.129394285714286e-06,
+      "loss": 1.2945,
+      "step": 40848
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1291257142857145e-06,
+      "loss": 0.5713,
+      "step": 40849
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.128857142857143e-06,
+      "loss": 1.0804,
+      "step": 40850
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1285885714285717e-06,
+      "loss": 1.4803,
+      "step": 40851
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.12832e-06,
+      "loss": 0.5604,
+      "step": 40852
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1280514285714285e-06,
+      "loss": 1.0206,
+      "step": 40853
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.127782857142857e-06,
+      "loss": 0.8012,
+      "step": 40854
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.127514285714286e-06,
+      "loss": 1.0609,
+      "step": 40855
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1272457142857143e-06,
+      "loss": 1.3336,
+      "step": 40856
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.126977142857143e-06,
+      "loss": 1.1001,
+      "step": 40857
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1267085714285715e-06,
+      "loss": 0.7173,
+      "step": 40858
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.12644e-06,
+      "loss": 0.7789,
+      "step": 40859
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1261714285714287e-06,
+      "loss": 0.9512,
+      "step": 40860
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.1259028571428573e-06,
+      "loss": 1.5443,
+      "step": 40861
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.125634285714286e-06,
+      "loss": 0.8968,
+      "step": 40862
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1253657142857145e-06,
+      "loss": 0.8517,
+      "step": 40863
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.125097142857143e-06,
+      "loss": 1.0059,
+      "step": 40864
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1248285714285717e-06,
+      "loss": 1.09,
+      "step": 40865
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.12456e-06,
+      "loss": 0.6723,
+      "step": 40866
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1242914285714285e-06,
+      "loss": 1.3397,
+      "step": 40867
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1240228571428575e-06,
+      "loss": 1.3967,
+      "step": 40868
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.123754285714286e-06,
+      "loss": 0.9707,
+      "step": 40869
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1234857142857143e-06,
+      "loss": 1.0799,
+      "step": 40870
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.123217142857143e-06,
+      "loss": 1.1068,
+      "step": 40871
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1229485714285715e-06,
+      "loss": 1.0446,
+      "step": 40872
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.12268e-06,
+      "loss": 1.4188,
+      "step": 40873
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1224114285714287e-06,
+      "loss": 0.9212,
+      "step": 40874
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1221428571428573e-06,
+      "loss": 1.0965,
+      "step": 40875
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.121874285714286e-06,
+      "loss": 1.338,
+      "step": 40876
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1216057142857145e-06,
+      "loss": 1.1595,
+      "step": 40877
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.121337142857143e-06,
+      "loss": 1.0128,
+      "step": 40878
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1210685714285712e-06,
+      "loss": 1.2492,
+      "step": 40879
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1208e-06,
+      "loss": 1.1265,
+      "step": 40880
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.120531428571429e-06,
+      "loss": 1.0202,
+      "step": 40881
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1202628571428575e-06,
+      "loss": 1.2256,
+      "step": 40882
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.119994285714286e-06,
+      "loss": 0.9288,
+      "step": 40883
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1197257142857142e-06,
+      "loss": 1.2231,
+      "step": 40884
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.119457142857143e-06,
+      "loss": 1.0359,
+      "step": 40885
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1191885714285714e-06,
+      "loss": 0.6501,
+      "step": 40886
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.11892e-06,
+      "loss": 0.9584,
+      "step": 40887
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.118651428571429e-06,
+      "loss": 0.7785,
+      "step": 40888
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1183828571428572e-06,
+      "loss": 1.3195,
+      "step": 40889
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.118114285714286e-06,
+      "loss": 0.7607,
+      "step": 40890
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1178457142857144e-06,
+      "loss": 1.4949,
+      "step": 40891
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.117577142857143e-06,
+      "loss": 1.3594,
+      "step": 40892
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.117308571428571e-06,
+      "loss": 1.0574,
+      "step": 40893
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1170400000000002e-06,
+      "loss": 1.2051,
+      "step": 40894
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.116771428571429e-06,
+      "loss": 1.1974,
+      "step": 40895
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1165028571428574e-06,
+      "loss": 1.2455,
+      "step": 40896
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.116234285714286e-06,
+      "loss": 0.9124,
+      "step": 40897
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.115965714285714e-06,
+      "loss": 1.1151,
+      "step": 40898
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.115697142857143e-06,
+      "loss": 0.9242,
+      "step": 40899
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1154285714285714e-06,
+      "loss": 1.2823,
+      "step": 40900
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1151600000000004e-06,
+      "loss": 1.112,
+      "step": 40901
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1148914285714286e-06,
+      "loss": 1.3688,
+      "step": 40902
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.114622857142857e-06,
+      "loss": 1.0525,
+      "step": 40903
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.114354285714286e-06,
+      "loss": 0.9951,
+      "step": 40904
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1140857142857144e-06,
+      "loss": 1.2363,
+      "step": 40905
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.113817142857143e-06,
+      "loss": 1.2272,
+      "step": 40906
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1135485714285716e-06,
+      "loss": 1.2871,
+      "step": 40907
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.11328e-06,
+      "loss": 1.0893,
+      "step": 40908
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.113011428571429e-06,
+      "loss": 0.8248,
+      "step": 40909
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1127428571428574e-06,
+      "loss": 1.381,
+      "step": 40910
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1124742857142856e-06,
+      "loss": 1.2761,
+      "step": 40911
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.112205714285714e-06,
+      "loss": 1.3264,
+      "step": 40912
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1119371428571428e-06,
+      "loss": 1.1205,
+      "step": 40913
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.111668571428572e-06,
+      "loss": 0.7059,
+      "step": 40914
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1114000000000004e-06,
+      "loss": 1.3607,
+      "step": 40915
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1111314285714286e-06,
+      "loss": 1.3161,
+      "step": 40916
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.110862857142857e-06,
+      "loss": 1.1205,
+      "step": 40917
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1105942857142858e-06,
+      "loss": 1.4078,
+      "step": 40918
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1103257142857144e-06,
+      "loss": 0.9434,
+      "step": 40919
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.110057142857143e-06,
+      "loss": 0.8177,
+      "step": 40920
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1097885714285716e-06,
+      "loss": 0.789,
+      "step": 40921
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.10952e-06,
+      "loss": 1.339,
+      "step": 40922
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1092514285714288e-06,
+      "loss": 0.8784,
+      "step": 40923
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1089828571428574e-06,
+      "loss": 1.2794,
+      "step": 40924
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1087142857142855e-06,
+      "loss": 1.4951,
+      "step": 40925
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.108445714285714e-06,
+      "loss": 0.6906,
+      "step": 40926
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.108177142857143e-06,
+      "loss": 1.6441,
+      "step": 40927
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1079085714285718e-06,
+      "loss": 1.3947,
+      "step": 40928
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1076400000000004e-06,
+      "loss": 0.9516,
+      "step": 40929
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1073714285714285e-06,
+      "loss": 0.9086,
+      "step": 40930
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.107102857142857e-06,
+      "loss": 1.1575,
+      "step": 40931
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1068342857142857e-06,
+      "loss": 0.9644,
+      "step": 40932
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1065657142857143e-06,
+      "loss": 1.3847,
+      "step": 40933
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.106297142857143e-06,
+      "loss": 1.178,
+      "step": 40934
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1060285714285715e-06,
+      "loss": 1.3838,
+      "step": 40935
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.10576e-06,
+      "loss": 1.3408,
+      "step": 40936
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1054914285714287e-06,
+      "loss": 1.1996,
+      "step": 40937
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1052228571428573e-06,
+      "loss": 1.0365,
+      "step": 40938
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1049542857142855e-06,
+      "loss": 1.5274,
+      "step": 40939
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1046857142857145e-06,
+      "loss": 1.1289,
+      "step": 40940
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.104417142857143e-06,
+      "loss": 1.0353,
+      "step": 40941
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1041485714285717e-06,
+      "loss": 1.2963,
+      "step": 40942
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.10388e-06,
+      "loss": 1.205,
+      "step": 40943
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1036114285714285e-06,
+      "loss": 1.034,
+      "step": 40944
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.103342857142857e-06,
+      "loss": 1.1779,
+      "step": 40945
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1030742857142857e-06,
+      "loss": 1.2164,
+      "step": 40946
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1028057142857147e-06,
+      "loss": 1.6087,
+      "step": 40947
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.102537142857143e-06,
+      "loss": 1.2969,
+      "step": 40948
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1022685714285715e-06,
+      "loss": 0.7776,
+      "step": 40949
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.102e-06,
+      "loss": 1.0157,
+      "step": 40950
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1017314285714287e-06,
+      "loss": 1.1067,
+      "step": 40951
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1014628571428573e-06,
+      "loss": 1.1304,
+      "step": 40952
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.101194285714286e-06,
+      "loss": 1.0645,
+      "step": 40953
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1009257142857145e-06,
+      "loss": 1.3318,
+      "step": 40954
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.100657142857143e-06,
+      "loss": 1.1917,
+      "step": 40955
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.1003885714285717e-06,
+      "loss": 1.4359,
+      "step": 40956
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.10012e-06,
+      "loss": 0.7311,
+      "step": 40957
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0998514285714285e-06,
+      "loss": 1.2473,
+      "step": 40958
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.099582857142857e-06,
+      "loss": 0.9236,
+      "step": 40959
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.099314285714286e-06,
+      "loss": 1.1996,
+      "step": 40960
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0990457142857147e-06,
+      "loss": 1.2592,
+      "step": 40961
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.098777142857143e-06,
+      "loss": 1.5433,
+      "step": 40962
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0985085714285715e-06,
+      "loss": 1.0535,
+      "step": 40963
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.09824e-06,
+      "loss": 1.0339,
+      "step": 40964
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0979714285714287e-06,
+      "loss": 1.1273,
+      "step": 40965
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0977028571428573e-06,
+      "loss": 0.9412,
+      "step": 40966
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.097434285714286e-06,
+      "loss": 0.9798,
+      "step": 40967
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0971657142857145e-06,
+      "loss": 0.8937,
+      "step": 40968
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.096897142857143e-06,
+      "loss": 1.0766,
+      "step": 40969
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0966285714285717e-06,
+      "loss": 1.5614,
+      "step": 40970
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.09636e-06,
+      "loss": 1.0042,
+      "step": 40971
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0960914285714285e-06,
+      "loss": 0.7591,
+      "step": 40972
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0958228571428575e-06,
+      "loss": 1.0929,
+      "step": 40973
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.095554285714286e-06,
+      "loss": 1.6057,
+      "step": 40974
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0952857142857143e-06,
+      "loss": 1.1774,
+      "step": 40975
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.095017142857143e-06,
+      "loss": 0.8415,
+      "step": 40976
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0947485714285715e-06,
+      "loss": 1.0693,
+      "step": 40977
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.09448e-06,
+      "loss": 1.4615,
+      "step": 40978
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0942114285714287e-06,
+      "loss": 1.2066,
+      "step": 40979
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0939428571428573e-06,
+      "loss": 0.8731,
+      "step": 40980
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.093674285714286e-06,
+      "loss": 0.8859,
+      "step": 40981
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0934057142857145e-06,
+      "loss": 1.207,
+      "step": 40982
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.093137142857143e-06,
+      "loss": 0.6382,
+      "step": 40983
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0928685714285717e-06,
+      "loss": 1.2115,
+      "step": 40984
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0926e-06,
+      "loss": 1.0312,
+      "step": 40985
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.092331428571429e-06,
+      "loss": 1.1548,
+      "step": 40986
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0920628571428575e-06,
+      "loss": 1.0209,
+      "step": 40987
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.091794285714286e-06,
+      "loss": 1.2686,
+      "step": 40988
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0915257142857142e-06,
+      "loss": 1.3441,
+      "step": 40989
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.091257142857143e-06,
+      "loss": 0.9018,
+      "step": 40990
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0909885714285714e-06,
+      "loss": 1.1709,
+      "step": 40991
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.09072e-06,
+      "loss": 1.2035,
+      "step": 40992
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.090451428571429e-06,
+      "loss": 1.1879,
+      "step": 40993
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0901828571428572e-06,
+      "loss": 0.6659,
+      "step": 40994
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.089914285714286e-06,
+      "loss": 1.3978,
+      "step": 40995
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0896457142857144e-06,
+      "loss": 1.6735,
+      "step": 40996
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.089377142857143e-06,
+      "loss": 1.0033,
+      "step": 40997
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.089108571428571e-06,
+      "loss": 1.0715,
+      "step": 40998
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.08884e-06,
+      "loss": 1.4182,
+      "step": 40999
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.088571428571429e-06,
+      "loss": 1.0245,
+      "step": 41000
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0883028571428574e-06,
+      "loss": 0.8153,
+      "step": 41001
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.088034285714286e-06,
+      "loss": 1.7112,
+      "step": 41002
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.087765714285714e-06,
+      "loss": 0.5866,
+      "step": 41003
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.087497142857143e-06,
+      "loss": 1.381,
+      "step": 41004
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0872285714285714e-06,
+      "loss": 1.1976,
+      "step": 41005
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0869600000000004e-06,
+      "loss": 1.1487,
+      "step": 41006
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.086691428571429e-06,
+      "loss": 1.3822,
+      "step": 41007
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.086422857142857e-06,
+      "loss": 1.4612,
+      "step": 41008
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.086154285714286e-06,
+      "loss": 1.0572,
+      "step": 41009
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0858857142857144e-06,
+      "loss": 1.2848,
+      "step": 41010
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.085617142857143e-06,
+      "loss": 1.4686,
+      "step": 41011
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.085348571428571e-06,
+      "loss": 1.1044,
+      "step": 41012
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.08508e-06,
+      "loss": 0.915,
+      "step": 41013
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.084811428571429e-06,
+      "loss": 1.1527,
+      "step": 41014
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0845428571428574e-06,
+      "loss": 1.4082,
+      "step": 41015
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.084274285714286e-06,
+      "loss": 1.182,
+      "step": 41016
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.084005714285714e-06,
+      "loss": 1.057,
+      "step": 41017
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0837371428571428e-06,
+      "loss": 1.1387,
+      "step": 41018
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.083468571428572e-06,
+      "loss": 1.3329,
+      "step": 41019
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0832000000000004e-06,
+      "loss": 1.5469,
+      "step": 41020
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0829314285714286e-06,
+      "loss": 0.9825,
+      "step": 41021
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.082662857142857e-06,
+      "loss": 1.0873,
+      "step": 41022
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0823942857142858e-06,
+      "loss": 1.2883,
+      "step": 41023
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0821257142857144e-06,
+      "loss": 1.226,
+      "step": 41024
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.081857142857143e-06,
+      "loss": 1.1488,
+      "step": 41025
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0815885714285716e-06,
+      "loss": 1.0821,
+      "step": 41026
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.08132e-06,
+      "loss": 0.8517,
+      "step": 41027
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0810514285714288e-06,
+      "loss": 1.4184,
+      "step": 41028
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0807828571428574e-06,
+      "loss": 1.0353,
+      "step": 41029
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0805142857142855e-06,
+      "loss": 1.0363,
+      "step": 41030
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.080245714285714e-06,
+      "loss": 0.7882,
+      "step": 41031
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0799771428571427e-06,
+      "loss": 1.4576,
+      "step": 41032
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0797085714285718e-06,
+      "loss": 0.894,
+      "step": 41033
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0794400000000004e-06,
+      "loss": 0.9862,
+      "step": 41034
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0791714285714285e-06,
+      "loss": 1.0336,
+      "step": 41035
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.078902857142857e-06,
+      "loss": 1.2161,
+      "step": 41036
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.0786342857142857e-06,
+      "loss": 1.0061,
+      "step": 41037
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0783657142857143e-06,
+      "loss": 1.277,
+      "step": 41038
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0780971428571434e-06,
+      "loss": 0.927,
+      "step": 41039
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0778285714285715e-06,
+      "loss": 1.4693,
+      "step": 41040
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.07756e-06,
+      "loss": 1.3154,
+      "step": 41041
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0772914285714287e-06,
+      "loss": 1.3856,
+      "step": 41042
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0770228571428573e-06,
+      "loss": 1.141,
+      "step": 41043
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0767542857142855e-06,
+      "loss": 0.9313,
+      "step": 41044
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.076485714285714e-06,
+      "loss": 1.395,
+      "step": 41045
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.076217142857143e-06,
+      "loss": 1.2258,
+      "step": 41046
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0759485714285717e-06,
+      "loss": 1.0621,
+      "step": 41047
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0756800000000003e-06,
+      "loss": 0.9554,
+      "step": 41048
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0754114285714285e-06,
+      "loss": 1.2899,
+      "step": 41049
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.075142857142857e-06,
+      "loss": 0.8089,
+      "step": 41050
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0748742857142857e-06,
+      "loss": 1.06,
+      "step": 41051
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0746057142857143e-06,
+      "loss": 0.935,
+      "step": 41052
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.074337142857143e-06,
+      "loss": 1.0959,
+      "step": 41053
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0740685714285715e-06,
+      "loss": 1.1134,
+      "step": 41054
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0738e-06,
+      "loss": 1.1581,
+      "step": 41055
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0735314285714287e-06,
+      "loss": 1.2599,
+      "step": 41056
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0732628571428573e-06,
+      "loss": 1.5392,
+      "step": 41057
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0729942857142855e-06,
+      "loss": 0.876,
+      "step": 41058
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0727257142857145e-06,
+      "loss": 1.3178,
+      "step": 41059
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.072457142857143e-06,
+      "loss": 1.3975,
+      "step": 41060
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0721885714285717e-06,
+      "loss": 0.9935,
+      "step": 41061
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.07192e-06,
+      "loss": 1.2872,
+      "step": 41062
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0716514285714285e-06,
+      "loss": 1.1485,
+      "step": 41063
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.071382857142857e-06,
+      "loss": 1.1697,
+      "step": 41064
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0711142857142857e-06,
+      "loss": 0.6694,
+      "step": 41065
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0708457142857147e-06,
+      "loss": 1.2835,
+      "step": 41066
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.070577142857143e-06,
+      "loss": 1.1953,
+      "step": 41067
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0703085714285715e-06,
+      "loss": 1.1449,
+      "step": 41068
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.07004e-06,
+      "loss": 0.9596,
+      "step": 41069
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0697714285714287e-06,
+      "loss": 1.152,
+      "step": 41070
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0695028571428573e-06,
+      "loss": 1.095,
+      "step": 41071
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.069234285714286e-06,
+      "loss": 1.1486,
+      "step": 41072
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0689657142857145e-06,
+      "loss": 0.7951,
+      "step": 41073
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.068697142857143e-06,
+      "loss": 0.9759,
+      "step": 41074
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0684285714285717e-06,
+      "loss": 1.0355,
+      "step": 41075
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.06816e-06,
+      "loss": 1.0843,
+      "step": 41076
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0678914285714285e-06,
+      "loss": 1.7923,
+      "step": 41077
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.067622857142857e-06,
+      "loss": 1.0028,
+      "step": 41078
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.067354285714286e-06,
+      "loss": 1.328,
+      "step": 41079
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0670857142857147e-06,
+      "loss": 1.218,
+      "step": 41080
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.066817142857143e-06,
+      "loss": 1.0725,
+      "step": 41081
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0665485714285715e-06,
+      "loss": 1.521,
+      "step": 41082
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.06628e-06,
+      "loss": 1.1889,
+      "step": 41083
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0660114285714287e-06,
+      "loss": 1.0099,
+      "step": 41084
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0657428571428573e-06,
+      "loss": 1.3905,
+      "step": 41085
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.065474285714286e-06,
+      "loss": 0.976,
+      "step": 41086
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0652057142857145e-06,
+      "loss": 1.4021,
+      "step": 41087
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.064937142857143e-06,
+      "loss": 1.1628,
+      "step": 41088
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0646685714285717e-06,
+      "loss": 1.0408,
+      "step": 41089
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0644e-06,
+      "loss": 0.9625,
+      "step": 41090
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0641314285714284e-06,
+      "loss": 1.0436,
+      "step": 41091
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0638628571428575e-06,
+      "loss": 0.7926,
+      "step": 41092
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.063594285714286e-06,
+      "loss": 1.349,
+      "step": 41093
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0633257142857147e-06,
+      "loss": 1.1828,
+      "step": 41094
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.063057142857143e-06,
+      "loss": 1.051,
+      "step": 41095
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0627885714285714e-06,
+      "loss": 0.8607,
+      "step": 41096
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.06252e-06,
+      "loss": 1.1878,
+      "step": 41097
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0622514285714286e-06,
+      "loss": 1.1804,
+      "step": 41098
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0619828571428572e-06,
+      "loss": 1.1503,
+      "step": 41099
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.061714285714286e-06,
+      "loss": 1.0501,
+      "step": 41100
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0614457142857144e-06,
+      "loss": 1.5143,
+      "step": 41101
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.061177142857143e-06,
+      "loss": 1.2982,
+      "step": 41102
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0609085714285716e-06,
+      "loss": 1.0559,
+      "step": 41103
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.06064e-06,
+      "loss": 1.2526,
+      "step": 41104
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.060371428571429e-06,
+      "loss": 1.2099,
+      "step": 41105
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0601028571428574e-06,
+      "loss": 0.8665,
+      "step": 41106
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.059834285714286e-06,
+      "loss": 1.0705,
+      "step": 41107
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.059565714285714e-06,
+      "loss": 1.0331,
+      "step": 41108
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.059297142857143e-06,
+      "loss": 1.2429,
+      "step": 41109
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0590285714285714e-06,
+      "loss": 0.9269,
+      "step": 41110
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.05876e-06,
+      "loss": 0.7079,
+      "step": 41111
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.058491428571429e-06,
+      "loss": 0.9207,
+      "step": 41112
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.058222857142857e-06,
+      "loss": 1.2312,
+      "step": 41113
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.057954285714286e-06,
+      "loss": 0.8027,
+      "step": 41114
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0576857142857144e-06,
+      "loss": 1.1941,
+      "step": 41115
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.057417142857143e-06,
+      "loss": 1.2075,
+      "step": 41116
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.057148571428571e-06,
+      "loss": 0.7561,
+      "step": 41117
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.05688e-06,
+      "loss": 1.1348,
+      "step": 41118
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.056611428571429e-06,
+      "loss": 1.4041,
+      "step": 41119
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0563428571428574e-06,
+      "loss": 1.3461,
+      "step": 41120
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.056074285714286e-06,
+      "loss": 1.0718,
+      "step": 41121
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.055805714285714e-06,
+      "loss": 0.8861,
+      "step": 41122
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0555371428571428e-06,
+      "loss": 1.0621,
+      "step": 41123
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0552685714285714e-06,
+      "loss": 1.2545,
+      "step": 41124
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0550000000000004e-06,
+      "loss": 1.3967,
+      "step": 41125
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.054731428571429e-06,
+      "loss": 0.9596,
+      "step": 41126
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.054462857142857e-06,
+      "loss": 1.0747,
+      "step": 41127
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0541942857142858e-06,
+      "loss": 0.8954,
+      "step": 41128
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0539257142857144e-06,
+      "loss": 0.9689,
+      "step": 41129
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.053657142857143e-06,
+      "loss": 0.7004,
+      "step": 41130
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0533885714285716e-06,
+      "loss": 1.2758,
+      "step": 41131
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.05312e-06,
+      "loss": 1.6036,
+      "step": 41132
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0528514285714288e-06,
+      "loss": 0.9915,
+      "step": 41133
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0525828571428574e-06,
+      "loss": 1.0288,
+      "step": 41134
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.052314285714286e-06,
+      "loss": 0.9295,
+      "step": 41135
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.052045714285714e-06,
+      "loss": 1.1906,
+      "step": 41136
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0517771428571427e-06,
+      "loss": 0.5852,
+      "step": 41137
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0515085714285718e-06,
+      "loss": 1.5541,
+      "step": 41138
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0512400000000004e-06,
+      "loss": 0.9613,
+      "step": 41139
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0509714285714285e-06,
+      "loss": 1.1625,
+      "step": 41140
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.050702857142857e-06,
+      "loss": 1.1809,
+      "step": 41141
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0504342857142857e-06,
+      "loss": 1.3673,
+      "step": 41142
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0501657142857143e-06,
+      "loss": 0.9593,
+      "step": 41143
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.049897142857143e-06,
+      "loss": 1.1707,
+      "step": 41144
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0496285714285715e-06,
+      "loss": 1.2918,
+      "step": 41145
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.04936e-06,
+      "loss": 0.8578,
+      "step": 41146
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0490914285714287e-06,
+      "loss": 1.3617,
+      "step": 41147
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0488228571428573e-06,
+      "loss": 1.1407,
+      "step": 41148
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0485542857142855e-06,
+      "loss": 1.0784,
+      "step": 41149
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.048285714285714e-06,
+      "loss": 1.0052,
+      "step": 41150
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.048017142857143e-06,
+      "loss": 1.1048,
+      "step": 41151
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0477485714285717e-06,
+      "loss": 1.0952,
+      "step": 41152
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0474800000000003e-06,
+      "loss": 1.0398,
+      "step": 41153
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0472114285714285e-06,
+      "loss": 1.0603,
+      "step": 41154
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.046942857142857e-06,
+      "loss": 1.1241,
+      "step": 41155
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0466742857142857e-06,
+      "loss": 1.1836,
+      "step": 41156
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0464057142857143e-06,
+      "loss": 1.4336,
+      "step": 41157
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0461371428571433e-06,
+      "loss": 1.3217,
+      "step": 41158
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0458685714285715e-06,
+      "loss": 0.9159,
+      "step": 41159
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0456e-06,
+      "loss": 1.1822,
+      "step": 41160
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0453314285714287e-06,
+      "loss": 0.9909,
+      "step": 41161
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0450628571428573e-06,
+      "loss": 1.0134,
+      "step": 41162
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0447942857142855e-06,
+      "loss": 1.1079,
+      "step": 41163
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0445257142857145e-06,
+      "loss": 1.1398,
+      "step": 41164
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.044257142857143e-06,
+      "loss": 1.101,
+      "step": 41165
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0439885714285717e-06,
+      "loss": 1.1386,
+      "step": 41166
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0437200000000003e-06,
+      "loss": 0.8835,
+      "step": 41167
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0434514285714285e-06,
+      "loss": 1.0395,
+      "step": 41168
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.043182857142857e-06,
+      "loss": 1.0655,
+      "step": 41169
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0429142857142857e-06,
+      "loss": 0.9569,
+      "step": 41170
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0426457142857147e-06,
+      "loss": 1.1984,
+      "step": 41171
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.042377142857143e-06,
+      "loss": 0.7038,
+      "step": 41172
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0421085714285715e-06,
+      "loss": 1.0498,
+      "step": 41173
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.04184e-06,
+      "loss": 1.1227,
+      "step": 41174
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0415714285714287e-06,
+      "loss": 1.6277,
+      "step": 41175
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0413028571428573e-06,
+      "loss": 1.5365,
+      "step": 41176
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.041034285714286e-06,
+      "loss": 1.3527,
+      "step": 41177
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0407657142857145e-06,
+      "loss": 0.684,
+      "step": 41178
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.040497142857143e-06,
+      "loss": 1.3836,
+      "step": 41179
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0402285714285717e-06,
+      "loss": 1.2144,
+      "step": 41180
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0399600000000003e-06,
+      "loss": 1.3344,
+      "step": 41181
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0396914285714285e-06,
+      "loss": 1.0913,
+      "step": 41182
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.039422857142857e-06,
+      "loss": 1.301,
+      "step": 41183
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.039154285714286e-06,
+      "loss": 1.7901,
+      "step": 41184
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0388857142857147e-06,
+      "loss": 1.3568,
+      "step": 41185
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.038617142857143e-06,
+      "loss": 0.9814,
+      "step": 41186
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0383485714285715e-06,
+      "loss": 0.8029,
+      "step": 41187
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.03808e-06,
+      "loss": 1.4933,
+      "step": 41188
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0378114285714287e-06,
+      "loss": 1.1266,
+      "step": 41189
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0375428571428573e-06,
+      "loss": 1.2244,
+      "step": 41190
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.037274285714286e-06,
+      "loss": 1.3868,
+      "step": 41191
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0370057142857145e-06,
+      "loss": 1.1755,
+      "step": 41192
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.036737142857143e-06,
+      "loss": 1.7296,
+      "step": 41193
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0364685714285717e-06,
+      "loss": 1.012,
+      "step": 41194
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0362e-06,
+      "loss": 1.3322,
+      "step": 41195
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0359314285714284e-06,
+      "loss": 0.759,
+      "step": 41196
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0356628571428575e-06,
+      "loss": 1.2331,
+      "step": 41197
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.035394285714286e-06,
+      "loss": 1.2036,
+      "step": 41198
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0351257142857147e-06,
+      "loss": 0.919,
+      "step": 41199
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.034857142857143e-06,
+      "loss": 0.907,
+      "step": 41200
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0345885714285714e-06,
+      "loss": 1.3305,
+      "step": 41201
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.03432e-06,
+      "loss": 0.8096,
+      "step": 41202
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0340514285714286e-06,
+      "loss": 1.3215,
+      "step": 41203
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0337828571428572e-06,
+      "loss": 0.9936,
+      "step": 41204
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.033514285714286e-06,
+      "loss": 1.3572,
+      "step": 41205
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0332457142857144e-06,
+      "loss": 1.2128,
+      "step": 41206
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.032977142857143e-06,
+      "loss": 1.388,
+      "step": 41207
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0327085714285716e-06,
+      "loss": 1.1075,
+      "step": 41208
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.03244e-06,
+      "loss": 0.9726,
+      "step": 41209
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.032171428571429e-06,
+      "loss": 0.8397,
+      "step": 41210
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.0319028571428574e-06,
+      "loss": 0.8993,
+      "step": 41211
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.031634285714286e-06,
+      "loss": 0.9513,
+      "step": 41212
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0313657142857146e-06,
+      "loss": 0.8772,
+      "step": 41213
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.031097142857143e-06,
+      "loss": 0.9696,
+      "step": 41214
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0308285714285714e-06,
+      "loss": 1.2634,
+      "step": 41215
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.03056e-06,
+      "loss": 0.98,
+      "step": 41216
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.030291428571429e-06,
+      "loss": 1.0412,
+      "step": 41217
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.030022857142857e-06,
+      "loss": 1.4152,
+      "step": 41218
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.029754285714286e-06,
+      "loss": 0.8754,
+      "step": 41219
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0294857142857144e-06,
+      "loss": 0.7437,
+      "step": 41220
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.029217142857143e-06,
+      "loss": 1.3015,
+      "step": 41221
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0289485714285716e-06,
+      "loss": 1.0664,
+      "step": 41222
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0286799999999998e-06,
+      "loss": 1.1643,
+      "step": 41223
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.028411428571429e-06,
+      "loss": 1.2053,
+      "step": 41224
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0281428571428574e-06,
+      "loss": 1.1234,
+      "step": 41225
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.027874285714286e-06,
+      "loss": 0.5624,
+      "step": 41226
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.027605714285714e-06,
+      "loss": 0.9956,
+      "step": 41227
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0273371428571428e-06,
+      "loss": 0.9101,
+      "step": 41228
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0270685714285714e-06,
+      "loss": 1.3066,
+      "step": 41229
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0268000000000004e-06,
+      "loss": 1.0008,
+      "step": 41230
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.026531428571429e-06,
+      "loss": 0.8522,
+      "step": 41231
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.026262857142857e-06,
+      "loss": 1.5177,
+      "step": 41232
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0259942857142858e-06,
+      "loss": 0.8077,
+      "step": 41233
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0257257142857144e-06,
+      "loss": 1.1214,
+      "step": 41234
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.025457142857143e-06,
+      "loss": 1.1274,
+      "step": 41235
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.025188571428571e-06,
+      "loss": 1.047,
+      "step": 41236
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.02492e-06,
+      "loss": 1.0244,
+      "step": 41237
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0246514285714288e-06,
+      "loss": 1.0734,
+      "step": 41238
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0243828571428574e-06,
+      "loss": 0.9654,
+      "step": 41239
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.024114285714286e-06,
+      "loss": 0.7763,
+      "step": 41240
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.023845714285714e-06,
+      "loss": 0.9333,
+      "step": 41241
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0235771428571427e-06,
+      "loss": 0.9897,
+      "step": 41242
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0233085714285718e-06,
+      "loss": 1.0067,
+      "step": 41243
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0230400000000004e-06,
+      "loss": 1.355,
+      "step": 41244
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.022771428571429e-06,
+      "loss": 1.2768,
+      "step": 41245
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.022502857142857e-06,
+      "loss": 0.9293,
+      "step": 41246
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0222342857142857e-06,
+      "loss": 1.3527,
+      "step": 41247
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0219657142857143e-06,
+      "loss": 1.2962,
+      "step": 41248
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.021697142857143e-06,
+      "loss": 1.1165,
+      "step": 41249
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0214285714285715e-06,
+      "loss": 1.3029,
+      "step": 41250
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.02116e-06,
+      "loss": 1.1762,
+      "step": 41251
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0208914285714287e-06,
+      "loss": 1.3545,
+      "step": 41252
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0206228571428573e-06,
+      "loss": 1.1057,
+      "step": 41253
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.020354285714286e-06,
+      "loss": 1.3035,
+      "step": 41254
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.020085714285714e-06,
+      "loss": 0.9915,
+      "step": 41255
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0198171428571427e-06,
+      "loss": 1.1071,
+      "step": 41256
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0195485714285717e-06,
+      "loss": 1.3116,
+      "step": 41257
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0192800000000003e-06,
+      "loss": 1.338,
+      "step": 41258
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0190114285714285e-06,
+      "loss": 1.2178,
+      "step": 41259
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.018742857142857e-06,
+      "loss": 1.5293,
+      "step": 41260
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0184742857142857e-06,
+      "loss": 0.9989,
+      "step": 41261
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0182057142857143e-06,
+      "loss": 0.8918,
+      "step": 41262
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0179371428571433e-06,
+      "loss": 1.0321,
+      "step": 41263
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0176685714285715e-06,
+      "loss": 0.9162,
+      "step": 41264
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0174e-06,
+      "loss": 1.1477,
+      "step": 41265
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0171314285714287e-06,
+      "loss": 0.8105,
+      "step": 41266
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0168628571428573e-06,
+      "loss": 1.0177,
+      "step": 41267
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0165942857142855e-06,
+      "loss": 0.5204,
+      "step": 41268
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.016325714285714e-06,
+      "loss": 1.1549,
+      "step": 41269
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.016057142857143e-06,
+      "loss": 1.203,
+      "step": 41270
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0157885714285717e-06,
+      "loss": 1.2883,
+      "step": 41271
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0155200000000003e-06,
+      "loss": 1.1869,
+      "step": 41272
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0152514285714285e-06,
+      "loss": 0.6613,
+      "step": 41273
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.014982857142857e-06,
+      "loss": 0.9526,
+      "step": 41274
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0147142857142857e-06,
+      "loss": 0.9818,
+      "step": 41275
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0144457142857143e-06,
+      "loss": 0.828,
+      "step": 41276
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0141771428571433e-06,
+      "loss": 1.1463,
+      "step": 41277
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0139085714285715e-06,
+      "loss": 1.0663,
+      "step": 41278
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.01364e-06,
+      "loss": 1.0864,
+      "step": 41279
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0133714285714287e-06,
+      "loss": 0.7298,
+      "step": 41280
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0131028571428573e-06,
+      "loss": 1.1623,
+      "step": 41281
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0128342857142855e-06,
+      "loss": 0.9836,
+      "step": 41282
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0125657142857145e-06,
+      "loss": 0.8464,
+      "step": 41283
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.012297142857143e-06,
+      "loss": 1.0749,
+      "step": 41284
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0120285714285717e-06,
+      "loss": 1.1128,
+      "step": 41285
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0117600000000003e-06,
+      "loss": 0.8943,
+      "step": 41286
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0114914285714285e-06,
+      "loss": 0.729,
+      "step": 41287
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.011222857142857e-06,
+      "loss": 0.9141,
+      "step": 41288
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0109542857142857e-06,
+      "loss": 1.6985,
+      "step": 41289
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0106857142857147e-06,
+      "loss": 0.9024,
+      "step": 41290
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.010417142857143e-06,
+      "loss": 1.3299,
+      "step": 41291
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0101485714285715e-06,
+      "loss": 1.0347,
+      "step": 41292
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.00988e-06,
+      "loss": 1.0584,
+      "step": 41293
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0096114285714287e-06,
+      "loss": 1.2509,
+      "step": 41294
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0093428571428573e-06,
+      "loss": 1.2199,
+      "step": 41295
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.009074285714286e-06,
+      "loss": 0.9038,
+      "step": 41296
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0088057142857145e-06,
+      "loss": 1.1034,
+      "step": 41297
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.008537142857143e-06,
+      "loss": 0.95,
+      "step": 41298
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0082685714285717e-06,
+      "loss": 1.0692,
+      "step": 41299
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0080000000000003e-06,
+      "loss": 0.9251,
+      "step": 41300
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0077314285714284e-06,
+      "loss": 0.9766,
+      "step": 41301
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.007462857142857e-06,
+      "loss": 1.7102,
+      "step": 41302
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.007194285714286e-06,
+      "loss": 0.9725,
+      "step": 41303
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0069257142857147e-06,
+      "loss": 0.8819,
+      "step": 41304
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.006657142857143e-06,
+      "loss": 0.9052,
+      "step": 41305
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0063885714285714e-06,
+      "loss": 0.6932,
+      "step": 41306
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.00612e-06,
+      "loss": 1.3121,
+      "step": 41307
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0058514285714286e-06,
+      "loss": 1.1007,
+      "step": 41308
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0055828571428572e-06,
+      "loss": 1.204,
+      "step": 41309
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.005314285714286e-06,
+      "loss": 1.3102,
+      "step": 41310
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0050457142857144e-06,
+      "loss": 1.1065,
+      "step": 41311
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.004777142857143e-06,
+      "loss": 1.2866,
+      "step": 41312
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0045085714285716e-06,
+      "loss": 1.4617,
+      "step": 41313
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.00424e-06,
+      "loss": 1.0705,
+      "step": 41314
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0039714285714284e-06,
+      "loss": 1.6987,
+      "step": 41315
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0037028571428574e-06,
+      "loss": 1.4885,
+      "step": 41316
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.003434285714286e-06,
+      "loss": 0.8674,
+      "step": 41317
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0031657142857146e-06,
+      "loss": 1.0807,
+      "step": 41318
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.002897142857143e-06,
+      "loss": 0.7759,
+      "step": 41319
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0026285714285714e-06,
+      "loss": 0.5843,
+      "step": 41320
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.00236e-06,
+      "loss": 1.2194,
+      "step": 41321
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0020914285714286e-06,
+      "loss": 1.0277,
+      "step": 41322
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.001822857142857e-06,
+      "loss": 0.679,
+      "step": 41323
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.001554285714286e-06,
+      "loss": 1.3322,
+      "step": 41324
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0012857142857144e-06,
+      "loss": 1.3131,
+      "step": 41325
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.001017142857143e-06,
+      "loss": 1.2004,
+      "step": 41326
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0007485714285716e-06,
+      "loss": 1.1415,
+      "step": 41327
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.0004799999999998e-06,
+      "loss": 0.5758,
+      "step": 41328
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.000211428571429e-06,
+      "loss": 1.2562,
+      "step": 41329
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9999428571428574e-06,
+      "loss": 1.046,
+      "step": 41330
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.999674285714286e-06,
+      "loss": 1.1057,
+      "step": 41331
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9994057142857146e-06,
+      "loss": 0.5687,
+      "step": 41332
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9991371428571428e-06,
+      "loss": 1.2648,
+      "step": 41333
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9988685714285714e-06,
+      "loss": 0.8043,
+      "step": 41334
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9986e-06,
+      "loss": 0.7409,
+      "step": 41335
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.998331428571429e-06,
+      "loss": 1.2276,
+      "step": 41336
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.998062857142857e-06,
+      "loss": 1.2313,
+      "step": 41337
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9977942857142858e-06,
+      "loss": 1.2763,
+      "step": 41338
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9975257142857144e-06,
+      "loss": 0.9754,
+      "step": 41339
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.997257142857143e-06,
+      "loss": 1.1624,
+      "step": 41340
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9969885714285716e-06,
+      "loss": 1.0985,
+      "step": 41341
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.99672e-06,
+      "loss": 0.9236,
+      "step": 41342
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9964514285714288e-06,
+      "loss": 1.0006,
+      "step": 41343
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9961828571428574e-06,
+      "loss": 0.894,
+      "step": 41344
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.995914285714286e-06,
+      "loss": 1.3149,
+      "step": 41345
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.995645714285714e-06,
+      "loss": 1.0817,
+      "step": 41346
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9953771428571427e-06,
+      "loss": 1.4599,
+      "step": 41347
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9951085714285713e-06,
+      "loss": 1.2341,
+      "step": 41348
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9948400000000004e-06,
+      "loss": 1.0011,
+      "step": 41349
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.994571428571429e-06,
+      "loss": 0.6995,
+      "step": 41350
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.994302857142857e-06,
+      "loss": 1.3123,
+      "step": 41351
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9940342857142857e-06,
+      "loss": 0.9805,
+      "step": 41352
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9937657142857143e-06,
+      "loss": 1.2914,
+      "step": 41353
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.993497142857143e-06,
+      "loss": 1.0358,
+      "step": 41354
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9932285714285715e-06,
+      "loss": 1.329,
+      "step": 41355
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.99296e-06,
+      "loss": 0.9341,
+      "step": 41356
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9926914285714287e-06,
+      "loss": 0.7219,
+      "step": 41357
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9924228571428573e-06,
+      "loss": 0.9252,
+      "step": 41358
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.992154285714286e-06,
+      "loss": 1.0407,
+      "step": 41359
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.991885714285714e-06,
+      "loss": 0.7974,
+      "step": 41360
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9916171428571427e-06,
+      "loss": 1.2507,
+      "step": 41361
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9913485714285717e-06,
+      "loss": 1.3572,
+      "step": 41362
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9910800000000003e-06,
+      "loss": 1.0617,
+      "step": 41363
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.990811428571429e-06,
+      "loss": 1.4098,
+      "step": 41364
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.990542857142857e-06,
+      "loss": 1.1211,
+      "step": 41365
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9902742857142857e-06,
+      "loss": 0.7426,
+      "step": 41366
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9900057142857143e-06,
+      "loss": 1.1174,
+      "step": 41367
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.989737142857143e-06,
+      "loss": 1.3262,
+      "step": 41368
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9894685714285715e-06,
+      "loss": 1.5359,
+      "step": 41369
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9892e-06,
+      "loss": 1.1348,
+      "step": 41370
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9889314285714287e-06,
+      "loss": 0.9628,
+      "step": 41371
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9886628571428573e-06,
+      "loss": 1.0173,
+      "step": 41372
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.988394285714286e-06,
+      "loss": 1.0951,
+      "step": 41373
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.988125714285714e-06,
+      "loss": 1.0644,
+      "step": 41374
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.987857142857143e-06,
+      "loss": 0.7878,
+      "step": 41375
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9875885714285717e-06,
+      "loss": 1.3752,
+      "step": 41376
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9873200000000003e-06,
+      "loss": 1.3558,
+      "step": 41377
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9870514285714285e-06,
+      "loss": 1.0836,
+      "step": 41378
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.986782857142857e-06,
+      "loss": 1.3012,
+      "step": 41379
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9865142857142857e-06,
+      "loss": 1.2209,
+      "step": 41380
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9862457142857143e-06,
+      "loss": 1.1228,
+      "step": 41381
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9859771428571433e-06,
+      "loss": 1.3005,
+      "step": 41382
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9857085714285715e-06,
+      "loss": 0.9586,
+      "step": 41383
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.98544e-06,
+      "loss": 0.8952,
+      "step": 41384
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9851714285714287e-06,
+      "loss": 0.7464,
+      "step": 41385
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9849028571428573e-06,
+      "loss": 1.2341,
+      "step": 41386
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.984634285714286e-06,
+      "loss": 0.9933,
+      "step": 41387
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9843657142857145e-06,
+      "loss": 0.8211,
+      "step": 41388
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.984097142857143e-06,
+      "loss": 1.2385,
+      "step": 41389
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9838285714285717e-06,
+      "loss": 1.0926,
+      "step": 41390
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9835600000000003e-06,
+      "loss": 1.1487,
+      "step": 41391
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9832914285714285e-06,
+      "loss": 0.9102,
+      "step": 41392
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.983022857142857e-06,
+      "loss": 1.2584,
+      "step": 41393
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9827542857142857e-06,
+      "loss": 0.9229,
+      "step": 41394
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9824857142857147e-06,
+      "loss": 1.1228,
+      "step": 41395
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9822171428571433e-06,
+      "loss": 1.0706,
+      "step": 41396
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9819485714285715e-06,
+      "loss": 0.9661,
+      "step": 41397
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.98168e-06,
+      "loss": 1.199,
+      "step": 41398
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9814114285714287e-06,
+      "loss": 1.0458,
+      "step": 41399
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9811428571428573e-06,
+      "loss": 0.7488,
+      "step": 41400
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.980874285714286e-06,
+      "loss": 1.1709,
+      "step": 41401
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9806057142857145e-06,
+      "loss": 1.578,
+      "step": 41402
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.980337142857143e-06,
+      "loss": 0.8022,
+      "step": 41403
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9800685714285717e-06,
+      "loss": 0.9756,
+      "step": 41404
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9798000000000003e-06,
+      "loss": 1.1432,
+      "step": 41405
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9795314285714284e-06,
+      "loss": 1.0664,
+      "step": 41406
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.979262857142857e-06,
+      "loss": 0.8689,
+      "step": 41407
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.978994285714286e-06,
+      "loss": 1.2207,
+      "step": 41408
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9787257142857147e-06,
+      "loss": 1.0665,
+      "step": 41409
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.978457142857143e-06,
+      "loss": 1.558,
+      "step": 41410
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9781885714285714e-06,
+      "loss": 1.0162,
+      "step": 41411
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.97792e-06,
+      "loss": 1.2696,
+      "step": 41412
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9776514285714286e-06,
+      "loss": 1.2176,
+      "step": 41413
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9773828571428572e-06,
+      "loss": 1.0332,
+      "step": 41414
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.977114285714286e-06,
+      "loss": 1.2684,
+      "step": 41415
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9768457142857144e-06,
+      "loss": 1.3664,
+      "step": 41416
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.976577142857143e-06,
+      "loss": 1.1226,
+      "step": 41417
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9763085714285716e-06,
+      "loss": 1.129,
+      "step": 41418
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9760400000000002e-06,
+      "loss": 1.2614,
+      "step": 41419
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9757714285714284e-06,
+      "loss": 0.8962,
+      "step": 41420
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9755028571428574e-06,
+      "loss": 0.9489,
+      "step": 41421
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.975234285714286e-06,
+      "loss": 1.336,
+      "step": 41422
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9749657142857146e-06,
+      "loss": 1.0251,
+      "step": 41423
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.974697142857143e-06,
+      "loss": 1.4227,
+      "step": 41424
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9744285714285714e-06,
+      "loss": 1.2437,
+      "step": 41425
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.97416e-06,
+      "loss": 0.9681,
+      "step": 41426
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9738914285714286e-06,
+      "loss": 0.557,
+      "step": 41427
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9736228571428576e-06,
+      "loss": 1.1332,
+      "step": 41428
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.973354285714286e-06,
+      "loss": 1.4516,
+      "step": 41429
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9730857142857144e-06,
+      "loss": 1.3868,
+      "step": 41430
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.972817142857143e-06,
+      "loss": 1.317,
+      "step": 41431
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9725485714285716e-06,
+      "loss": 0.9326,
+      "step": 41432
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9722799999999998e-06,
+      "loss": 1.0953,
+      "step": 41433
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.972011428571429e-06,
+      "loss": 1.5495,
+      "step": 41434
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9717428571428574e-06,
+      "loss": 1.6279,
+      "step": 41435
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.971474285714286e-06,
+      "loss": 1.2259,
+      "step": 41436
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9712057142857146e-06,
+      "loss": 1.0826,
+      "step": 41437
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9709371428571428e-06,
+      "loss": 1.2264,
+      "step": 41438
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9706685714285714e-06,
+      "loss": 1.159,
+      "step": 41439
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9704e-06,
+      "loss": 1.2754,
+      "step": 41440
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.970131428571429e-06,
+      "loss": 1.6613,
+      "step": 41441
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.969862857142857e-06,
+      "loss": 1.5354,
+      "step": 41442
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9695942857142858e-06,
+      "loss": 1.0256,
+      "step": 41443
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9693257142857144e-06,
+      "loss": 1.0937,
+      "step": 41444
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.969057142857143e-06,
+      "loss": 1.2442,
+      "step": 41445
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9687885714285716e-06,
+      "loss": 0.9107,
+      "step": 41446
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.96852e-06,
+      "loss": 1.1314,
+      "step": 41447
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9682514285714288e-06,
+      "loss": 1.7619,
+      "step": 41448
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9679828571428574e-06,
+      "loss": 0.9973,
+      "step": 41449
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.967714285714286e-06,
+      "loss": 1.0973,
+      "step": 41450
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9674457142857146e-06,
+      "loss": 1.0017,
+      "step": 41451
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9671771428571427e-06,
+      "loss": 1.3756,
+      "step": 41452
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9669085714285713e-06,
+      "loss": 1.3614,
+      "step": 41453
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9666400000000004e-06,
+      "loss": 1.0675,
+      "step": 41454
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.966371428571429e-06,
+      "loss": 0.9203,
+      "step": 41455
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.966102857142857e-06,
+      "loss": 1.0764,
+      "step": 41456
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9658342857142857e-06,
+      "loss": 1.6358,
+      "step": 41457
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9655657142857143e-06,
+      "loss": 1.0233,
+      "step": 41458
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.965297142857143e-06,
+      "loss": 1.2278,
+      "step": 41459
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9650285714285715e-06,
+      "loss": 0.9598,
+      "step": 41460
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.96476e-06,
+      "loss": 1.0282,
+      "step": 41461
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9644914285714287e-06,
+      "loss": 0.8925,
+      "step": 41462
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9642228571428573e-06,
+      "loss": 0.9651,
+      "step": 41463
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.963954285714286e-06,
+      "loss": 1.3314,
+      "step": 41464
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.963685714285714e-06,
+      "loss": 0.8604,
+      "step": 41465
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9634171428571427e-06,
+      "loss": 1.474,
+      "step": 41466
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9631485714285717e-06,
+      "loss": 1.189,
+      "step": 41467
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9628800000000003e-06,
+      "loss": 1.1658,
+      "step": 41468
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.962611428571429e-06,
+      "loss": 1.2225,
+      "step": 41469
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.962342857142857e-06,
+      "loss": 1.026,
+      "step": 41470
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9620742857142857e-06,
+      "loss": 0.9934,
+      "step": 41471
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9618057142857143e-06,
+      "loss": 1.2291,
+      "step": 41472
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.961537142857143e-06,
+      "loss": 1.1129,
+      "step": 41473
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9612685714285715e-06,
+      "loss": 0.8626,
+      "step": 41474
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.961e-06,
+      "loss": 1.5049,
+      "step": 41475
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9607314285714287e-06,
+      "loss": 1.0929,
+      "step": 41476
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9604628571428573e-06,
+      "loss": 0.9916,
+      "step": 41477
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.960194285714286e-06,
+      "loss": 0.9504,
+      "step": 41478
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.959925714285714e-06,
+      "loss": 1.3249,
+      "step": 41479
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9596571428571427e-06,
+      "loss": 1.2102,
+      "step": 41480
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9593885714285717e-06,
+      "loss": 0.9144,
+      "step": 41481
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9591200000000003e-06,
+      "loss": 1.2551,
+      "step": 41482
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.958851428571429e-06,
+      "loss": 1.1522,
+      "step": 41483
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.958582857142857e-06,
+      "loss": 1.376,
+      "step": 41484
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9583142857142857e-06,
+      "loss": 1.2341,
+      "step": 41485
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9580457142857143e-06,
+      "loss": 1.4332,
+      "step": 41486
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9577771428571433e-06,
+      "loss": 1.0676,
+      "step": 41487
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9575085714285715e-06,
+      "loss": 1.1332,
+      "step": 41488
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.95724e-06,
+      "loss": 1.0931,
+      "step": 41489
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9569714285714287e-06,
+      "loss": 1.0349,
+      "step": 41490
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9567028571428573e-06,
+      "loss": 1.2896,
+      "step": 41491
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.956434285714286e-06,
+      "loss": 1.5524,
+      "step": 41492
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.956165714285714e-06,
+      "loss": 1.3796,
+      "step": 41493
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.955897142857143e-06,
+      "loss": 0.916,
+      "step": 41494
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9556285714285717e-06,
+      "loss": 1.2052,
+      "step": 41495
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9553600000000003e-06,
+      "loss": 1.2301,
+      "step": 41496
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9550914285714285e-06,
+      "loss": 0.9629,
+      "step": 41497
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.954822857142857e-06,
+      "loss": 1.0774,
+      "step": 41498
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9545542857142857e-06,
+      "loss": 1.1377,
+      "step": 41499
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9542857142857147e-06,
+      "loss": 1.0075,
+      "step": 41500
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9540171428571433e-06,
+      "loss": 0.8499,
+      "step": 41501
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9537485714285715e-06,
+      "loss": 1.4799,
+      "step": 41502
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.95348e-06,
+      "loss": 0.6979,
+      "step": 41503
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9532114285714287e-06,
+      "loss": 1.3476,
+      "step": 41504
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9529428571428573e-06,
+      "loss": 1.8026,
+      "step": 41505
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.952674285714286e-06,
+      "loss": 1.3991,
+      "step": 41506
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9524057142857145e-06,
+      "loss": 1.0037,
+      "step": 41507
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.952137142857143e-06,
+      "loss": 1.0536,
+      "step": 41508
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9518685714285717e-06,
+      "loss": 0.8264,
+      "step": 41509
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9516000000000003e-06,
+      "loss": 1.0114,
+      "step": 41510
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9513314285714284e-06,
+      "loss": 0.7969,
+      "step": 41511
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.951062857142857e-06,
+      "loss": 1.2091,
+      "step": 41512
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9507942857142856e-06,
+      "loss": 1.3942,
+      "step": 41513
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9505257142857147e-06,
+      "loss": 1.2637,
+      "step": 41514
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9502571428571432e-06,
+      "loss": 1.1888,
+      "step": 41515
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9499885714285714e-06,
+      "loss": 1.477,
+      "step": 41516
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.94972e-06,
+      "loss": 1.1254,
+      "step": 41517
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9494514285714286e-06,
+      "loss": 1.143,
+      "step": 41518
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9491828571428572e-06,
+      "loss": 1.1485,
+      "step": 41519
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.948914285714286e-06,
+      "loss": 1.1629,
+      "step": 41520
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9486457142857144e-06,
+      "loss": 0.9942,
+      "step": 41521
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.948377142857143e-06,
+      "loss": 1.1169,
+      "step": 41522
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9481085714285716e-06,
+      "loss": 1.1354,
+      "step": 41523
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9478400000000002e-06,
+      "loss": 0.8022,
+      "step": 41524
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9475714285714284e-06,
+      "loss": 0.876,
+      "step": 41525
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.947302857142857e-06,
+      "loss": 1.3002,
+      "step": 41526
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.947034285714286e-06,
+      "loss": 0.9295,
+      "step": 41527
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9467657142857146e-06,
+      "loss": 1.1405,
+      "step": 41528
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.946497142857143e-06,
+      "loss": 1.4294,
+      "step": 41529
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9462285714285714e-06,
+      "loss": 1.224,
+      "step": 41530
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.94596e-06,
+      "loss": 0.7211,
+      "step": 41531
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9456914285714286e-06,
+      "loss": 0.8634,
+      "step": 41532
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.945422857142857e-06,
+      "loss": 1.1759,
+      "step": 41533
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.945154285714286e-06,
+      "loss": 1.1144,
+      "step": 41534
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9448857142857144e-06,
+      "loss": 0.8938,
+      "step": 41535
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.944617142857143e-06,
+      "loss": 1.3007,
+      "step": 41536
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9443485714285716e-06,
+      "loss": 1.1346,
+      "step": 41537
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.94408e-06,
+      "loss": 1.5967,
+      "step": 41538
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9438114285714284e-06,
+      "loss": 1.0395,
+      "step": 41539
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9435428571428574e-06,
+      "loss": 1.4343,
+      "step": 41540
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.943274285714286e-06,
+      "loss": 1.2722,
+      "step": 41541
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9430057142857146e-06,
+      "loss": 1.1767,
+      "step": 41542
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9427371428571428e-06,
+      "loss": 0.971,
+      "step": 41543
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9424685714285714e-06,
+      "loss": 1.1064,
+      "step": 41544
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9422e-06,
+      "loss": 1.1402,
+      "step": 41545
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9419314285714286e-06,
+      "loss": 1.1176,
+      "step": 41546
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9416628571428576e-06,
+      "loss": 1.0823,
+      "step": 41547
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9413942857142858e-06,
+      "loss": 1.2128,
+      "step": 41548
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9411257142857144e-06,
+      "loss": 0.9331,
+      "step": 41549
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.940857142857143e-06,
+      "loss": 1.4951,
+      "step": 41550
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9405885714285716e-06,
+      "loss": 1.2372,
+      "step": 41551
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9403199999999997e-06,
+      "loss": 1.0286,
+      "step": 41552
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9400514285714288e-06,
+      "loss": 1.1237,
+      "step": 41553
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9397828571428574e-06,
+      "loss": 0.8931,
+      "step": 41554
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.939514285714286e-06,
+      "loss": 1.3464,
+      "step": 41555
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9392457142857146e-06,
+      "loss": 1.2351,
+      "step": 41556
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9389771428571427e-06,
+      "loss": 1.1359,
+      "step": 41557
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9387085714285713e-06,
+      "loss": 1.2301,
+      "step": 41558
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.93844e-06,
+      "loss": 1.2447,
+      "step": 41559
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.938171428571429e-06,
+      "loss": 1.4833,
+      "step": 41560
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.937902857142857e-06,
+      "loss": 0.9991,
+      "step": 41561
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.9376342857142857e-06,
+      "loss": 1.3435,
+      "step": 41562
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9373657142857143e-06,
+      "loss": 0.8538,
+      "step": 41563
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.937097142857143e-06,
+      "loss": 1.251,
+      "step": 41564
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9368285714285715e-06,
+      "loss": 1.0319,
+      "step": 41565
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.93656e-06,
+      "loss": 1.04,
+      "step": 41566
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9362914285714287e-06,
+      "loss": 0.8388,
+      "step": 41567
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9360228571428573e-06,
+      "loss": 1.1877,
+      "step": 41568
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.935754285714286e-06,
+      "loss": 0.8562,
+      "step": 41569
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9354857142857145e-06,
+      "loss": 0.9321,
+      "step": 41570
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9352171428571427e-06,
+      "loss": 1.4695,
+      "step": 41571
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9349485714285713e-06,
+      "loss": 0.7377,
+      "step": 41572
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9346800000000003e-06,
+      "loss": 0.8798,
+      "step": 41573
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.934411428571429e-06,
+      "loss": 1.1472,
+      "step": 41574
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.934142857142857e-06,
+      "loss": 0.8627,
+      "step": 41575
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9338742857142857e-06,
+      "loss": 0.8958,
+      "step": 41576
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9336057142857143e-06,
+      "loss": 0.8703,
+      "step": 41577
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.933337142857143e-06,
+      "loss": 0.7345,
+      "step": 41578
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9330685714285715e-06,
+      "loss": 0.7867,
+      "step": 41579
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9328e-06,
+      "loss": 1.1265,
+      "step": 41580
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9325314285714287e-06,
+      "loss": 1.2309,
+      "step": 41581
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9322628571428573e-06,
+      "loss": 0.9531,
+      "step": 41582
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.931994285714286e-06,
+      "loss": 1.1741,
+      "step": 41583
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.931725714285714e-06,
+      "loss": 1.3046,
+      "step": 41584
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9314571428571427e-06,
+      "loss": 1.382,
+      "step": 41585
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9311885714285717e-06,
+      "loss": 1.1282,
+      "step": 41586
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9309200000000003e-06,
+      "loss": 1.2213,
+      "step": 41587
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.930651428571429e-06,
+      "loss": 1.1816,
+      "step": 41588
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.930382857142857e-06,
+      "loss": 1.0036,
+      "step": 41589
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9301142857142857e-06,
+      "loss": 0.8577,
+      "step": 41590
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9298457142857143e-06,
+      "loss": 1.1122,
+      "step": 41591
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.929577142857143e-06,
+      "loss": 1.0311,
+      "step": 41592
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.929308571428572e-06,
+      "loss": 0.8426,
+      "step": 41593
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.92904e-06,
+      "loss": 1.0809,
+      "step": 41594
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9287714285714287e-06,
+      "loss": 0.9831,
+      "step": 41595
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9285028571428573e-06,
+      "loss": 0.8847,
+      "step": 41596
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.928234285714286e-06,
+      "loss": 1.0441,
+      "step": 41597
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.927965714285714e-06,
+      "loss": 1.0194,
+      "step": 41598
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.927697142857143e-06,
+      "loss": 1.0527,
+      "step": 41599
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9274285714285717e-06,
+      "loss": 1.1572,
+      "step": 41600
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9271600000000003e-06,
+      "loss": 1.1398,
+      "step": 41601
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.926891428571429e-06,
+      "loss": 1.3297,
+      "step": 41602
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.926622857142857e-06,
+      "loss": 1.2931,
+      "step": 41603
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9263542857142857e-06,
+      "loss": 1.2198,
+      "step": 41604
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9260857142857143e-06,
+      "loss": 0.9843,
+      "step": 41605
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9258171428571433e-06,
+      "loss": 1.4479,
+      "step": 41606
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9255485714285715e-06,
+      "loss": 1.1437,
+      "step": 41607
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.92528e-06,
+      "loss": 0.8127,
+      "step": 41608
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9250114285714287e-06,
+      "loss": 1.0828,
+      "step": 41609
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9247428571428573e-06,
+      "loss": 1.2099,
+      "step": 41610
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.924474285714286e-06,
+      "loss": 1.4826,
+      "step": 41611
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9242057142857145e-06,
+      "loss": 0.9279,
+      "step": 41612
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.923937142857143e-06,
+      "loss": 0.8671,
+      "step": 41613
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9236685714285717e-06,
+      "loss": 1.2498,
+      "step": 41614
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9234000000000003e-06,
+      "loss": 0.912,
+      "step": 41615
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9231314285714284e-06,
+      "loss": 1.0135,
+      "step": 41616
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.922862857142857e-06,
+      "loss": 0.8976,
+      "step": 41617
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9225942857142856e-06,
+      "loss": 1.2735,
+      "step": 41618
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9223257142857146e-06,
+      "loss": 0.8481,
+      "step": 41619
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9220571428571432e-06,
+      "loss": 1.0109,
+      "step": 41620
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9217885714285714e-06,
+      "loss": 1.1385,
+      "step": 41621
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.92152e-06,
+      "loss": 0.9837,
+      "step": 41622
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9212514285714286e-06,
+      "loss": 1.2901,
+      "step": 41623
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9209828571428572e-06,
+      "loss": 1.258,
+      "step": 41624
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.920714285714286e-06,
+      "loss": 1.1881,
+      "step": 41625
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9204457142857144e-06,
+      "loss": 0.8622,
+      "step": 41626
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.920177142857143e-06,
+      "loss": 1.4534,
+      "step": 41627
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9199085714285716e-06,
+      "loss": 0.9285,
+      "step": 41628
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9196400000000002e-06,
+      "loss": 1.1522,
+      "step": 41629
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9193714285714284e-06,
+      "loss": 1.171,
+      "step": 41630
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.919102857142857e-06,
+      "loss": 1.0631,
+      "step": 41631
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.918834285714286e-06,
+      "loss": 0.6159,
+      "step": 41632
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9185657142857146e-06,
+      "loss": 1.123,
+      "step": 41633
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9182971428571432e-06,
+      "loss": 1.0696,
+      "step": 41634
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9180285714285714e-06,
+      "loss": 0.6314,
+      "step": 41635
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.91776e-06,
+      "loss": 1.3667,
+      "step": 41636
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9174914285714286e-06,
+      "loss": 1.3034,
+      "step": 41637
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.917222857142857e-06,
+      "loss": 0.7581,
+      "step": 41638
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.916954285714286e-06,
+      "loss": 1.0852,
+      "step": 41639
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9166857142857144e-06,
+      "loss": 0.6191,
+      "step": 41640
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.916417142857143e-06,
+      "loss": 1.0846,
+      "step": 41641
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9161485714285716e-06,
+      "loss": 1.253,
+      "step": 41642
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.91588e-06,
+      "loss": 1.2419,
+      "step": 41643
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9156114285714284e-06,
+      "loss": 1.2898,
+      "step": 41644
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9153428571428574e-06,
+      "loss": 1.215,
+      "step": 41645
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.915074285714286e-06,
+      "loss": 0.977,
+      "step": 41646
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9148057142857146e-06,
+      "loss": 1.2842,
+      "step": 41647
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9145371428571428e-06,
+      "loss": 0.9942,
+      "step": 41648
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9142685714285714e-06,
+      "loss": 0.8714,
+      "step": 41649
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.914e-06,
+      "loss": 0.9187,
+      "step": 41650
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9137314285714286e-06,
+      "loss": 0.9499,
+      "step": 41651
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9134628571428576e-06,
+      "loss": 1.353,
+      "step": 41652
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9131942857142858e-06,
+      "loss": 1.5088,
+      "step": 41653
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9129257142857144e-06,
+      "loss": 0.8406,
+      "step": 41654
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.912657142857143e-06,
+      "loss": 1.0066,
+      "step": 41655
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9123885714285716e-06,
+      "loss": 0.8911,
+      "step": 41656
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.91212e-06,
+      "loss": 1.4392,
+      "step": 41657
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9118514285714288e-06,
+      "loss": 0.94,
+      "step": 41658
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9115828571428574e-06,
+      "loss": 1.0776,
+      "step": 41659
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.911314285714286e-06,
+      "loss": 1.3632,
+      "step": 41660
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9110457142857146e-06,
+      "loss": 1.1484,
+      "step": 41661
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9107771428571427e-06,
+      "loss": 1.3494,
+      "step": 41662
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9105085714285713e-06,
+      "loss": 0.9968,
+      "step": 41663
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.91024e-06,
+      "loss": 1.1911,
+      "step": 41664
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.909971428571429e-06,
+      "loss": 1.0703,
+      "step": 41665
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9097028571428576e-06,
+      "loss": 1.5415,
+      "step": 41666
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9094342857142857e-06,
+      "loss": 0.7693,
+      "step": 41667
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9091657142857143e-06,
+      "loss": 1.0661,
+      "step": 41668
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.908897142857143e-06,
+      "loss": 1.4441,
+      "step": 41669
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9086285714285715e-06,
+      "loss": 0.897,
+      "step": 41670
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.90836e-06,
+      "loss": 1.0339,
+      "step": 41671
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9080914285714287e-06,
+      "loss": 0.9429,
+      "step": 41672
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9078228571428573e-06,
+      "loss": 0.7464,
+      "step": 41673
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.907554285714286e-06,
+      "loss": 1.2712,
+      "step": 41674
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9072857142857145e-06,
+      "loss": 1.06,
+      "step": 41675
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9070171428571427e-06,
+      "loss": 1.1617,
+      "step": 41676
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9067485714285713e-06,
+      "loss": 0.7477,
+      "step": 41677
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9064800000000003e-06,
+      "loss": 1.1306,
+      "step": 41678
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.906211428571429e-06,
+      "loss": 1.2035,
+      "step": 41679
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9059428571428575e-06,
+      "loss": 1.2617,
+      "step": 41680
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9056742857142857e-06,
+      "loss": 1.1055,
+      "step": 41681
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9054057142857143e-06,
+      "loss": 1.2473,
+      "step": 41682
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.905137142857143e-06,
+      "loss": 1.3464,
+      "step": 41683
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9048685714285715e-06,
+      "loss": 0.6099,
+      "step": 41684
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9046e-06,
+      "loss": 1.1339,
+      "step": 41685
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9043314285714287e-06,
+      "loss": 1.2359,
+      "step": 41686
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9040628571428573e-06,
+      "loss": 1.1424,
+      "step": 41687
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.903794285714286e-06,
+      "loss": 0.7502,
+      "step": 41688
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9035257142857145e-06,
+      "loss": 1.0299,
+      "step": 41689
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9032571428571427e-06,
+      "loss": 0.9758,
+      "step": 41690
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9029885714285717e-06,
+      "loss": 1.1727,
+      "step": 41691
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9027200000000003e-06,
+      "loss": 1.3677,
+      "step": 41692
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.902451428571429e-06,
+      "loss": 1.3504,
+      "step": 41693
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.902182857142857e-06,
+      "loss": 1.1062,
+      "step": 41694
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9019142857142857e-06,
+      "loss": 1.327,
+      "step": 41695
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9016457142857143e-06,
+      "loss": 0.9851,
+      "step": 41696
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.901377142857143e-06,
+      "loss": 0.8718,
+      "step": 41697
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.901108571428572e-06,
+      "loss": 1.0587,
+      "step": 41698
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.90084e-06,
+      "loss": 1.1549,
+      "step": 41699
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9005714285714287e-06,
+      "loss": 0.859,
+      "step": 41700
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.9003028571428573e-06,
+      "loss": 1.0507,
+      "step": 41701
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.900034285714286e-06,
+      "loss": 1.3855,
+      "step": 41702
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.899765714285714e-06,
+      "loss": 0.738,
+      "step": 41703
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8994971428571427e-06,
+      "loss": 1.4476,
+      "step": 41704
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8992285714285717e-06,
+      "loss": 0.8789,
+      "step": 41705
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8989600000000003e-06,
+      "loss": 1.105,
+      "step": 41706
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.898691428571429e-06,
+      "loss": 1.0805,
+      "step": 41707
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.898422857142857e-06,
+      "loss": 1.0422,
+      "step": 41708
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8981542857142857e-06,
+      "loss": 1.5056,
+      "step": 41709
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8978857142857143e-06,
+      "loss": 0.7601,
+      "step": 41710
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8976171428571433e-06,
+      "loss": 1.3675,
+      "step": 41711
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.897348571428572e-06,
+      "loss": 0.838,
+      "step": 41712
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.89708e-06,
+      "loss": 1.4575,
+      "step": 41713
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8968114285714287e-06,
+      "loss": 1.4666,
+      "step": 41714
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8965428571428573e-06,
+      "loss": 0.9402,
+      "step": 41715
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.896274285714286e-06,
+      "loss": 0.9647,
+      "step": 41716
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.896005714285714e-06,
+      "loss": 1.4851,
+      "step": 41717
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.895737142857143e-06,
+      "loss": 1.3099,
+      "step": 41718
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8954685714285716e-06,
+      "loss": 1.0427,
+      "step": 41719
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8952000000000002e-06,
+      "loss": 0.794,
+      "step": 41720
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.894931428571429e-06,
+      "loss": 0.8546,
+      "step": 41721
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.894662857142857e-06,
+      "loss": 1.132,
+      "step": 41722
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8943942857142856e-06,
+      "loss": 1.2285,
+      "step": 41723
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8941257142857146e-06,
+      "loss": 1.2142,
+      "step": 41724
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8938571428571432e-06,
+      "loss": 0.7997,
+      "step": 41725
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8935885714285714e-06,
+      "loss": 1.2229,
+      "step": 41726
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.89332e-06,
+      "loss": 1.2082,
+      "step": 41727
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8930514285714286e-06,
+      "loss": 0.7848,
+      "step": 41728
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8927828571428572e-06,
+      "loss": 1.0835,
+      "step": 41729
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.892514285714286e-06,
+      "loss": 1.4559,
+      "step": 41730
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8922457142857144e-06,
+      "loss": 1.1981,
+      "step": 41731
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.891977142857143e-06,
+      "loss": 0.6079,
+      "step": 41732
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8917085714285716e-06,
+      "loss": 0.9627,
+      "step": 41733
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8914400000000002e-06,
+      "loss": 1.3731,
+      "step": 41734
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8911714285714284e-06,
+      "loss": 1.3158,
+      "step": 41735
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.890902857142857e-06,
+      "loss": 1.5535,
+      "step": 41736
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8906342857142856e-06,
+      "loss": 1.0947,
+      "step": 41737
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8903657142857146e-06,
+      "loss": 0.8758,
+      "step": 41738
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8900971428571432e-06,
+      "loss": 1.1097,
+      "step": 41739
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8898285714285714e-06,
+      "loss": 1.1892,
+      "step": 41740
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.88956e-06,
+      "loss": 0.8842,
+      "step": 41741
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8892914285714286e-06,
+      "loss": 1.4608,
+      "step": 41742
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.889022857142857e-06,
+      "loss": 1.7127,
+      "step": 41743
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8887542857142862e-06,
+      "loss": 0.6973,
+      "step": 41744
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8884857142857144e-06,
+      "loss": 0.982,
+      "step": 41745
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.888217142857143e-06,
+      "loss": 1.0061,
+      "step": 41746
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8879485714285716e-06,
+      "loss": 1.3308,
+      "step": 41747
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.88768e-06,
+      "loss": 1.578,
+      "step": 41748
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8874114285714284e-06,
+      "loss": 0.8307,
+      "step": 41749
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.887142857142857e-06,
+      "loss": 0.9746,
+      "step": 41750
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.886874285714286e-06,
+      "loss": 0.9578,
+      "step": 41751
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8866057142857146e-06,
+      "loss": 1.0339,
+      "step": 41752
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.886337142857143e-06,
+      "loss": 1.2492,
+      "step": 41753
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8860685714285714e-06,
+      "loss": 1.3416,
+      "step": 41754
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8858e-06,
+      "loss": 1.1794,
+      "step": 41755
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8855314285714286e-06,
+      "loss": 1.1614,
+      "step": 41756
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.885262857142857e-06,
+      "loss": 1.1677,
+      "step": 41757
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8849942857142858e-06,
+      "loss": 1.2334,
+      "step": 41758
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8847257142857144e-06,
+      "loss": 0.901,
+      "step": 41759
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.884457142857143e-06,
+      "loss": 1.3107,
+      "step": 41760
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8841885714285716e-06,
+      "loss": 1.1099,
+      "step": 41761
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.88392e-06,
+      "loss": 1.2305,
+      "step": 41762
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8836514285714283e-06,
+      "loss": 0.8356,
+      "step": 41763
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8833828571428574e-06,
+      "loss": 0.9092,
+      "step": 41764
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.883114285714286e-06,
+      "loss": 0.9418,
+      "step": 41765
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8828457142857146e-06,
+      "loss": 1.005,
+      "step": 41766
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8825771428571427e-06,
+      "loss": 0.9838,
+      "step": 41767
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8823085714285713e-06,
+      "loss": 1.4059,
+      "step": 41768
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.88204e-06,
+      "loss": 1.3951,
+      "step": 41769
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8817714285714285e-06,
+      "loss": 0.8457,
+      "step": 41770
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8815028571428576e-06,
+      "loss": 0.8583,
+      "step": 41771
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8812342857142857e-06,
+      "loss": 1.0767,
+      "step": 41772
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8809657142857143e-06,
+      "loss": 1.1969,
+      "step": 41773
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.880697142857143e-06,
+      "loss": 1.02,
+      "step": 41774
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8804285714285715e-06,
+      "loss": 1.4253,
+      "step": 41775
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.88016e-06,
+      "loss": 1.3142,
+      "step": 41776
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8798914285714287e-06,
+      "loss": 1.2845,
+      "step": 41777
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8796228571428573e-06,
+      "loss": 1.1319,
+      "step": 41778
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.879354285714286e-06,
+      "loss": 0.9226,
+      "step": 41779
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8790857142857145e-06,
+      "loss": 1.0758,
+      "step": 41780
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8788171428571427e-06,
+      "loss": 1.1412,
+      "step": 41781
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8785485714285713e-06,
+      "loss": 0.9832,
+      "step": 41782
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.87828e-06,
+      "loss": 1.287,
+      "step": 41783
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.878011428571429e-06,
+      "loss": 1.6767,
+      "step": 41784
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8777428571428575e-06,
+      "loss": 1.2104,
+      "step": 41785
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8774742857142857e-06,
+      "loss": 0.9115,
+      "step": 41786
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8772057142857143e-06,
+      "loss": 0.9357,
+      "step": 41787
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.876937142857143e-06,
+      "loss": 1.0326,
+      "step": 41788
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8766685714285715e-06,
+      "loss": 1.1244,
+      "step": 41789
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8764e-06,
+      "loss": 1.2809,
+      "step": 41790
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8761314285714287e-06,
+      "loss": 1.6431,
+      "step": 41791
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8758628571428573e-06,
+      "loss": 1.3266,
+      "step": 41792
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.875594285714286e-06,
+      "loss": 1.1538,
+      "step": 41793
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8753257142857145e-06,
+      "loss": 1.3211,
+      "step": 41794
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8750571428571427e-06,
+      "loss": 1.0337,
+      "step": 41795
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8747885714285713e-06,
+      "loss": 0.8637,
+      "step": 41796
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8745200000000003e-06,
+      "loss": 1.0829,
+      "step": 41797
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.874251428571429e-06,
+      "loss": 1.3575,
+      "step": 41798
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8739828571428575e-06,
+      "loss": 0.7465,
+      "step": 41799
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8737142857142857e-06,
+      "loss": 0.9603,
+      "step": 41800
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8734457142857143e-06,
+      "loss": 1.1343,
+      "step": 41801
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.873177142857143e-06,
+      "loss": 1.1388,
+      "step": 41802
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8729085714285715e-06,
+      "loss": 1.2479,
+      "step": 41803
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.87264e-06,
+      "loss": 1.3456,
+      "step": 41804
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8723714285714287e-06,
+      "loss": 0.898,
+      "step": 41805
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8721028571428573e-06,
+      "loss": 0.7752,
+      "step": 41806
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.871834285714286e-06,
+      "loss": 1.2033,
+      "step": 41807
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8715657142857145e-06,
+      "loss": 1.0905,
+      "step": 41808
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8712971428571427e-06,
+      "loss": 1.0904,
+      "step": 41809
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8710285714285717e-06,
+      "loss": 1.1164,
+      "step": 41810
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8707600000000003e-06,
+      "loss": 0.8621,
+      "step": 41811
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.870491428571429e-06,
+      "loss": 1.0893,
+      "step": 41812
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.870222857142857e-06,
+      "loss": 1.1705,
+      "step": 41813
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8699542857142857e-06,
+      "loss": 1.1351,
+      "step": 41814
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8696857142857143e-06,
+      "loss": 0.8706,
+      "step": 41815
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.869417142857143e-06,
+      "loss": 1.1021,
+      "step": 41816
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.869148571428572e-06,
+      "loss": 1.3473,
+      "step": 41817
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.86888e-06,
+      "loss": 0.7818,
+      "step": 41818
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8686114285714287e-06,
+      "loss": 1.1654,
+      "step": 41819
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8683428571428572e-06,
+      "loss": 1.2333,
+      "step": 41820
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.868074285714286e-06,
+      "loss": 1.0638,
+      "step": 41821
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.867805714285714e-06,
+      "loss": 1.0509,
+      "step": 41822
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.867537142857143e-06,
+      "loss": 1.2939,
+      "step": 41823
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8672685714285716e-06,
+      "loss": 1.2468,
+      "step": 41824
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8670000000000002e-06,
+      "loss": 0.9222,
+      "step": 41825
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.866731428571429e-06,
+      "loss": 0.7814,
+      "step": 41826
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.866462857142857e-06,
+      "loss": 0.7081,
+      "step": 41827
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8661942857142856e-06,
+      "loss": 1.3329,
+      "step": 41828
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8659257142857142e-06,
+      "loss": 1.0619,
+      "step": 41829
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8656571428571432e-06,
+      "loss": 1.4124,
+      "step": 41830
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.865388571428572e-06,
+      "loss": 1.0439,
+      "step": 41831
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.86512e-06,
+      "loss": 1.1566,
+      "step": 41832
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8648514285714286e-06,
+      "loss": 0.8503,
+      "step": 41833
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8645828571428572e-06,
+      "loss": 1.183,
+      "step": 41834
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.864314285714286e-06,
+      "loss": 1.3161,
+      "step": 41835
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8640457142857144e-06,
+      "loss": 1.4732,
+      "step": 41836
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.863777142857143e-06,
+      "loss": 1.2237,
+      "step": 41837
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8635085714285716e-06,
+      "loss": 1.223,
+      "step": 41838
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8632400000000002e-06,
+      "loss": 1.6155,
+      "step": 41839
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.862971428571429e-06,
+      "loss": 1.1235,
+      "step": 41840
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.862702857142857e-06,
+      "loss": 1.3507,
+      "step": 41841
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8624342857142856e-06,
+      "loss": 0.5596,
+      "step": 41842
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8621657142857146e-06,
+      "loss": 1.2458,
+      "step": 41843
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8618971428571432e-06,
+      "loss": 1.195,
+      "step": 41844
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8616285714285714e-06,
+      "loss": 1.1962,
+      "step": 41845
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.86136e-06,
+      "loss": 0.8865,
+      "step": 41846
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8610914285714286e-06,
+      "loss": 1.1998,
+      "step": 41847
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.860822857142857e-06,
+      "loss": 1.3,
+      "step": 41848
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.860554285714286e-06,
+      "loss": 1.4868,
+      "step": 41849
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8602857142857144e-06,
+      "loss": 1.4985,
+      "step": 41850
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.860017142857143e-06,
+      "loss": 1.2455,
+      "step": 41851
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8597485714285716e-06,
+      "loss": 0.9743,
+      "step": 41852
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.85948e-06,
+      "loss": 1.0818,
+      "step": 41853
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8592114285714284e-06,
+      "loss": 1.2178,
+      "step": 41854
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.858942857142857e-06,
+      "loss": 0.9524,
+      "step": 41855
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.858674285714286e-06,
+      "loss": 0.9129,
+      "step": 41856
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8584057142857146e-06,
+      "loss": 1.2754,
+      "step": 41857
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.858137142857143e-06,
+      "loss": 0.8154,
+      "step": 41858
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8578685714285714e-06,
+      "loss": 1.0534,
+      "step": 41859
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8576e-06,
+      "loss": 0.9184,
+      "step": 41860
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8573314285714286e-06,
+      "loss": 1.1899,
+      "step": 41861
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.857062857142857e-06,
+      "loss": 0.8882,
+      "step": 41862
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.856794285714286e-06,
+      "loss": 1.5035,
+      "step": 41863
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8565257142857144e-06,
+      "loss": 1.1969,
+      "step": 41864
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.856257142857143e-06,
+      "loss": 0.7006,
+      "step": 41865
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8559885714285716e-06,
+      "loss": 1.4038,
+      "step": 41866
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.85572e-06,
+      "loss": 1.0357,
+      "step": 41867
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8554514285714283e-06,
+      "loss": 0.963,
+      "step": 41868
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8551828571428574e-06,
+      "loss": 1.0374,
+      "step": 41869
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.854914285714286e-06,
+      "loss": 1.2734,
+      "step": 41870
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8546457142857146e-06,
+      "loss": 1.6378,
+      "step": 41871
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.854377142857143e-06,
+      "loss": 1.4807,
+      "step": 41872
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8541085714285713e-06,
+      "loss": 1.0953,
+      "step": 41873
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.85384e-06,
+      "loss": 0.7583,
+      "step": 41874
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8535714285714285e-06,
+      "loss": 0.9759,
+      "step": 41875
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8533028571428576e-06,
+      "loss": 1.0836,
+      "step": 41876
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8530342857142857e-06,
+      "loss": 1.2537,
+      "step": 41877
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8527657142857143e-06,
+      "loss": 1.2629,
+      "step": 41878
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.852497142857143e-06,
+      "loss": 0.9927,
+      "step": 41879
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8522285714285715e-06,
+      "loss": 0.8009,
+      "step": 41880
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.85196e-06,
+      "loss": 0.9429,
+      "step": 41881
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8516914285714287e-06,
+      "loss": 0.9944,
+      "step": 41882
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8514228571428573e-06,
+      "loss": 0.982,
+      "step": 41883
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.851154285714286e-06,
+      "loss": 1.4288,
+      "step": 41884
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8508857142857145e-06,
+      "loss": 0.6426,
+      "step": 41885
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.850617142857143e-06,
+      "loss": 1.089,
+      "step": 41886
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8503485714285713e-06,
+      "loss": 0.7348,
+      "step": 41887
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.85008e-06,
+      "loss": 0.8426,
+      "step": 41888
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.849811428571429e-06,
+      "loss": 1.5379,
+      "step": 41889
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8495428571428575e-06,
+      "loss": 1.1413,
+      "step": 41890
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8492742857142857e-06,
+      "loss": 1.2612,
+      "step": 41891
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8490057142857143e-06,
+      "loss": 0.9872,
+      "step": 41892
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.848737142857143e-06,
+      "loss": 0.8736,
+      "step": 41893
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8484685714285715e-06,
+      "loss": 1.2598,
+      "step": 41894
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8482e-06,
+      "loss": 1.1558,
+      "step": 41895
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8479314285714287e-06,
+      "loss": 1.3373,
+      "step": 41896
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8476628571428573e-06,
+      "loss": 1.059,
+      "step": 41897
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.847394285714286e-06,
+      "loss": 1.5358,
+      "step": 41898
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8471257142857145e-06,
+      "loss": 1.0037,
+      "step": 41899
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8468571428571427e-06,
+      "loss": 1.1116,
+      "step": 41900
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8465885714285713e-06,
+      "loss": 1.4067,
+      "step": 41901
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8463200000000003e-06,
+      "loss": 0.9784,
+      "step": 41902
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.846051428571429e-06,
+      "loss": 0.8037,
+      "step": 41903
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8457828571428575e-06,
+      "loss": 1.1998,
+      "step": 41904
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8455142857142857e-06,
+      "loss": 1.3087,
+      "step": 41905
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8452457142857143e-06,
+      "loss": 0.8309,
+      "step": 41906
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.844977142857143e-06,
+      "loss": 1.1655,
+      "step": 41907
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8447085714285715e-06,
+      "loss": 1.337,
+      "step": 41908
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.84444e-06,
+      "loss": 0.7824,
+      "step": 41909
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8441714285714287e-06,
+      "loss": 0.9484,
+      "step": 41910
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8439028571428573e-06,
+      "loss": 0.6776,
+      "step": 41911
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.843634285714286e-06,
+      "loss": 1.0331,
+      "step": 41912
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8433657142857145e-06,
+      "loss": 1.2207,
+      "step": 41913
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8430971428571427e-06,
+      "loss": 1.3238,
+      "step": 41914
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8428285714285717e-06,
+      "loss": 1.0267,
+      "step": 41915
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8425600000000003e-06,
+      "loss": 1.0216,
+      "step": 41916
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.842291428571429e-06,
+      "loss": 1.2821,
+      "step": 41917
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8420228571428575e-06,
+      "loss": 1.1,
+      "step": 41918
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8417542857142857e-06,
+      "loss": 1.3095,
+      "step": 41919
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8414857142857143e-06,
+      "loss": 1.0261,
+      "step": 41920
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.841217142857143e-06,
+      "loss": 0.8032,
+      "step": 41921
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.840948571428572e-06,
+      "loss": 0.903,
+      "step": 41922
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.84068e-06,
+      "loss": 0.9168,
+      "step": 41923
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8404114285714286e-06,
+      "loss": 1.0516,
+      "step": 41924
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8401428571428572e-06,
+      "loss": 1.6037,
+      "step": 41925
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.839874285714286e-06,
+      "loss": 1.2599,
+      "step": 41926
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8396057142857144e-06,
+      "loss": 1.0402,
+      "step": 41927
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8393371428571426e-06,
+      "loss": 1.1486,
+      "step": 41928
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8390685714285716e-06,
+      "loss": 1.4153,
+      "step": 41929
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8388000000000002e-06,
+      "loss": 1.3299,
+      "step": 41930
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.838531428571429e-06,
+      "loss": 1.3173,
+      "step": 41931
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.838262857142857e-06,
+      "loss": 1.0119,
+      "step": 41932
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8379942857142856e-06,
+      "loss": 0.9773,
+      "step": 41933
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8377257142857142e-06,
+      "loss": 0.6962,
+      "step": 41934
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8374571428571432e-06,
+      "loss": 1.1791,
+      "step": 41935
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.837188571428572e-06,
+      "loss": 1.4131,
+      "step": 41936
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.83692e-06,
+      "loss": 1.0591,
+      "step": 41937
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8366514285714286e-06,
+      "loss": 1.262,
+      "step": 41938
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8363828571428572e-06,
+      "loss": 0.8701,
+      "step": 41939
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.836114285714286e-06,
+      "loss": 1.2004,
+      "step": 41940
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.835845714285714e-06,
+      "loss": 1.3049,
+      "step": 41941
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.835577142857143e-06,
+      "loss": 1.0724,
+      "step": 41942
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8353085714285716e-06,
+      "loss": 0.8503,
+      "step": 41943
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8350400000000002e-06,
+      "loss": 0.8989,
+      "step": 41944
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.834771428571429e-06,
+      "loss": 1.3355,
+      "step": 41945
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.834502857142857e-06,
+      "loss": 1.4721,
+      "step": 41946
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8342342857142856e-06,
+      "loss": 0.9046,
+      "step": 41947
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8339657142857146e-06,
+      "loss": 0.905,
+      "step": 41948
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8336971428571432e-06,
+      "loss": 0.9115,
+      "step": 41949
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.833428571428572e-06,
+      "loss": 1.2215,
+      "step": 41950
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.83316e-06,
+      "loss": 0.8817,
+      "step": 41951
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8328914285714286e-06,
+      "loss": 0.9987,
+      "step": 41952
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.832622857142857e-06,
+      "loss": 1.167,
+      "step": 41953
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.832354285714286e-06,
+      "loss": 0.9166,
+      "step": 41954
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8320857142857144e-06,
+      "loss": 0.7422,
+      "step": 41955
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.831817142857143e-06,
+      "loss": 1.0203,
+      "step": 41956
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8315485714285716e-06,
+      "loss": 1.201,
+      "step": 41957
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.83128e-06,
+      "loss": 1.2255,
+      "step": 41958
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.831011428571429e-06,
+      "loss": 0.6939,
+      "step": 41959
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.830742857142857e-06,
+      "loss": 0.8505,
+      "step": 41960
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8304742857142856e-06,
+      "loss": 1.444,
+      "step": 41961
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8302057142857146e-06,
+      "loss": 1.3796,
+      "step": 41962
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.829937142857143e-06,
+      "loss": 0.9315,
+      "step": 41963
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8296685714285714e-06,
+      "loss": 0.612,
+      "step": 41964
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8294e-06,
+      "loss": 1.0765,
+      "step": 41965
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8291314285714286e-06,
+      "loss": 1.0968,
+      "step": 41966
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.828862857142857e-06,
+      "loss": 1.0697,
+      "step": 41967
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.828594285714286e-06,
+      "loss": 0.8734,
+      "step": 41968
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8283257142857144e-06,
+      "loss": 1.1237,
+      "step": 41969
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.828057142857143e-06,
+      "loss": 1.0903,
+      "step": 41970
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8277885714285716e-06,
+      "loss": 1.2052,
+      "step": 41971
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.82752e-06,
+      "loss": 0.9695,
+      "step": 41972
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8272514285714283e-06,
+      "loss": 1.0113,
+      "step": 41973
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.826982857142857e-06,
+      "loss": 1.2465,
+      "step": 41974
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.826714285714286e-06,
+      "loss": 1.5547,
+      "step": 41975
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8264457142857146e-06,
+      "loss": 1.6134,
+      "step": 41976
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.826177142857143e-06,
+      "loss": 1.3339,
+      "step": 41977
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8259085714285713e-06,
+      "loss": 1.1706,
+      "step": 41978
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.82564e-06,
+      "loss": 0.9819,
+      "step": 41979
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8253714285714285e-06,
+      "loss": 0.841,
+      "step": 41980
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.825102857142857e-06,
+      "loss": 0.8703,
+      "step": 41981
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.824834285714286e-06,
+      "loss": 1.0527,
+      "step": 41982
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8245657142857143e-06,
+      "loss": 1.0916,
+      "step": 41983
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.824297142857143e-06,
+      "loss": 1.446,
+      "step": 41984
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8240285714285715e-06,
+      "loss": 1.0134,
+      "step": 41985
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.82376e-06,
+      "loss": 1.258,
+      "step": 41986
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8234914285714283e-06,
+      "loss": 1.1861,
+      "step": 41987
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8232228571428573e-06,
+      "loss": 1.1252,
+      "step": 41988
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.822954285714286e-06,
+      "loss": 0.9707,
+      "step": 41989
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8226857142857145e-06,
+      "loss": 1.0548,
+      "step": 41990
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.822417142857143e-06,
+      "loss": 0.7402,
+      "step": 41991
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8221485714285713e-06,
+      "loss": 1.4283,
+      "step": 41992
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.82188e-06,
+      "loss": 1.1285,
+      "step": 41993
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8216114285714285e-06,
+      "loss": 0.7934,
+      "step": 41994
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8213428571428575e-06,
+      "loss": 1.2153,
+      "step": 41995
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8210742857142857e-06,
+      "loss": 1.2384,
+      "step": 41996
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8208057142857143e-06,
+      "loss": 0.587,
+      "step": 41997
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.820537142857143e-06,
+      "loss": 1.3292,
+      "step": 41998
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8202685714285715e-06,
+      "loss": 1.3486,
+      "step": 41999
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.82e-06,
+      "loss": 1.0951,
+      "step": 42000
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8197314285714287e-06,
+      "loss": 1.2104,
+      "step": 42001
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8194628571428573e-06,
+      "loss": 1.5559,
+      "step": 42002
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.819194285714286e-06,
+      "loss": 1.0365,
+      "step": 42003
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8189257142857145e-06,
+      "loss": 1.08,
+      "step": 42004
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.818657142857143e-06,
+      "loss": 1.3845,
+      "step": 42005
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8183885714285713e-06,
+      "loss": 1.1197,
+      "step": 42006
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.81812e-06,
+      "loss": 1.3602,
+      "step": 42007
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.817851428571429e-06,
+      "loss": 1.1719,
+      "step": 42008
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8175828571428575e-06,
+      "loss": 1.0725,
+      "step": 42009
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8173142857142857e-06,
+      "loss": 1.2402,
+      "step": 42010
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8170457142857143e-06,
+      "loss": 1.0108,
+      "step": 42011
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.816777142857143e-06,
+      "loss": 1.0182,
+      "step": 42012
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8165085714285715e-06,
+      "loss": 1.3536,
+      "step": 42013
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.81624e-06,
+      "loss": 1.4525,
+      "step": 42014
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8159714285714287e-06,
+      "loss": 0.9796,
+      "step": 42015
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8157028571428573e-06,
+      "loss": 1.196,
+      "step": 42016
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.815434285714286e-06,
+      "loss": 1.1319,
+      "step": 42017
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8151657142857145e-06,
+      "loss": 0.9333,
+      "step": 42018
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8148971428571427e-06,
+      "loss": 0.9635,
+      "step": 42019
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8146285714285713e-06,
+      "loss": 1.2827,
+      "step": 42020
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8143600000000003e-06,
+      "loss": 1.2934,
+      "step": 42021
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.814091428571429e-06,
+      "loss": 1.0361,
+      "step": 42022
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8138228571428575e-06,
+      "loss": 0.8656,
+      "step": 42023
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8135542857142856e-06,
+      "loss": 0.9309,
+      "step": 42024
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8132857142857142e-06,
+      "loss": 1.0117,
+      "step": 42025
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.813017142857143e-06,
+      "loss": 0.9536,
+      "step": 42026
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8127485714285714e-06,
+      "loss": 1.0814,
+      "step": 42027
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.81248e-06,
+      "loss": 0.8706,
+      "step": 42028
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8122114285714286e-06,
+      "loss": 1.2155,
+      "step": 42029
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8119428571428572e-06,
+      "loss": 1.2887,
+      "step": 42030
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.811674285714286e-06,
+      "loss": 0.8513,
+      "step": 42031
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8114057142857144e-06,
+      "loss": 1.3654,
+      "step": 42032
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8111371428571426e-06,
+      "loss": 1.2677,
+      "step": 42033
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8108685714285716e-06,
+      "loss": 0.9312,
+      "step": 42034
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8106000000000002e-06,
+      "loss": 0.9282,
+      "step": 42035
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.810331428571429e-06,
+      "loss": 1.4015,
+      "step": 42036
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8100628571428574e-06,
+      "loss": 1.3935,
+      "step": 42037
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8097942857142856e-06,
+      "loss": 1.0292,
+      "step": 42038
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8095257142857142e-06,
+      "loss": 1.1107,
+      "step": 42039
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.809257142857143e-06,
+      "loss": 1.3636,
+      "step": 42040
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.808988571428572e-06,
+      "loss": 0.9418,
+      "step": 42041
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.80872e-06,
+      "loss": 0.9036,
+      "step": 42042
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8084514285714286e-06,
+      "loss": 1.2446,
+      "step": 42043
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8081828571428572e-06,
+      "loss": 0.8442,
+      "step": 42044
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.807914285714286e-06,
+      "loss": 1.1989,
+      "step": 42045
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8076457142857144e-06,
+      "loss": 0.7007,
+      "step": 42046
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.807377142857143e-06,
+      "loss": 1.1047,
+      "step": 42047
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8071085714285716e-06,
+      "loss": 1.0871,
+      "step": 42048
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8068400000000002e-06,
+      "loss": 0.8878,
+      "step": 42049
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.806571428571429e-06,
+      "loss": 1.131,
+      "step": 42050
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.806302857142857e-06,
+      "loss": 1.0955,
+      "step": 42051
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8060342857142856e-06,
+      "loss": 1.4851,
+      "step": 42052
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.805765714285714e-06,
+      "loss": 1.5366,
+      "step": 42053
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8054971428571432e-06,
+      "loss": 1.2948,
+      "step": 42054
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.805228571428572e-06,
+      "loss": 1.0731,
+      "step": 42055
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.80496e-06,
+      "loss": 1.2143,
+      "step": 42056
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8046914285714286e-06,
+      "loss": 1.159,
+      "step": 42057
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.804422857142857e-06,
+      "loss": 1.3252,
+      "step": 42058
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.804154285714286e-06,
+      "loss": 0.8285,
+      "step": 42059
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8038857142857144e-06,
+      "loss": 1.3331,
+      "step": 42060
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.803617142857143e-06,
+      "loss": 1.0462,
+      "step": 42061
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8033485714285716e-06,
+      "loss": 1.3052,
+      "step": 42062
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.80308e-06,
+      "loss": 0.9569,
+      "step": 42063
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.802811428571429e-06,
+      "loss": 1.2851,
+      "step": 42064
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.802542857142857e-06,
+      "loss": 0.7872,
+      "step": 42065
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8022742857142856e-06,
+      "loss": 0.9703,
+      "step": 42066
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8020057142857146e-06,
+      "loss": 1.2565,
+      "step": 42067
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.801737142857143e-06,
+      "loss": 0.8438,
+      "step": 42068
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8014685714285718e-06,
+      "loss": 1.0563,
+      "step": 42069
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8012e-06,
+      "loss": 1.3253,
+      "step": 42070
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8009314285714286e-06,
+      "loss": 0.8056,
+      "step": 42071
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.800662857142857e-06,
+      "loss": 1.1083,
+      "step": 42072
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8003942857142858e-06,
+      "loss": 1.5576,
+      "step": 42073
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8001257142857144e-06,
+      "loss": 1.0537,
+      "step": 42074
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.799857142857143e-06,
+      "loss": 1.0865,
+      "step": 42075
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.7995885714285716e-06,
+      "loss": 1.1812,
+      "step": 42076
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.79932e-06,
+      "loss": 1.0289,
+      "step": 42077
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.7990514285714288e-06,
+      "loss": 1.0852,
+      "step": 42078
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.798782857142857e-06,
+      "loss": 0.8405,
+      "step": 42079
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.798514285714286e-06,
+      "loss": 1.6662,
+      "step": 42080
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.7982457142857146e-06,
+      "loss": 1.4015,
+      "step": 42081
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.797977142857143e-06,
+      "loss": 1.1624,
+      "step": 42082
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.7977085714285713e-06,
+      "loss": 0.8113,
+      "step": 42083
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.79744e-06,
+      "loss": 1.3754,
+      "step": 42084
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.7971714285714285e-06,
+      "loss": 1.0547,
+      "step": 42085
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.796902857142857e-06,
+      "loss": 1.0558,
+      "step": 42086
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.796634285714286e-06,
+      "loss": 1.0777,
+      "step": 42087
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7963657142857143e-06,
+      "loss": 1.7485,
+      "step": 42088
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.796097142857143e-06,
+      "loss": 1.2096,
+      "step": 42089
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7958285714285715e-06,
+      "loss": 1.0182,
+      "step": 42090
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.79556e-06,
+      "loss": 0.7617,
+      "step": 42091
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7952914285714287e-06,
+      "loss": 1.057,
+      "step": 42092
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7950228571428573e-06,
+      "loss": 1.1653,
+      "step": 42093
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.794754285714286e-06,
+      "loss": 0.8909,
+      "step": 42094
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7944857142857145e-06,
+      "loss": 1.5087,
+      "step": 42095
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.794217142857143e-06,
+      "loss": 1.2649,
+      "step": 42096
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7939485714285713e-06,
+      "loss": 1.0771,
+      "step": 42097
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.79368e-06,
+      "loss": 1.3245,
+      "step": 42098
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7934114285714285e-06,
+      "loss": 1.0678,
+      "step": 42099
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7931428571428575e-06,
+      "loss": 1.3687,
+      "step": 42100
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.792874285714286e-06,
+      "loss": 1.0575,
+      "step": 42101
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7926057142857143e-06,
+      "loss": 1.0985,
+      "step": 42102
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.792337142857143e-06,
+      "loss": 1.1159,
+      "step": 42103
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7920685714285715e-06,
+      "loss": 1.1922,
+      "step": 42104
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7918e-06,
+      "loss": 1.1373,
+      "step": 42105
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7915314285714287e-06,
+      "loss": 1.1914,
+      "step": 42106
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7912628571428573e-06,
+      "loss": 1.1984,
+      "step": 42107
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.790994285714286e-06,
+      "loss": 0.8517,
+      "step": 42108
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7907257142857145e-06,
+      "loss": 1.166,
+      "step": 42109
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.790457142857143e-06,
+      "loss": 1.6765,
+      "step": 42110
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7901885714285713e-06,
+      "loss": 1.0597,
+      "step": 42111
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.78992e-06,
+      "loss": 0.9923,
+      "step": 42112
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.789651428571429e-06,
+      "loss": 1.0663,
+      "step": 42113
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7893828571428575e-06,
+      "loss": 1.0863,
+      "step": 42114
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7891142857142857e-06,
+      "loss": 0.7716,
+      "step": 42115
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7888457142857143e-06,
+      "loss": 0.8801,
+      "step": 42116
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.788577142857143e-06,
+      "loss": 1.2393,
+      "step": 42117
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7883085714285715e-06,
+      "loss": 1.0827,
+      "step": 42118
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.78804e-06,
+      "loss": 1.1488,
+      "step": 42119
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7877714285714287e-06,
+      "loss": 1.4626,
+      "step": 42120
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7875028571428573e-06,
+      "loss": 0.8498,
+      "step": 42121
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.787234285714286e-06,
+      "loss": 0.9227,
+      "step": 42122
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7869657142857145e-06,
+      "loss": 0.5715,
+      "step": 42123
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.786697142857143e-06,
+      "loss": 1.0511,
+      "step": 42124
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7864285714285713e-06,
+      "loss": 1.2418,
+      "step": 42125
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7861600000000003e-06,
+      "loss": 1.2154,
+      "step": 42126
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.785891428571429e-06,
+      "loss": 1.2351,
+      "step": 42127
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7856228571428575e-06,
+      "loss": 0.7706,
+      "step": 42128
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7853542857142856e-06,
+      "loss": 1.0742,
+      "step": 42129
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7850857142857142e-06,
+      "loss": 0.7315,
+      "step": 42130
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.784817142857143e-06,
+      "loss": 1.0347,
+      "step": 42131
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7845485714285714e-06,
+      "loss": 1.3237,
+      "step": 42132
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7842800000000005e-06,
+      "loss": 1.2544,
+      "step": 42133
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7840114285714286e-06,
+      "loss": 0.9218,
+      "step": 42134
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7837428571428572e-06,
+      "loss": 0.9818,
+      "step": 42135
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.783474285714286e-06,
+      "loss": 0.8614,
+      "step": 42136
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7832057142857144e-06,
+      "loss": 1.199,
+      "step": 42137
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7829371428571426e-06,
+      "loss": 1.1507,
+      "step": 42138
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7826685714285716e-06,
+      "loss": 0.9207,
+      "step": 42139
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7824000000000002e-06,
+      "loss": 0.8944,
+      "step": 42140
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.782131428571429e-06,
+      "loss": 1.4382,
+      "step": 42141
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7818628571428574e-06,
+      "loss": 1.1684,
+      "step": 42142
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7815942857142856e-06,
+      "loss": 0.8887,
+      "step": 42143
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7813257142857142e-06,
+      "loss": 0.6173,
+      "step": 42144
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.781057142857143e-06,
+      "loss": 1.2301,
+      "step": 42145
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.780788571428572e-06,
+      "loss": 0.769,
+      "step": 42146
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.78052e-06,
+      "loss": 1.1288,
+      "step": 42147
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7802514285714286e-06,
+      "loss": 1.1019,
+      "step": 42148
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7799828571428572e-06,
+      "loss": 1.4237,
+      "step": 42149
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.779714285714286e-06,
+      "loss": 1.1037,
+      "step": 42150
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7794457142857144e-06,
+      "loss": 0.8137,
+      "step": 42151
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7791771428571426e-06,
+      "loss": 1.0628,
+      "step": 42152
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7789085714285716e-06,
+      "loss": 0.7325,
+      "step": 42153
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7786400000000002e-06,
+      "loss": 1.5252,
+      "step": 42154
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.778371428571429e-06,
+      "loss": 1.4353,
+      "step": 42155
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7781028571428574e-06,
+      "loss": 1.1231,
+      "step": 42156
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7778342857142856e-06,
+      "loss": 1.3694,
+      "step": 42157
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.777565714285714e-06,
+      "loss": 0.9477,
+      "step": 42158
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7772971428571432e-06,
+      "loss": 1.2777,
+      "step": 42159
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.777028571428572e-06,
+      "loss": 0.9676,
+      "step": 42160
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.77676e-06,
+      "loss": 0.8398,
+      "step": 42161
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7764914285714286e-06,
+      "loss": 1.1961,
+      "step": 42162
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.776222857142857e-06,
+      "loss": 1.1056,
+      "step": 42163
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.775954285714286e-06,
+      "loss": 0.7673,
+      "step": 42164
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7756857142857144e-06,
+      "loss": 1.036,
+      "step": 42165
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.775417142857143e-06,
+      "loss": 1.1839,
+      "step": 42166
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7751485714285716e-06,
+      "loss": 0.8473,
+      "step": 42167
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.77488e-06,
+      "loss": 1.0224,
+      "step": 42168
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7746114285714288e-06,
+      "loss": 1.1708,
+      "step": 42169
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.774342857142857e-06,
+      "loss": 1.2402,
+      "step": 42170
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7740742857142856e-06,
+      "loss": 0.9674,
+      "step": 42171
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7738057142857146e-06,
+      "loss": 0.9881,
+      "step": 42172
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.773537142857143e-06,
+      "loss": 1.1199,
+      "step": 42173
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7732685714285718e-06,
+      "loss": 1.2378,
+      "step": 42174
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.773e-06,
+      "loss": 0.9463,
+      "step": 42175
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7727314285714286e-06,
+      "loss": 1.3307,
+      "step": 42176
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.772462857142857e-06,
+      "loss": 1.0002,
+      "step": 42177
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7721942857142858e-06,
+      "loss": 0.9007,
+      "step": 42178
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7719257142857144e-06,
+      "loss": 1.0288,
+      "step": 42179
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.771657142857143e-06,
+      "loss": 1.1417,
+      "step": 42180
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7713885714285716e-06,
+      "loss": 1.1732,
+      "step": 42181
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.77112e-06,
+      "loss": 1.3585,
+      "step": 42182
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7708514285714288e-06,
+      "loss": 1.16,
+      "step": 42183
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.770582857142857e-06,
+      "loss": 0.9093,
+      "step": 42184
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7703142857142855e-06,
+      "loss": 1.2194,
+      "step": 42185
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7700457142857146e-06,
+      "loss": 0.9884,
+      "step": 42186
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.769777142857143e-06,
+      "loss": 1.2219,
+      "step": 42187
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7695085714285718e-06,
+      "loss": 1.7661,
+      "step": 42188
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.76924e-06,
+      "loss": 0.9754,
+      "step": 42189
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7689714285714285e-06,
+      "loss": 1.1934,
+      "step": 42190
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.768702857142857e-06,
+      "loss": 1.3419,
+      "step": 42191
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.768434285714286e-06,
+      "loss": 0.7638,
+      "step": 42192
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7681657142857143e-06,
+      "loss": 1.2551,
+      "step": 42193
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.767897142857143e-06,
+      "loss": 1.2283,
+      "step": 42194
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7676285714285715e-06,
+      "loss": 0.8824,
+      "step": 42195
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.76736e-06,
+      "loss": 1.2602,
+      "step": 42196
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7670914285714287e-06,
+      "loss": 0.8081,
+      "step": 42197
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.766822857142857e-06,
+      "loss": 1.4188,
+      "step": 42198
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.766554285714286e-06,
+      "loss": 1.1226,
+      "step": 42199
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7662857142857145e-06,
+      "loss": 1.424,
+      "step": 42200
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.766017142857143e-06,
+      "loss": 1.0664,
+      "step": 42201
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7657485714285713e-06,
+      "loss": 1.2726,
+      "step": 42202
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.76548e-06,
+      "loss": 1.3082,
+      "step": 42203
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7652114285714285e-06,
+      "loss": 0.7385,
+      "step": 42204
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.764942857142857e-06,
+      "loss": 0.7154,
+      "step": 42205
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.764674285714286e-06,
+      "loss": 1.1557,
+      "step": 42206
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7644057142857143e-06,
+      "loss": 1.1259,
+      "step": 42207
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.764137142857143e-06,
+      "loss": 0.8357,
+      "step": 42208
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7638685714285715e-06,
+      "loss": 1.3599,
+      "step": 42209
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7636e-06,
+      "loss": 1.1319,
+      "step": 42210
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7633314285714287e-06,
+      "loss": 1.1381,
+      "step": 42211
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7630628571428573e-06,
+      "loss": 0.9226,
+      "step": 42212
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.762794285714286e-06,
+      "loss": 0.9605,
+      "step": 42213
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7625257142857145e-06,
+      "loss": 1.4895,
+      "step": 42214
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.762257142857143e-06,
+      "loss": 1.0155,
+      "step": 42215
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7619885714285713e-06,
+      "loss": 1.1394,
+      "step": 42216
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.76172e-06,
+      "loss": 1.0546,
+      "step": 42217
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7614514285714285e-06,
+      "loss": 1.0686,
+      "step": 42218
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7611828571428575e-06,
+      "loss": 1.0335,
+      "step": 42219
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.760914285714286e-06,
+      "loss": 1.2616,
+      "step": 42220
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7606457142857143e-06,
+      "loss": 1.2614,
+      "step": 42221
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.760377142857143e-06,
+      "loss": 1.0857,
+      "step": 42222
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7601085714285715e-06,
+      "loss": 0.9678,
+      "step": 42223
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.75984e-06,
+      "loss": 1.317,
+      "step": 42224
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7595714285714287e-06,
+      "loss": 1.1177,
+      "step": 42225
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7593028571428573e-06,
+      "loss": 1.1472,
+      "step": 42226
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.759034285714286e-06,
+      "loss": 0.7691,
+      "step": 42227
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7587657142857145e-06,
+      "loss": 1.1161,
+      "step": 42228
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.758497142857143e-06,
+      "loss": 0.9371,
+      "step": 42229
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7582285714285712e-06,
+      "loss": 1.2754,
+      "step": 42230
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.75796e-06,
+      "loss": 0.797,
+      "step": 42231
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.757691428571429e-06,
+      "loss": 1.2781,
+      "step": 42232
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7574228571428575e-06,
+      "loss": 0.8804,
+      "step": 42233
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7571542857142856e-06,
+      "loss": 1.0158,
+      "step": 42234
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7568857142857142e-06,
+      "loss": 0.9561,
+      "step": 42235
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.756617142857143e-06,
+      "loss": 0.9754,
+      "step": 42236
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7563485714285714e-06,
+      "loss": 1.2388,
+      "step": 42237
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.75608e-06,
+      "loss": 1.4299,
+      "step": 42238
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7558114285714286e-06,
+      "loss": 1.124,
+      "step": 42239
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7555428571428572e-06,
+      "loss": 1.2214,
+      "step": 42240
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.755274285714286e-06,
+      "loss": 0.8074,
+      "step": 42241
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7550057142857144e-06,
+      "loss": 1.1705,
+      "step": 42242
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.754737142857143e-06,
+      "loss": 1.2398,
+      "step": 42243
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7544685714285712e-06,
+      "loss": 1.3955,
+      "step": 42244
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7542000000000002e-06,
+      "loss": 1.6566,
+      "step": 42245
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.753931428571429e-06,
+      "loss": 1.0015,
+      "step": 42246
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7536628571428574e-06,
+      "loss": 1.1705,
+      "step": 42247
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7533942857142856e-06,
+      "loss": 1.5497,
+      "step": 42248
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7531257142857142e-06,
+      "loss": 1.1264,
+      "step": 42249
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.752857142857143e-06,
+      "loss": 1.297,
+      "step": 42250
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7525885714285714e-06,
+      "loss": 0.878,
+      "step": 42251
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7523200000000004e-06,
+      "loss": 1.1014,
+      "step": 42252
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7520514285714286e-06,
+      "loss": 1.2441,
+      "step": 42253
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7517828571428572e-06,
+      "loss": 1.0667,
+      "step": 42254
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.751514285714286e-06,
+      "loss": 1.2808,
+      "step": 42255
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7512457142857144e-06,
+      "loss": 1.1371,
+      "step": 42256
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7509771428571426e-06,
+      "loss": 1.38,
+      "step": 42257
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7507085714285716e-06,
+      "loss": 0.9621,
+      "step": 42258
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7504400000000002e-06,
+      "loss": 1.0761,
+      "step": 42259
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.750171428571429e-06,
+      "loss": 1.3498,
+      "step": 42260
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7499028571428574e-06,
+      "loss": 0.9036,
+      "step": 42261
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7496342857142856e-06,
+      "loss": 1.375,
+      "step": 42262
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.749365714285714e-06,
+      "loss": 1.0651,
+      "step": 42263
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.749097142857143e-06,
+      "loss": 1.4399,
+      "step": 42264
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.748828571428572e-06,
+      "loss": 1.3293,
+      "step": 42265
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.74856e-06,
+      "loss": 0.8391,
+      "step": 42266
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7482914285714286e-06,
+      "loss": 1.0797,
+      "step": 42267
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.748022857142857e-06,
+      "loss": 0.9014,
+      "step": 42268
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.747754285714286e-06,
+      "loss": 1.3508,
+      "step": 42269
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7474857142857144e-06,
+      "loss": 1.1186,
+      "step": 42270
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.747217142857143e-06,
+      "loss": 1.0853,
+      "step": 42271
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7469485714285716e-06,
+      "loss": 0.8055,
+      "step": 42272
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.74668e-06,
+      "loss": 0.9348,
+      "step": 42273
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7464114285714288e-06,
+      "loss": 0.9113,
+      "step": 42274
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7461428571428574e-06,
+      "loss": 1.1361,
+      "step": 42275
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7458742857142856e-06,
+      "loss": 1.1569,
+      "step": 42276
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.745605714285714e-06,
+      "loss": 0.9652,
+      "step": 42277
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.745337142857143e-06,
+      "loss": 1.306,
+      "step": 42278
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7450685714285718e-06,
+      "loss": 1.1877,
+      "step": 42279
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7448e-06,
+      "loss": 1.0364,
+      "step": 42280
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7445314285714286e-06,
+      "loss": 0.9297,
+      "step": 42281
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.744262857142857e-06,
+      "loss": 1.1797,
+      "step": 42282
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7439942857142858e-06,
+      "loss": 1.2229,
+      "step": 42283
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7437257142857144e-06,
+      "loss": 1.2209,
+      "step": 42284
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.743457142857143e-06,
+      "loss": 1.3142,
+      "step": 42285
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7431885714285716e-06,
+      "loss": 1.0597,
+      "step": 42286
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.74292e-06,
+      "loss": 1.383,
+      "step": 42287
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7426514285714288e-06,
+      "loss": 0.6502,
+      "step": 42288
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.742382857142857e-06,
+      "loss": 1.2642,
+      "step": 42289
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7421142857142855e-06,
+      "loss": 1.3594,
+      "step": 42290
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7418457142857146e-06,
+      "loss": 1.1949,
+      "step": 42291
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.741577142857143e-06,
+      "loss": 0.7405,
+      "step": 42292
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7413085714285718e-06,
+      "loss": 1.3754,
+      "step": 42293
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.74104e-06,
+      "loss": 0.7889,
+      "step": 42294
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7407714285714285e-06,
+      "loss": 1.1403,
+      "step": 42295
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.740502857142857e-06,
+      "loss": 1.4563,
+      "step": 42296
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7402342857142857e-06,
+      "loss": 0.7707,
+      "step": 42297
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7399657142857148e-06,
+      "loss": 0.9777,
+      "step": 42298
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.739697142857143e-06,
+      "loss": 1.3361,
+      "step": 42299
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7394285714285715e-06,
+      "loss": 0.7159,
+      "step": 42300
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.73916e-06,
+      "loss": 1.121,
+      "step": 42301
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7388914285714287e-06,
+      "loss": 1.3495,
+      "step": 42302
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.738622857142857e-06,
+      "loss": 1.4288,
+      "step": 42303
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.738354285714286e-06,
+      "loss": 1.1015,
+      "step": 42304
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7380857142857145e-06,
+      "loss": 0.9995,
+      "step": 42305
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.737817142857143e-06,
+      "loss": 1.196,
+      "step": 42306
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7375485714285717e-06,
+      "loss": 1.0781,
+      "step": 42307
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.73728e-06,
+      "loss": 0.8984,
+      "step": 42308
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7370114285714285e-06,
+      "loss": 0.9899,
+      "step": 42309
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.736742857142857e-06,
+      "loss": 1.2409,
+      "step": 42310
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.736474285714286e-06,
+      "loss": 0.8928,
+      "step": 42311
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7362057142857143e-06,
+      "loss": 1.5631,
+      "step": 42312
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.735937142857143e-06,
+      "loss": 1.0613,
+      "step": 42313
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7356685714285715e-06,
+      "loss": 0.7627,
+      "step": 42314
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7354e-06,
+      "loss": 1.111,
+      "step": 42315
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7351314285714287e-06,
+      "loss": 1.0858,
+      "step": 42316
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7348628571428573e-06,
+      "loss": 0.8649,
+      "step": 42317
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.734594285714286e-06,
+      "loss": 1.5877,
+      "step": 42318
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7343257142857145e-06,
+      "loss": 1.4206,
+      "step": 42319
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.734057142857143e-06,
+      "loss": 0.89,
+      "step": 42320
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7337885714285713e-06,
+      "loss": 0.9951,
+      "step": 42321
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.73352e-06,
+      "loss": 0.8431,
+      "step": 42322
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7332514285714285e-06,
+      "loss": 1.1058,
+      "step": 42323
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7329828571428575e-06,
+      "loss": 1.1251,
+      "step": 42324
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.732714285714286e-06,
+      "loss": 1.3163,
+      "step": 42325
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7324457142857143e-06,
+      "loss": 0.7996,
+      "step": 42326
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.732177142857143e-06,
+      "loss": 0.8753,
+      "step": 42327
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7319085714285715e-06,
+      "loss": 0.8693,
+      "step": 42328
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.73164e-06,
+      "loss": 1.4858,
+      "step": 42329
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7313714285714287e-06,
+      "loss": 1.0523,
+      "step": 42330
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7311028571428573e-06,
+      "loss": 1.5103,
+      "step": 42331
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.730834285714286e-06,
+      "loss": 1.2568,
+      "step": 42332
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7305657142857145e-06,
+      "loss": 0.9775,
+      "step": 42333
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.730297142857143e-06,
+      "loss": 0.9874,
+      "step": 42334
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7300285714285712e-06,
+      "loss": 1.2807,
+      "step": 42335
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.72976e-06,
+      "loss": 1.135,
+      "step": 42336
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.729491428571429e-06,
+      "loss": 1.1228,
+      "step": 42337
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7292228571428575e-06,
+      "loss": 1.2981,
+      "step": 42338
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.728954285714286e-06,
+      "loss": 0.9999,
+      "step": 42339
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7286857142857142e-06,
+      "loss": 0.8334,
+      "step": 42340
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.728417142857143e-06,
+      "loss": 1.116,
+      "step": 42341
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7281485714285714e-06,
+      "loss": 0.99,
+      "step": 42342
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.72788e-06,
+      "loss": 1.0385,
+      "step": 42343
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7276114285714286e-06,
+      "loss": 1.3136,
+      "step": 42344
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7273428571428572e-06,
+      "loss": 1.0043,
+      "step": 42345
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.727074285714286e-06,
+      "loss": 0.9405,
+      "step": 42346
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7268057142857144e-06,
+      "loss": 1.0215,
+      "step": 42347
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.726537142857143e-06,
+      "loss": 1.1659,
+      "step": 42348
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7262685714285712e-06,
+      "loss": 0.8674,
+      "step": 42349
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7260000000000002e-06,
+      "loss": 1.4082,
+      "step": 42350
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.725731428571429e-06,
+      "loss": 1.2845,
+      "step": 42351
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7254628571428574e-06,
+      "loss": 1.0209,
+      "step": 42352
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7251942857142856e-06,
+      "loss": 1.4754,
+      "step": 42353
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7249257142857142e-06,
+      "loss": 1.1257,
+      "step": 42354
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.724657142857143e-06,
+      "loss": 0.9585,
+      "step": 42355
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7243885714285714e-06,
+      "loss": 1.0261,
+      "step": 42356
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7241200000000004e-06,
+      "loss": 1.0776,
+      "step": 42357
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7238514285714286e-06,
+      "loss": 0.7596,
+      "step": 42358
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7235828571428572e-06,
+      "loss": 1.0221,
+      "step": 42359
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.723314285714286e-06,
+      "loss": 1.1437,
+      "step": 42360
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7230457142857144e-06,
+      "loss": 1.0772,
+      "step": 42361
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.722777142857143e-06,
+      "loss": 1.5937,
+      "step": 42362
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7225085714285716e-06,
+      "loss": 1.4939,
+      "step": 42363
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7222400000000002e-06,
+      "loss": 1.2101,
+      "step": 42364
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.721971428571429e-06,
+      "loss": 0.8011,
+      "step": 42365
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7217028571428574e-06,
+      "loss": 1.1701,
+      "step": 42366
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7214342857142856e-06,
+      "loss": 0.762,
+      "step": 42367
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.721165714285714e-06,
+      "loss": 1.391,
+      "step": 42368
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.720897142857143e-06,
+      "loss": 1.054,
+      "step": 42369
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.720628571428572e-06,
+      "loss": 1.2415,
+      "step": 42370
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7203600000000004e-06,
+      "loss": 1.1796,
+      "step": 42371
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7200914285714286e-06,
+      "loss": 1.3334,
+      "step": 42372
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.719822857142857e-06,
+      "loss": 1.3424,
+      "step": 42373
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7195542857142858e-06,
+      "loss": 1.3681,
+      "step": 42374
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7192857142857144e-06,
+      "loss": 1.2536,
+      "step": 42375
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.719017142857143e-06,
+      "loss": 1.2847,
+      "step": 42376
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7187485714285716e-06,
+      "loss": 1.3752,
+      "step": 42377
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.71848e-06,
+      "loss": 1.2856,
+      "step": 42378
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7182114285714288e-06,
+      "loss": 1.0061,
+      "step": 42379
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7179428571428574e-06,
+      "loss": 1.037,
+      "step": 42380
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7176742857142856e-06,
+      "loss": 1.3179,
+      "step": 42381
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.717405714285714e-06,
+      "loss": 0.835,
+      "step": 42382
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.717137142857143e-06,
+      "loss": 0.9895,
+      "step": 42383
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7168685714285718e-06,
+      "loss": 0.7924,
+      "step": 42384
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7166000000000004e-06,
+      "loss": 0.9634,
+      "step": 42385
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7163314285714286e-06,
+      "loss": 1.3904,
+      "step": 42386
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.716062857142857e-06,
+      "loss": 1.0673,
+      "step": 42387
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7157942857142858e-06,
+      "loss": 1.1553,
+      "step": 42388
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7155257142857144e-06,
+      "loss": 0.9352,
+      "step": 42389
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.715257142857143e-06,
+      "loss": 0.7189,
+      "step": 42390
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7149885714285716e-06,
+      "loss": 0.77,
+      "step": 42391
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.71472e-06,
+      "loss": 1.2998,
+      "step": 42392
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7144514285714288e-06,
+      "loss": 1.0692,
+      "step": 42393
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7141828571428574e-06,
+      "loss": 1.2595,
+      "step": 42394
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7139142857142855e-06,
+      "loss": 1.3074,
+      "step": 42395
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7136457142857146e-06,
+      "loss": 1.0111,
+      "step": 42396
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.713377142857143e-06,
+      "loss": 1.1339,
+      "step": 42397
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7131085714285718e-06,
+      "loss": 1.0142,
+      "step": 42398
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.71284e-06,
+      "loss": 1.1361,
+      "step": 42399
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7125714285714285e-06,
+      "loss": 1.386,
+      "step": 42400
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.712302857142857e-06,
+      "loss": 0.9875,
+      "step": 42401
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7120342857142857e-06,
+      "loss": 1.1847,
+      "step": 42402
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7117657142857148e-06,
+      "loss": 1.1098,
+      "step": 42403
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.711497142857143e-06,
+      "loss": 0.9529,
+      "step": 42404
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7112285714285715e-06,
+      "loss": 0.7915,
+      "step": 42405
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.71096e-06,
+      "loss": 1.1827,
+      "step": 42406
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7106914285714287e-06,
+      "loss": 1.4743,
+      "step": 42407
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.710422857142857e-06,
+      "loss": 0.8087,
+      "step": 42408
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7101542857142855e-06,
+      "loss": 1.506,
+      "step": 42409
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7098857142857145e-06,
+      "loss": 1.0972,
+      "step": 42410
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.709617142857143e-06,
+      "loss": 1.3107,
+      "step": 42411
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7093485714285717e-06,
+      "loss": 1.0063,
+      "step": 42412
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.70908e-06,
+      "loss": 1.6293,
+      "step": 42413
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7088114285714285e-06,
+      "loss": 1.2834,
+      "step": 42414
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.708542857142857e-06,
+      "loss": 0.6722,
+      "step": 42415
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.708274285714286e-06,
+      "loss": 0.5286,
+      "step": 42416
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7080057142857147e-06,
+      "loss": 0.9385,
+      "step": 42417
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.707737142857143e-06,
+      "loss": 1.4336,
+      "step": 42418
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7074685714285715e-06,
+      "loss": 1.4527,
+      "step": 42419
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7072e-06,
+      "loss": 0.5486,
+      "step": 42420
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7069314285714287e-06,
+      "loss": 1.2102,
+      "step": 42421
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.706662857142857e-06,
+      "loss": 1.3704,
+      "step": 42422
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.706394285714286e-06,
+      "loss": 0.9979,
+      "step": 42423
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7061257142857145e-06,
+      "loss": 1.2901,
+      "step": 42424
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.705857142857143e-06,
+      "loss": 1.3455,
+      "step": 42425
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7055885714285717e-06,
+      "loss": 1.0363,
+      "step": 42426
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.70532e-06,
+      "loss": 1.0768,
+      "step": 42427
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7050514285714285e-06,
+      "loss": 0.9986,
+      "step": 42428
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7047828571428575e-06,
+      "loss": 0.747,
+      "step": 42429
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.704514285714286e-06,
+      "loss": 1.0004,
+      "step": 42430
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7042457142857143e-06,
+      "loss": 0.8778,
+      "step": 42431
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.703977142857143e-06,
+      "loss": 1.0806,
+      "step": 42432
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7037085714285715e-06,
+      "loss": 0.8948,
+      "step": 42433
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.70344e-06,
+      "loss": 1.4679,
+      "step": 42434
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7031714285714287e-06,
+      "loss": 1.3669,
+      "step": 42435
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7029028571428573e-06,
+      "loss": 0.8231,
+      "step": 42436
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.702634285714286e-06,
+      "loss": 1.1278,
+      "step": 42437
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.7023657142857145e-06,
+      "loss": 1.568,
+      "step": 42438
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.702097142857143e-06,
+      "loss": 1.2709,
+      "step": 42439
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.7018285714285712e-06,
+      "loss": 0.945,
+      "step": 42440
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.70156e-06,
+      "loss": 0.9598,
+      "step": 42441
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.7012914285714284e-06,
+      "loss": 0.9748,
+      "step": 42442
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.7010228571428575e-06,
+      "loss": 1.0068,
+      "step": 42443
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.700754285714286e-06,
+      "loss": 0.9839,
+      "step": 42444
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.7004857142857142e-06,
+      "loss": 1.3417,
+      "step": 42445
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.700217142857143e-06,
+      "loss": 1.4227,
+      "step": 42446
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6999485714285714e-06,
+      "loss": 1.3329,
+      "step": 42447
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.69968e-06,
+      "loss": 1.0579,
+      "step": 42448
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.699411428571429e-06,
+      "loss": 1.463,
+      "step": 42449
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6991428571428572e-06,
+      "loss": 1.0562,
+      "step": 42450
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.698874285714286e-06,
+      "loss": 1.198,
+      "step": 42451
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6986057142857144e-06,
+      "loss": 1.4096,
+      "step": 42452
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.698337142857143e-06,
+      "loss": 1.0188,
+      "step": 42453
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6980685714285712e-06,
+      "loss": 1.3619,
+      "step": 42454
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6978e-06,
+      "loss": 1.3374,
+      "step": 42455
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.697531428571429e-06,
+      "loss": 0.8381,
+      "step": 42456
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6972628571428574e-06,
+      "loss": 0.8954,
+      "step": 42457
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.696994285714286e-06,
+      "loss": 1.2645,
+      "step": 42458
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6967257142857142e-06,
+      "loss": 0.8933,
+      "step": 42459
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.696457142857143e-06,
+      "loss": 1.0738,
+      "step": 42460
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6961885714285714e-06,
+      "loss": 1.1479,
+      "step": 42461
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.69592e-06,
+      "loss": 0.8159,
+      "step": 42462
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6956514285714286e-06,
+      "loss": 1.3218,
+      "step": 42463
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6953828571428572e-06,
+      "loss": 0.7826,
+      "step": 42464
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.695114285714286e-06,
+      "loss": 1.2248,
+      "step": 42465
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6948457142857144e-06,
+      "loss": 0.9206,
+      "step": 42466
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.694577142857143e-06,
+      "loss": 1.1189,
+      "step": 42467
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.694308571428571e-06,
+      "loss": 1.4987,
+      "step": 42468
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6940400000000002e-06,
+      "loss": 0.8902,
+      "step": 42469
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.693771428571429e-06,
+      "loss": 0.8343,
+      "step": 42470
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6935028571428574e-06,
+      "loss": 0.8802,
+      "step": 42471
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6932342857142856e-06,
+      "loss": 1.2094,
+      "step": 42472
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.692965714285714e-06,
+      "loss": 1.2757,
+      "step": 42473
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.692697142857143e-06,
+      "loss": 1.04,
+      "step": 42474
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6924285714285714e-06,
+      "loss": 0.9928,
+      "step": 42475
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6921600000000004e-06,
+      "loss": 0.9773,
+      "step": 42476
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6918914285714286e-06,
+      "loss": 1.0771,
+      "step": 42477
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.691622857142857e-06,
+      "loss": 0.6261,
+      "step": 42478
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6913542857142858e-06,
+      "loss": 1.0892,
+      "step": 42479
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6910857142857144e-06,
+      "loss": 1.0144,
+      "step": 42480
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.690817142857143e-06,
+      "loss": 1.1547,
+      "step": 42481
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6905485714285716e-06,
+      "loss": 0.8948,
+      "step": 42482
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.69028e-06,
+      "loss": 1.0569,
+      "step": 42483
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6900114285714288e-06,
+      "loss": 0.8972,
+      "step": 42484
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6897428571428574e-06,
+      "loss": 1.4511,
+      "step": 42485
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6894742857142856e-06,
+      "loss": 1.1103,
+      "step": 42486
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.689205714285714e-06,
+      "loss": 1.2346,
+      "step": 42487
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6889371428571428e-06,
+      "loss": 1.2815,
+      "step": 42488
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6886685714285718e-06,
+      "loss": 1.1587,
+      "step": 42489
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6884000000000004e-06,
+      "loss": 1.3949,
+      "step": 42490
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6881314285714286e-06,
+      "loss": 1.377,
+      "step": 42491
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.687862857142857e-06,
+      "loss": 0.9371,
+      "step": 42492
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6875942857142858e-06,
+      "loss": 1.2801,
+      "step": 42493
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6873257142857144e-06,
+      "loss": 1.0921,
+      "step": 42494
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.687057142857143e-06,
+      "loss": 0.8923,
+      "step": 42495
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6867885714285716e-06,
+      "loss": 1.0574,
+      "step": 42496
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.68652e-06,
+      "loss": 1.503,
+      "step": 42497
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6862514285714288e-06,
+      "loss": 0.9925,
+      "step": 42498
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6859828571428574e-06,
+      "loss": 0.8276,
+      "step": 42499
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6857142857142855e-06,
+      "loss": 1.0031,
+      "step": 42500
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.685445714285714e-06,
+      "loss": 0.7704,
+      "step": 42501
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.685177142857143e-06,
+      "loss": 0.9762,
+      "step": 42502
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6849085714285718e-06,
+      "loss": 0.9856,
+      "step": 42503
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6846400000000004e-06,
+      "loss": 1.4883,
+      "step": 42504
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6843714285714285e-06,
+      "loss": 1.3139,
+      "step": 42505
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.684102857142857e-06,
+      "loss": 1.0872,
+      "step": 42506
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6838342857142857e-06,
+      "loss": 1.1072,
+      "step": 42507
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6835657142857143e-06,
+      "loss": 0.8928,
+      "step": 42508
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.683297142857143e-06,
+      "loss": 1.0397,
+      "step": 42509
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6830285714285715e-06,
+      "loss": 1.2298,
+      "step": 42510
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.68276e-06,
+      "loss": 1.1815,
+      "step": 42511
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6824914285714287e-06,
+      "loss": 1.3254,
+      "step": 42512
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6822228571428573e-06,
+      "loss": 1.2552,
+      "step": 42513
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6819542857142855e-06,
+      "loss": 1.4743,
+      "step": 42514
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6816857142857145e-06,
+      "loss": 1.1664,
+      "step": 42515
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.681417142857143e-06,
+      "loss": 0.9122,
+      "step": 42516
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6811485714285717e-06,
+      "loss": 1.2352,
+      "step": 42517
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.68088e-06,
+      "loss": 1.0669,
+      "step": 42518
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6806114285714285e-06,
+      "loss": 0.819,
+      "step": 42519
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.680342857142857e-06,
+      "loss": 1.4042,
+      "step": 42520
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6800742857142857e-06,
+      "loss": 1.3585,
+      "step": 42521
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6798057142857147e-06,
+      "loss": 1.3182,
+      "step": 42522
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.679537142857143e-06,
+      "loss": 1.3944,
+      "step": 42523
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6792685714285715e-06,
+      "loss": 1.2102,
+      "step": 42524
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.679e-06,
+      "loss": 1.0777,
+      "step": 42525
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6787314285714287e-06,
+      "loss": 1.3768,
+      "step": 42526
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.678462857142857e-06,
+      "loss": 1.5097,
+      "step": 42527
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.678194285714286e-06,
+      "loss": 0.6482,
+      "step": 42528
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6779257142857145e-06,
+      "loss": 1.0109,
+      "step": 42529
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.677657142857143e-06,
+      "loss": 1.2837,
+      "step": 42530
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6773885714285717e-06,
+      "loss": 1.2042,
+      "step": 42531
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.67712e-06,
+      "loss": 1.0023,
+      "step": 42532
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6768514285714285e-06,
+      "loss": 0.9781,
+      "step": 42533
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.676582857142857e-06,
+      "loss": 0.9068,
+      "step": 42534
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.676314285714286e-06,
+      "loss": 1.3112,
+      "step": 42535
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6760457142857147e-06,
+      "loss": 1.2263,
+      "step": 42536
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.675777142857143e-06,
+      "loss": 1.1278,
+      "step": 42537
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6755085714285715e-06,
+      "loss": 1.0221,
+      "step": 42538
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.67524e-06,
+      "loss": 0.9921,
+      "step": 42539
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6749714285714287e-06,
+      "loss": 1.2458,
+      "step": 42540
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6747028571428573e-06,
+      "loss": 0.8387,
+      "step": 42541
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.674434285714286e-06,
+      "loss": 0.9746,
+      "step": 42542
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6741657142857145e-06,
+      "loss": 1.2151,
+      "step": 42543
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.673897142857143e-06,
+      "loss": 1.4456,
+      "step": 42544
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6736285714285717e-06,
+      "loss": 1.239,
+      "step": 42545
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.67336e-06,
+      "loss": 1.093,
+      "step": 42546
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6730914285714284e-06,
+      "loss": 0.8116,
+      "step": 42547
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6728228571428575e-06,
+      "loss": 0.8746,
+      "step": 42548
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.672554285714286e-06,
+      "loss": 0.7288,
+      "step": 42549
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6722857142857142e-06,
+      "loss": 1.2603,
+      "step": 42550
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.672017142857143e-06,
+      "loss": 0.9593,
+      "step": 42551
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6717485714285714e-06,
+      "loss": 1.0076,
+      "step": 42552
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.67148e-06,
+      "loss": 0.9915,
+      "step": 42553
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6712114285714286e-06,
+      "loss": 1.1111,
+      "step": 42554
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6709428571428572e-06,
+      "loss": 1.0295,
+      "step": 42555
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.670674285714286e-06,
+      "loss": 1.3019,
+      "step": 42556
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6704057142857144e-06,
+      "loss": 1.1638,
+      "step": 42557
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.670137142857143e-06,
+      "loss": 1.1106,
+      "step": 42558
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6698685714285712e-06,
+      "loss": 0.6931,
+      "step": 42559
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6696e-06,
+      "loss": 1.0875,
+      "step": 42560
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.669331428571429e-06,
+      "loss": 1.2291,
+      "step": 42561
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6690628571428574e-06,
+      "loss": 0.7453,
+      "step": 42562
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.668794285714286e-06,
+      "loss": 1.1528,
+      "step": 42563
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6685257142857142e-06,
+      "loss": 1.2856,
+      "step": 42564
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.668257142857143e-06,
+      "loss": 1.3342,
+      "step": 42565
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6679885714285714e-06,
+      "loss": 1.1183,
+      "step": 42566
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.66772e-06,
+      "loss": 1.003,
+      "step": 42567
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.667451428571429e-06,
+      "loss": 0.7817,
+      "step": 42568
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6671828571428572e-06,
+      "loss": 1.0985,
+      "step": 42569
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.666914285714286e-06,
+      "loss": 0.8961,
+      "step": 42570
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6666457142857144e-06,
+      "loss": 1.3896,
+      "step": 42571
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.666377142857143e-06,
+      "loss": 1.4748,
+      "step": 42572
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.666108571428571e-06,
+      "loss": 1.1286,
+      "step": 42573
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6658400000000002e-06,
+      "loss": 1.0019,
+      "step": 42574
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.665571428571429e-06,
+      "loss": 1.2127,
+      "step": 42575
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6653028571428574e-06,
+      "loss": 0.9403,
+      "step": 42576
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.665034285714286e-06,
+      "loss": 0.8949,
+      "step": 42577
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.664765714285714e-06,
+      "loss": 1.2341,
+      "step": 42578
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6644971428571428e-06,
+      "loss": 1.2204,
+      "step": 42579
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6642285714285714e-06,
+      "loss": 1.0877,
+      "step": 42580
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6639600000000004e-06,
+      "loss": 1.2349,
+      "step": 42581
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6636914285714286e-06,
+      "loss": 1.3573,
+      "step": 42582
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.663422857142857e-06,
+      "loss": 1.4261,
+      "step": 42583
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6631542857142858e-06,
+      "loss": 1.0025,
+      "step": 42584
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6628857142857144e-06,
+      "loss": 0.7999,
+      "step": 42585
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.662617142857143e-06,
+      "loss": 1.1317,
+      "step": 42586
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6623485714285716e-06,
+      "loss": 0.9996,
+      "step": 42587
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.66208e-06,
+      "loss": 1.3411,
+      "step": 42588
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6618114285714288e-06,
+      "loss": 0.8632,
+      "step": 42589
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6615428571428574e-06,
+      "loss": 0.8059,
+      "step": 42590
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.661274285714286e-06,
+      "loss": 1.4842,
+      "step": 42591
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.661005714285714e-06,
+      "loss": 1.1646,
+      "step": 42592
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6607371428571428e-06,
+      "loss": 1.439,
+      "step": 42593
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6604685714285718e-06,
+      "loss": 1.1004,
+      "step": 42594
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6602000000000004e-06,
+      "loss": 1.3,
+      "step": 42595
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6599314285714286e-06,
+      "loss": 1.0749,
+      "step": 42596
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.659662857142857e-06,
+      "loss": 0.8856,
+      "step": 42597
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6593942857142858e-06,
+      "loss": 1.3247,
+      "step": 42598
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6591257142857144e-06,
+      "loss": 0.6553,
+      "step": 42599
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.658857142857143e-06,
+      "loss": 1.0558,
+      "step": 42600
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6585885714285716e-06,
+      "loss": 0.9386,
+      "step": 42601
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.65832e-06,
+      "loss": 1.3511,
+      "step": 42602
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6580514285714288e-06,
+      "loss": 0.9854,
+      "step": 42603
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6577828571428574e-06,
+      "loss": 0.752,
+      "step": 42604
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6575142857142855e-06,
+      "loss": 1.2252,
+      "step": 42605
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.657245714285714e-06,
+      "loss": 1.1342,
+      "step": 42606
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.656977142857143e-06,
+      "loss": 1.2189,
+      "step": 42607
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6567085714285718e-06,
+      "loss": 1.0325,
+      "step": 42608
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6564400000000004e-06,
+      "loss": 0.9676,
+      "step": 42609
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6561714285714285e-06,
+      "loss": 1.0575,
+      "step": 42610
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.655902857142857e-06,
+      "loss": 0.9084,
+      "step": 42611
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6556342857142857e-06,
+      "loss": 0.9029,
+      "step": 42612
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6553657142857143e-06,
+      "loss": 0.6296,
+      "step": 42613
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.655097142857143e-06,
+      "loss": 1.553,
+      "step": 42614
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6548285714285715e-06,
+      "loss": 1.1067,
+      "step": 42615
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.65456e-06,
+      "loss": 0.9104,
+      "step": 42616
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6542914285714287e-06,
+      "loss": 1.3852,
+      "step": 42617
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6540228571428573e-06,
+      "loss": 1.417,
+      "step": 42618
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6537542857142855e-06,
+      "loss": 1.3825,
+      "step": 42619
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6534857142857145e-06,
+      "loss": 0.9475,
+      "step": 42620
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.653217142857143e-06,
+      "loss": 1.1961,
+      "step": 42621
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6529485714285717e-06,
+      "loss": 1.3845,
+      "step": 42622
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6526800000000003e-06,
+      "loss": 1.2136,
+      "step": 42623
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6524114285714285e-06,
+      "loss": 1.3655,
+      "step": 42624
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.652142857142857e-06,
+      "loss": 1.5438,
+      "step": 42625
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6518742857142857e-06,
+      "loss": 1.1417,
+      "step": 42626
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6516057142857147e-06,
+      "loss": 0.9823,
+      "step": 42627
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.651337142857143e-06,
+      "loss": 1.3467,
+      "step": 42628
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6510685714285715e-06,
+      "loss": 1.0809,
+      "step": 42629
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6508e-06,
+      "loss": 1.0189,
+      "step": 42630
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6505314285714287e-06,
+      "loss": 1.2258,
+      "step": 42631
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6502628571428573e-06,
+      "loss": 1.5798,
+      "step": 42632
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6499942857142855e-06,
+      "loss": 1.2515,
+      "step": 42633
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6497257142857145e-06,
+      "loss": 1.427,
+      "step": 42634
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.649457142857143e-06,
+      "loss": 1.1637,
+      "step": 42635
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6491885714285717e-06,
+      "loss": 0.8304,
+      "step": 42636
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.64892e-06,
+      "loss": 0.9411,
+      "step": 42637
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6486514285714285e-06,
+      "loss": 1.0956,
+      "step": 42638
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.648382857142857e-06,
+      "loss": 1.0252,
+      "step": 42639
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.648114285714286e-06,
+      "loss": 1.1579,
+      "step": 42640
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6478457142857147e-06,
+      "loss": 1.3832,
+      "step": 42641
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.647577142857143e-06,
+      "loss": 1.0792,
+      "step": 42642
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6473085714285715e-06,
+      "loss": 0.9963,
+      "step": 42643
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.64704e-06,
+      "loss": 1.338,
+      "step": 42644
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6467714285714287e-06,
+      "loss": 0.9959,
+      "step": 42645
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.646502857142857e-06,
+      "loss": 1.1107,
+      "step": 42646
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.646234285714286e-06,
+      "loss": 0.7925,
+      "step": 42647
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6459657142857145e-06,
+      "loss": 0.9587,
+      "step": 42648
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.645697142857143e-06,
+      "loss": 1.0009,
+      "step": 42649
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6454285714285717e-06,
+      "loss": 1.067,
+      "step": 42650
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.64516e-06,
+      "loss": 1.5454,
+      "step": 42651
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6448914285714284e-06,
+      "loss": 1.2337,
+      "step": 42652
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6446228571428575e-06,
+      "loss": 0.9174,
+      "step": 42653
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.644354285714286e-06,
+      "loss": 1.0717,
+      "step": 42654
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6440857142857147e-06,
+      "loss": 0.6179,
+      "step": 42655
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.643817142857143e-06,
+      "loss": 1.0724,
+      "step": 42656
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6435485714285714e-06,
+      "loss": 0.9654,
+      "step": 42657
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.64328e-06,
+      "loss": 1.1727,
+      "step": 42658
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6430114285714286e-06,
+      "loss": 1.4057,
+      "step": 42659
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6427428571428572e-06,
+      "loss": 1.3887,
+      "step": 42660
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.642474285714286e-06,
+      "loss": 1.2591,
+      "step": 42661
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6422057142857144e-06,
+      "loss": 1.471,
+      "step": 42662
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.641937142857143e-06,
+      "loss": 0.828,
+      "step": 42663
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6416685714285716e-06,
+      "loss": 1.4149,
+      "step": 42664
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6414e-06,
+      "loss": 0.9512,
+      "step": 42665
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6411314285714284e-06,
+      "loss": 1.0318,
+      "step": 42666
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6408628571428574e-06,
+      "loss": 1.1315,
+      "step": 42667
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.640594285714286e-06,
+      "loss": 1.2496,
+      "step": 42668
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6403257142857142e-06,
+      "loss": 1.4063,
+      "step": 42669
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.640057142857143e-06,
+      "loss": 0.7381,
+      "step": 42670
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6397885714285714e-06,
+      "loss": 1.1704,
+      "step": 42671
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.63952e-06,
+      "loss": 0.9335,
+      "step": 42672
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.639251428571429e-06,
+      "loss": 1.3048,
+      "step": 42673
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6389828571428572e-06,
+      "loss": 1.0172,
+      "step": 42674
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.638714285714286e-06,
+      "loss": 1.0807,
+      "step": 42675
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6384457142857144e-06,
+      "loss": 1.0872,
+      "step": 42676
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.638177142857143e-06,
+      "loss": 1.2723,
+      "step": 42677
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.637908571428571e-06,
+      "loss": 1.1912,
+      "step": 42678
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6376399999999998e-06,
+      "loss": 0.9087,
+      "step": 42679
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.637371428571429e-06,
+      "loss": 1.4867,
+      "step": 42680
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6371028571428574e-06,
+      "loss": 0.9292,
+      "step": 42681
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.636834285714286e-06,
+      "loss": 1.355,
+      "step": 42682
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.636565714285714e-06,
+      "loss": 0.6891,
+      "step": 42683
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6362971428571428e-06,
+      "loss": 0.8858,
+      "step": 42684
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6360285714285714e-06,
+      "loss": 1.1011,
+      "step": 42685
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.63576e-06,
+      "loss": 0.8673,
+      "step": 42686
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.635491428571429e-06,
+      "loss": 0.6096,
+      "step": 42687
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.635222857142857e-06,
+      "loss": 0.6961,
+      "step": 42688
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6349542857142858e-06,
+      "loss": 0.9662,
+      "step": 42689
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6346857142857144e-06,
+      "loss": 1.229,
+      "step": 42690
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.634417142857143e-06,
+      "loss": 1.2123,
+      "step": 42691
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.634148571428571e-06,
+      "loss": 1.2096,
+      "step": 42692
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.63388e-06,
+      "loss": 0.9942,
+      "step": 42693
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6336114285714288e-06,
+      "loss": 1.2671,
+      "step": 42694
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6333428571428574e-06,
+      "loss": 1.0384,
+      "step": 42695
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.633074285714286e-06,
+      "loss": 0.9592,
+      "step": 42696
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.632805714285714e-06,
+      "loss": 1.0946,
+      "step": 42697
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6325371428571428e-06,
+      "loss": 1.1315,
+      "step": 42698
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6322685714285714e-06,
+      "loss": 0.6584,
+      "step": 42699
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6320000000000004e-06,
+      "loss": 1.1357,
+      "step": 42700
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6317314285714286e-06,
+      "loss": 1.2579,
+      "step": 42701
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.631462857142857e-06,
+      "loss": 1.0957,
+      "step": 42702
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6311942857142858e-06,
+      "loss": 0.8811,
+      "step": 42703
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6309257142857144e-06,
+      "loss": 1.1614,
+      "step": 42704
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.630657142857143e-06,
+      "loss": 1.2993,
+      "step": 42705
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6303885714285716e-06,
+      "loss": 0.7341,
+      "step": 42706
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.63012e-06,
+      "loss": 1.1773,
+      "step": 42707
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6298514285714288e-06,
+      "loss": 0.8612,
+      "step": 42708
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6295828571428574e-06,
+      "loss": 1.1954,
+      "step": 42709
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.629314285714286e-06,
+      "loss": 1.1,
+      "step": 42710
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.629045714285714e-06,
+      "loss": 0.9691,
+      "step": 42711
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6287771428571427e-06,
+      "loss": 1.2182,
+      "step": 42712
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6285085714285718e-06,
+      "loss": 1.1174,
+      "step": 42713
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6282400000000004e-06,
+      "loss": 1.3744,
+      "step": 42714
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6279714285714285e-06,
+      "loss": 1.5417,
+      "step": 42715
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.627702857142857e-06,
+      "loss": 0.9765,
+      "step": 42716
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6274342857142857e-06,
+      "loss": 1.0376,
+      "step": 42717
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6271657142857143e-06,
+      "loss": 1.2359,
+      "step": 42718
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.626897142857143e-06,
+      "loss": 1.4791,
+      "step": 42719
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6266285714285715e-06,
+      "loss": 1.3489,
+      "step": 42720
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.62636e-06,
+      "loss": 1.0363,
+      "step": 42721
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6260914285714287e-06,
+      "loss": 1.3913,
+      "step": 42722
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6258228571428573e-06,
+      "loss": 1.0678,
+      "step": 42723
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6255542857142855e-06,
+      "loss": 0.9129,
+      "step": 42724
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.625285714285714e-06,
+      "loss": 1.1666,
+      "step": 42725
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.625017142857143e-06,
+      "loss": 1.1839,
+      "step": 42726
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6247485714285717e-06,
+      "loss": 0.9668,
+      "step": 42727
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6244800000000003e-06,
+      "loss": 0.8836,
+      "step": 42728
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6242114285714285e-06,
+      "loss": 1.1384,
+      "step": 42729
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.623942857142857e-06,
+      "loss": 1.1876,
+      "step": 42730
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6236742857142857e-06,
+      "loss": 1.0006,
+      "step": 42731
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6234057142857143e-06,
+      "loss": 1.0881,
+      "step": 42732
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.623137142857143e-06,
+      "loss": 1.1222,
+      "step": 42733
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6228685714285715e-06,
+      "loss": 0.8628,
+      "step": 42734
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6226e-06,
+      "loss": 1.1874,
+      "step": 42735
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6223314285714287e-06,
+      "loss": 0.9345,
+      "step": 42736
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6220628571428573e-06,
+      "loss": 1.3611,
+      "step": 42737
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6217942857142855e-06,
+      "loss": 1.1229,
+      "step": 42738
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6215257142857145e-06,
+      "loss": 1.1809,
+      "step": 42739
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.621257142857143e-06,
+      "loss": 1.1959,
+      "step": 42740
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6209885714285717e-06,
+      "loss": 1.1869,
+      "step": 42741
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6207200000000003e-06,
+      "loss": 1.3422,
+      "step": 42742
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6204514285714285e-06,
+      "loss": 0.8616,
+      "step": 42743
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.620182857142857e-06,
+      "loss": 1.0374,
+      "step": 42744
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6199142857142857e-06,
+      "loss": 1.431,
+      "step": 42745
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6196457142857147e-06,
+      "loss": 1.2494,
+      "step": 42746
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.619377142857143e-06,
+      "loss": 1.0461,
+      "step": 42747
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6191085714285715e-06,
+      "loss": 1.1909,
+      "step": 42748
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.61884e-06,
+      "loss": 1.2161,
+      "step": 42749
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6185714285714287e-06,
+      "loss": 1.0891,
+      "step": 42750
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6183028571428573e-06,
+      "loss": 1.3302,
+      "step": 42751
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.618034285714286e-06,
+      "loss": 1.2673,
+      "step": 42752
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6177657142857145e-06,
+      "loss": 0.974,
+      "step": 42753
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.617497142857143e-06,
+      "loss": 1.3073,
+      "step": 42754
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6172285714285717e-06,
+      "loss": 1.1645,
+      "step": 42755
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.61696e-06,
+      "loss": 1.2982,
+      "step": 42756
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6166914285714284e-06,
+      "loss": 1.3242,
+      "step": 42757
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.616422857142857e-06,
+      "loss": 1.249,
+      "step": 42758
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.616154285714286e-06,
+      "loss": 1.4234,
+      "step": 42759
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6158857142857147e-06,
+      "loss": 0.911,
+      "step": 42760
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.615617142857143e-06,
+      "loss": 1.1095,
+      "step": 42761
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6153485714285714e-06,
+      "loss": 1.0578,
+      "step": 42762
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.61508e-06,
+      "loss": 0.9838,
+      "step": 42763
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6148114285714286e-06,
+      "loss": 0.9637,
+      "step": 42764
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6145428571428572e-06,
+      "loss": 0.7124,
+      "step": 42765
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.614274285714286e-06,
+      "loss": 1.5357,
+      "step": 42766
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6140057142857144e-06,
+      "loss": 0.8771,
+      "step": 42767
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.613737142857143e-06,
+      "loss": 1.5445,
+      "step": 42768
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6134685714285716e-06,
+      "loss": 1.3306,
+      "step": 42769
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6132e-06,
+      "loss": 0.9407,
+      "step": 42770
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6129314285714284e-06,
+      "loss": 1.504,
+      "step": 42771
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6126628571428574e-06,
+      "loss": 1.0836,
+      "step": 42772
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.612394285714286e-06,
+      "loss": 1.0548,
+      "step": 42773
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6121257142857146e-06,
+      "loss": 1.207,
+      "step": 42774
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.611857142857143e-06,
+      "loss": 1.2391,
+      "step": 42775
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6115885714285714e-06,
+      "loss": 0.8987,
+      "step": 42776
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.61132e-06,
+      "loss": 1.1788,
+      "step": 42777
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6110514285714286e-06,
+      "loss": 0.8487,
+      "step": 42778
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6107828571428572e-06,
+      "loss": 1.1872,
+      "step": 42779
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.610514285714286e-06,
+      "loss": 1.3712,
+      "step": 42780
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6102457142857144e-06,
+      "loss": 1.537,
+      "step": 42781
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.609977142857143e-06,
+      "loss": 1.0321,
+      "step": 42782
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6097085714285716e-06,
+      "loss": 1.4641,
+      "step": 42783
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6094399999999998e-06,
+      "loss": 0.9458,
+      "step": 42784
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.609171428571429e-06,
+      "loss": 1.3699,
+      "step": 42785
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6089028571428574e-06,
+      "loss": 1.204,
+      "step": 42786
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.608634285714286e-06,
+      "loss": 1.0375,
+      "step": 42787
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.608365714285714e-06,
+      "loss": 1.1907,
+      "step": 42788
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6080971428571428e-06,
+      "loss": 1.0526,
+      "step": 42789
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6078285714285714e-06,
+      "loss": 1.0492,
+      "step": 42790
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.60756e-06,
+      "loss": 1.2384,
+      "step": 42791
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.607291428571429e-06,
+      "loss": 1.0819,
+      "step": 42792
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.607022857142857e-06,
+      "loss": 0.9669,
+      "step": 42793
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6067542857142858e-06,
+      "loss": 1.0218,
+      "step": 42794
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6064857142857144e-06,
+      "loss": 1.0044,
+      "step": 42795
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.606217142857143e-06,
+      "loss": 0.9076,
+      "step": 42796
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6059485714285716e-06,
+      "loss": 0.81,
+      "step": 42797
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.60568e-06,
+      "loss": 1.1498,
+      "step": 42798
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6054114285714288e-06,
+      "loss": 1.2399,
+      "step": 42799
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6051428571428574e-06,
+      "loss": 1.1499,
+      "step": 42800
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.604874285714286e-06,
+      "loss": 1.0446,
+      "step": 42801
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.604605714285714e-06,
+      "loss": 0.907,
+      "step": 42802
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6043371428571428e-06,
+      "loss": 1.0575,
+      "step": 42803
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6040685714285714e-06,
+      "loss": 1.7189,
+      "step": 42804
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6038000000000004e-06,
+      "loss": 0.8756,
+      "step": 42805
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.603531428571429e-06,
+      "loss": 1.2064,
+      "step": 42806
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.603262857142857e-06,
+      "loss": 0.7807,
+      "step": 42807
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6029942857142858e-06,
+      "loss": 1.2186,
+      "step": 42808
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6027257142857144e-06,
+      "loss": 1.1389,
+      "step": 42809
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.602457142857143e-06,
+      "loss": 1.1729,
+      "step": 42810
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6021885714285716e-06,
+      "loss": 1.3488,
+      "step": 42811
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.60192e-06,
+      "loss": 1.4528,
+      "step": 42812
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6016514285714288e-06,
+      "loss": 0.9829,
+      "step": 42813
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6013828571428574e-06,
+      "loss": 0.7397,
+      "step": 42814
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.601114285714286e-06,
+      "loss": 1.6085,
+      "step": 42815
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.600845714285714e-06,
+      "loss": 1.308,
+      "step": 42816
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6005771428571427e-06,
+      "loss": 1.0928,
+      "step": 42817
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6003085714285718e-06,
+      "loss": 0.9258,
+      "step": 42818
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6000400000000004e-06,
+      "loss": 0.8629,
+      "step": 42819
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5997714285714285e-06,
+      "loss": 1.3023,
+      "step": 42820
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.599502857142857e-06,
+      "loss": 1.0731,
+      "step": 42821
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5992342857142857e-06,
+      "loss": 1.1566,
+      "step": 42822
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5989657142857143e-06,
+      "loss": 1.614,
+      "step": 42823
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.598697142857143e-06,
+      "loss": 1.3582,
+      "step": 42824
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5984285714285715e-06,
+      "loss": 1.3486,
+      "step": 42825
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.59816e-06,
+      "loss": 1.3788,
+      "step": 42826
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5978914285714287e-06,
+      "loss": 1.7792,
+      "step": 42827
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5976228571428573e-06,
+      "loss": 0.8632,
+      "step": 42828
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.597354285714286e-06,
+      "loss": 1.1295,
+      "step": 42829
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.597085714285714e-06,
+      "loss": 1.2029,
+      "step": 42830
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.596817142857143e-06,
+      "loss": 1.0468,
+      "step": 42831
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5965485714285717e-06,
+      "loss": 1.0333,
+      "step": 42832
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5962800000000003e-06,
+      "loss": 1.1496,
+      "step": 42833
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5960114285714285e-06,
+      "loss": 1.1515,
+      "step": 42834
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.595742857142857e-06,
+      "loss": 0.9894,
+      "step": 42835
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5954742857142857e-06,
+      "loss": 1.3107,
+      "step": 42836
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5952057142857143e-06,
+      "loss": 1.0242,
+      "step": 42837
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5949371428571433e-06,
+      "loss": 1.3581,
+      "step": 42838
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5946685714285715e-06,
+      "loss": 0.9836,
+      "step": 42839
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5944e-06,
+      "loss": 0.8619,
+      "step": 42840
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5941314285714287e-06,
+      "loss": 1.2017,
+      "step": 42841
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5938628571428573e-06,
+      "loss": 0.973,
+      "step": 42842
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5935942857142855e-06,
+      "loss": 1.5135,
+      "step": 42843
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5933257142857145e-06,
+      "loss": 1.1125,
+      "step": 42844
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.593057142857143e-06,
+      "loss": 0.9424,
+      "step": 42845
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5927885714285717e-06,
+      "loss": 1.3332,
+      "step": 42846
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5925200000000003e-06,
+      "loss": 1.2183,
+      "step": 42847
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5922514285714285e-06,
+      "loss": 0.8999,
+      "step": 42848
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.591982857142857e-06,
+      "loss": 1.1398,
+      "step": 42849
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5917142857142857e-06,
+      "loss": 0.7286,
+      "step": 42850
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5914457142857147e-06,
+      "loss": 1.2245,
+      "step": 42851
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.591177142857143e-06,
+      "loss": 1.1612,
+      "step": 42852
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5909085714285715e-06,
+      "loss": 1.0342,
+      "step": 42853
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.59064e-06,
+      "loss": 0.7901,
+      "step": 42854
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5903714285714287e-06,
+      "loss": 1.1566,
+      "step": 42855
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5901028571428573e-06,
+      "loss": 1.4108,
+      "step": 42856
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5898342857142854e-06,
+      "loss": 0.9714,
+      "step": 42857
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5895657142857145e-06,
+      "loss": 1.5629,
+      "step": 42858
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.589297142857143e-06,
+      "loss": 0.9926,
+      "step": 42859
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5890285714285717e-06,
+      "loss": 1.3187,
+      "step": 42860
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5887600000000003e-06,
+      "loss": 1.2051,
+      "step": 42861
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5884914285714284e-06,
+      "loss": 0.9236,
+      "step": 42862
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.588222857142857e-06,
+      "loss": 1.1338,
+      "step": 42863
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.587954285714286e-06,
+      "loss": 1.1862,
+      "step": 42864
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5876857142857147e-06,
+      "loss": 1.0118,
+      "step": 42865
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.587417142857143e-06,
+      "loss": 1.5631,
+      "step": 42866
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5871485714285714e-06,
+      "loss": 1.135,
+      "step": 42867
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.58688e-06,
+      "loss": 1.3127,
+      "step": 42868
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5866114285714286e-06,
+      "loss": 0.965,
+      "step": 42869
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5863428571428572e-06,
+      "loss": 1.2216,
+      "step": 42870
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.586074285714286e-06,
+      "loss": 0.9555,
+      "step": 42871
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5858057142857144e-06,
+      "loss": 0.8019,
+      "step": 42872
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.585537142857143e-06,
+      "loss": 0.974,
+      "step": 42873
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5852685714285716e-06,
+      "loss": 0.6624,
+      "step": 42874
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.585e-06,
+      "loss": 1.3556,
+      "step": 42875
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5847314285714284e-06,
+      "loss": 0.9913,
+      "step": 42876
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5844628571428574e-06,
+      "loss": 1.2085,
+      "step": 42877
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.584194285714286e-06,
+      "loss": 1.3086,
+      "step": 42878
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5839257142857146e-06,
+      "loss": 1.4665,
+      "step": 42879
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.583657142857143e-06,
+      "loss": 1.374,
+      "step": 42880
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5833885714285714e-06,
+      "loss": 1.4328,
+      "step": 42881
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.58312e-06,
+      "loss": 0.9036,
+      "step": 42882
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5828514285714286e-06,
+      "loss": 0.8991,
+      "step": 42883
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5825828571428576e-06,
+      "loss": 1.0551,
+      "step": 42884
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.582314285714286e-06,
+      "loss": 0.651,
+      "step": 42885
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5820457142857144e-06,
+      "loss": 1.3167,
+      "step": 42886
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.581777142857143e-06,
+      "loss": 1.446,
+      "step": 42887
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5815085714285716e-06,
+      "loss": 0.7503,
+      "step": 42888
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5812399999999998e-06,
+      "loss": 0.9173,
+      "step": 42889
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5809714285714284e-06,
+      "loss": 1.0352,
+      "step": 42890
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5807028571428574e-06,
+      "loss": 1.6851,
+      "step": 42891
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.580434285714286e-06,
+      "loss": 1.059,
+      "step": 42892
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5801657142857146e-06,
+      "loss": 1.2053,
+      "step": 42893
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5798971428571428e-06,
+      "loss": 0.9901,
+      "step": 42894
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5796285714285714e-06,
+      "loss": 0.8276,
+      "step": 42895
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.57936e-06,
+      "loss": 1.1956,
+      "step": 42896
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.579091428571429e-06,
+      "loss": 0.9311,
+      "step": 42897
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.578822857142857e-06,
+      "loss": 1.3703,
+      "step": 42898
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5785542857142858e-06,
+      "loss": 1.1828,
+      "step": 42899
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5782857142857144e-06,
+      "loss": 0.6162,
+      "step": 42900
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.578017142857143e-06,
+      "loss": 0.9639,
+      "step": 42901
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5777485714285716e-06,
+      "loss": 0.9638,
+      "step": 42902
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5774799999999998e-06,
+      "loss": 0.8987,
+      "step": 42903
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5772114285714288e-06,
+      "loss": 1.0654,
+      "step": 42904
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5769428571428574e-06,
+      "loss": 0.9606,
+      "step": 42905
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.576674285714286e-06,
+      "loss": 1.3952,
+      "step": 42906
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.576405714285714e-06,
+      "loss": 1.5762,
+      "step": 42907
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5761371428571428e-06,
+      "loss": 0.9118,
+      "step": 42908
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5758685714285714e-06,
+      "loss": 0.9066,
+      "step": 42909
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5756e-06,
+      "loss": 1.131,
+      "step": 42910
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.575331428571429e-06,
+      "loss": 0.8741,
+      "step": 42911
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.575062857142857e-06,
+      "loss": 1.1264,
+      "step": 42912
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5747942857142858e-06,
+      "loss": 0.6813,
+      "step": 42913
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5745257142857144e-06,
+      "loss": 1.0672,
+      "step": 42914
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.574257142857143e-06,
+      "loss": 1.3278,
+      "step": 42915
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5739885714285716e-06,
+      "loss": 0.933,
+      "step": 42916
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.57372e-06,
+      "loss": 0.9926,
+      "step": 42917
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5734514285714288e-06,
+      "loss": 1.2666,
+      "step": 42918
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5731828571428574e-06,
+      "loss": 0.8871,
+      "step": 42919
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.572914285714286e-06,
+      "loss": 0.8835,
+      "step": 42920
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.572645714285714e-06,
+      "loss": 1.531,
+      "step": 42921
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5723771428571427e-06,
+      "loss": 0.8801,
+      "step": 42922
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5721085714285713e-06,
+      "loss": 0.9248,
+      "step": 42923
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5718400000000004e-06,
+      "loss": 0.7645,
+      "step": 42924
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.571571428571429e-06,
+      "loss": 1.1678,
+      "step": 42925
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.571302857142857e-06,
+      "loss": 1.1656,
+      "step": 42926
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5710342857142857e-06,
+      "loss": 1.4659,
+      "step": 42927
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5707657142857143e-06,
+      "loss": 1.5729,
+      "step": 42928
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.570497142857143e-06,
+      "loss": 1.2247,
+      "step": 42929
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5702285714285715e-06,
+      "loss": 0.977,
+      "step": 42930
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.56996e-06,
+      "loss": 1.5134,
+      "step": 42931
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5696914285714287e-06,
+      "loss": 1.1241,
+      "step": 42932
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5694228571428573e-06,
+      "loss": 1.1548,
+      "step": 42933
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.569154285714286e-06,
+      "loss": 0.6045,
+      "step": 42934
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.568885714285714e-06,
+      "loss": 1.0939,
+      "step": 42935
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5686171428571427e-06,
+      "loss": 1.398,
+      "step": 42936
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5683485714285717e-06,
+      "loss": 1.229,
+      "step": 42937
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5680800000000003e-06,
+      "loss": 1.0708,
+      "step": 42938
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5678114285714285e-06,
+      "loss": 0.8945,
+      "step": 42939
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.567542857142857e-06,
+      "loss": 1.1703,
+      "step": 42940
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5672742857142857e-06,
+      "loss": 0.7539,
+      "step": 42941
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5670057142857143e-06,
+      "loss": 1.0755,
+      "step": 42942
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.566737142857143e-06,
+      "loss": 0.7587,
+      "step": 42943
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5664685714285715e-06,
+      "loss": 1.1881,
+      "step": 42944
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5662e-06,
+      "loss": 1.152,
+      "step": 42945
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5659314285714287e-06,
+      "loss": 1.009,
+      "step": 42946
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5656628571428573e-06,
+      "loss": 0.8394,
+      "step": 42947
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.565394285714286e-06,
+      "loss": 1.3104,
+      "step": 42948
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.565125714285714e-06,
+      "loss": 1.0376,
+      "step": 42949
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.564857142857143e-06,
+      "loss": 1.0165,
+      "step": 42950
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5645885714285717e-06,
+      "loss": 0.8067,
+      "step": 42951
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5643200000000003e-06,
+      "loss": 1.0463,
+      "step": 42952
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5640514285714285e-06,
+      "loss": 1.0355,
+      "step": 42953
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.563782857142857e-06,
+      "loss": 0.7679,
+      "step": 42954
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5635142857142857e-06,
+      "loss": 1.3066,
+      "step": 42955
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5632457142857143e-06,
+      "loss": 1.4145,
+      "step": 42956
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5629771428571433e-06,
+      "loss": 1.4269,
+      "step": 42957
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5627085714285715e-06,
+      "loss": 1.0364,
+      "step": 42958
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.56244e-06,
+      "loss": 1.0128,
+      "step": 42959
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5621714285714287e-06,
+      "loss": 1.5627,
+      "step": 42960
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5619028571428573e-06,
+      "loss": 1.0625,
+      "step": 42961
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.5616342857142854e-06,
+      "loss": 1.0692,
+      "step": 42962
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5613657142857145e-06,
+      "loss": 1.1209,
+      "step": 42963
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.561097142857143e-06,
+      "loss": 1.0901,
+      "step": 42964
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5608285714285717e-06,
+      "loss": 0.7549,
+      "step": 42965
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5605600000000003e-06,
+      "loss": 1.0079,
+      "step": 42966
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5602914285714284e-06,
+      "loss": 1.1328,
+      "step": 42967
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.560022857142857e-06,
+      "loss": 1.4322,
+      "step": 42968
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5597542857142856e-06,
+      "loss": 1.09,
+      "step": 42969
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5594857142857147e-06,
+      "loss": 1.1731,
+      "step": 42970
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.559217142857143e-06,
+      "loss": 1.0291,
+      "step": 42971
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5589485714285714e-06,
+      "loss": 1.7015,
+      "step": 42972
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.55868e-06,
+      "loss": 0.9248,
+      "step": 42973
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5584114285714286e-06,
+      "loss": 0.7209,
+      "step": 42974
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5581428571428572e-06,
+      "loss": 1.1521,
+      "step": 42975
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.557874285714286e-06,
+      "loss": 0.9664,
+      "step": 42976
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5576057142857144e-06,
+      "loss": 1.0446,
+      "step": 42977
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.557337142857143e-06,
+      "loss": 0.9631,
+      "step": 42978
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5570685714285716e-06,
+      "loss": 0.9999,
+      "step": 42979
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5568000000000002e-06,
+      "loss": 1.0069,
+      "step": 42980
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5565314285714284e-06,
+      "loss": 1.0229,
+      "step": 42981
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.556262857142857e-06,
+      "loss": 1.1645,
+      "step": 42982
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.555994285714286e-06,
+      "loss": 0.8898,
+      "step": 42983
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5557257142857146e-06,
+      "loss": 1.0813,
+      "step": 42984
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.555457142857143e-06,
+      "loss": 1.2965,
+      "step": 42985
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5551885714285714e-06,
+      "loss": 1.0176,
+      "step": 42986
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.55492e-06,
+      "loss": 1.1889,
+      "step": 42987
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5546514285714286e-06,
+      "loss": 1.1373,
+      "step": 42988
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.554382857142857e-06,
+      "loss": 0.7126,
+      "step": 42989
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.554114285714286e-06,
+      "loss": 1.2817,
+      "step": 42990
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5538457142857144e-06,
+      "loss": 1.2376,
+      "step": 42991
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.553577142857143e-06,
+      "loss": 1.0748,
+      "step": 42992
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5533085714285716e-06,
+      "loss": 1.3926,
+      "step": 42993
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5530399999999998e-06,
+      "loss": 1.2987,
+      "step": 42994
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5527714285714284e-06,
+      "loss": 0.761,
+      "step": 42995
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5525028571428574e-06,
+      "loss": 1.2071,
+      "step": 42996
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.552234285714286e-06,
+      "loss": 1.2083,
+      "step": 42997
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5519657142857146e-06,
+      "loss": 0.9094,
+      "step": 42998
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5516971428571428e-06,
+      "loss": 1.3472,
+      "step": 42999
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5514285714285714e-06,
+      "loss": 1.114,
+      "step": 43000
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.55116e-06,
+      "loss": 1.2324,
+      "step": 43001
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5508914285714286e-06,
+      "loss": 0.9805,
+      "step": 43002
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5506228571428576e-06,
+      "loss": 1.0208,
+      "step": 43003
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5503542857142858e-06,
+      "loss": 1.0068,
+      "step": 43004
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5500857142857144e-06,
+      "loss": 1.0279,
+      "step": 43005
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.549817142857143e-06,
+      "loss": 0.9183,
+      "step": 43006
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5495485714285716e-06,
+      "loss": 0.9643,
+      "step": 43007
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5492799999999998e-06,
+      "loss": 1.3654,
+      "step": 43008
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5490114285714288e-06,
+      "loss": 1.2205,
+      "step": 43009
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5487428571428574e-06,
+      "loss": 1.1184,
+      "step": 43010
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.548474285714286e-06,
+      "loss": 1.0295,
+      "step": 43011
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5482057142857146e-06,
+      "loss": 1.2418,
+      "step": 43012
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5479371428571428e-06,
+      "loss": 1.6734,
+      "step": 43013
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5476685714285714e-06,
+      "loss": 1.1753,
+      "step": 43014
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5474e-06,
+      "loss": 1.0744,
+      "step": 43015
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.547131428571429e-06,
+      "loss": 1.0469,
+      "step": 43016
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.546862857142857e-06,
+      "loss": 1.3758,
+      "step": 43017
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5465942857142858e-06,
+      "loss": 1.5419,
+      "step": 43018
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5463257142857144e-06,
+      "loss": 1.31,
+      "step": 43019
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.546057142857143e-06,
+      "loss": 0.8327,
+      "step": 43020
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5457885714285716e-06,
+      "loss": 0.9354,
+      "step": 43021
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.54552e-06,
+      "loss": 0.9433,
+      "step": 43022
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5452514285714288e-06,
+      "loss": 1.253,
+      "step": 43023
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5449828571428574e-06,
+      "loss": 1.256,
+      "step": 43024
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.544714285714286e-06,
+      "loss": 1.5713,
+      "step": 43025
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.544445714285714e-06,
+      "loss": 1.3539,
+      "step": 43026
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5441771428571427e-06,
+      "loss": 1.5476,
+      "step": 43027
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5439085714285713e-06,
+      "loss": 1.4522,
+      "step": 43028
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5436400000000003e-06,
+      "loss": 1.2746,
+      "step": 43029
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.543371428571429e-06,
+      "loss": 0.6973,
+      "step": 43030
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.543102857142857e-06,
+      "loss": 1.1515,
+      "step": 43031
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5428342857142857e-06,
+      "loss": 1.313,
+      "step": 43032
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5425657142857143e-06,
+      "loss": 1.1527,
+      "step": 43033
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.542297142857143e-06,
+      "loss": 0.7208,
+      "step": 43034
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5420285714285715e-06,
+      "loss": 1.6365,
+      "step": 43035
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.54176e-06,
+      "loss": 1.3348,
+      "step": 43036
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5414914285714287e-06,
+      "loss": 0.8981,
+      "step": 43037
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5412228571428573e-06,
+      "loss": 0.8725,
+      "step": 43038
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.540954285714286e-06,
+      "loss": 1.458,
+      "step": 43039
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.540685714285714e-06,
+      "loss": 1.0331,
+      "step": 43040
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5404171428571427e-06,
+      "loss": 0.9639,
+      "step": 43041
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5401485714285717e-06,
+      "loss": 0.7723,
+      "step": 43042
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5398800000000003e-06,
+      "loss": 1.1307,
+      "step": 43043
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.539611428571429e-06,
+      "loss": 0.877,
+      "step": 43044
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.539342857142857e-06,
+      "loss": 1.2436,
+      "step": 43045
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5390742857142857e-06,
+      "loss": 0.6748,
+      "step": 43046
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5388057142857143e-06,
+      "loss": 0.8858,
+      "step": 43047
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.538537142857143e-06,
+      "loss": 1.3415,
+      "step": 43048
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5382685714285715e-06,
+      "loss": 1.2285,
+      "step": 43049
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.538e-06,
+      "loss": 0.833,
+      "step": 43050
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5377314285714287e-06,
+      "loss": 1.0365,
+      "step": 43051
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5374628571428573e-06,
+      "loss": 1.0831,
+      "step": 43052
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.537194285714286e-06,
+      "loss": 1.3621,
+      "step": 43053
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.536925714285714e-06,
+      "loss": 0.7964,
+      "step": 43054
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.536657142857143e-06,
+      "loss": 0.7677,
+      "step": 43055
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5363885714285717e-06,
+      "loss": 1.3042,
+      "step": 43056
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5361200000000003e-06,
+      "loss": 0.8648,
+      "step": 43057
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5358514285714285e-06,
+      "loss": 0.9576,
+      "step": 43058
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.535582857142857e-06,
+      "loss": 1.0318,
+      "step": 43059
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5353142857142857e-06,
+      "loss": 0.9873,
+      "step": 43060
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5350457142857143e-06,
+      "loss": 0.8757,
+      "step": 43061
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5347771428571433e-06,
+      "loss": 0.9918,
+      "step": 43062
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5345085714285715e-06,
+      "loss": 1.4549,
+      "step": 43063
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.53424e-06,
+      "loss": 1.3495,
+      "step": 43064
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5339714285714287e-06,
+      "loss": 1.2168,
+      "step": 43065
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5337028571428573e-06,
+      "loss": 0.7803,
+      "step": 43066
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.533434285714286e-06,
+      "loss": 0.867,
+      "step": 43067
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5331657142857145e-06,
+      "loss": 1.3158,
+      "step": 43068
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.532897142857143e-06,
+      "loss": 1.0857,
+      "step": 43069
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5326285714285717e-06,
+      "loss": 1.182,
+      "step": 43070
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5323600000000003e-06,
+      "loss": 0.9189,
+      "step": 43071
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5320914285714284e-06,
+      "loss": 1.5841,
+      "step": 43072
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.531822857142857e-06,
+      "loss": 1.0831,
+      "step": 43073
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5315542857142856e-06,
+      "loss": 1.1898,
+      "step": 43074
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5312857142857147e-06,
+      "loss": 1.1074,
+      "step": 43075
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5310171428571433e-06,
+      "loss": 0.9527,
+      "step": 43076
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5307485714285714e-06,
+      "loss": 0.867,
+      "step": 43077
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.53048e-06,
+      "loss": 1.2741,
+      "step": 43078
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5302114285714286e-06,
+      "loss": 1.2479,
+      "step": 43079
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5299428571428572e-06,
+      "loss": 1.3352,
+      "step": 43080
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.529674285714286e-06,
+      "loss": 1.2153,
+      "step": 43081
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5294057142857144e-06,
+      "loss": 1.1749,
+      "step": 43082
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.529137142857143e-06,
+      "loss": 1.2192,
+      "step": 43083
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5288685714285716e-06,
+      "loss": 0.9431,
+      "step": 43084
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5286000000000002e-06,
+      "loss": 0.9674,
+      "step": 43085
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5283314285714284e-06,
+      "loss": 1.2125,
+      "step": 43086
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.528062857142857e-06,
+      "loss": 1.041,
+      "step": 43087
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.527794285714286e-06,
+      "loss": 0.7485,
+      "step": 43088
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5275257142857146e-06,
+      "loss": 0.5863,
+      "step": 43089
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5272571428571432e-06,
+      "loss": 1.5025,
+      "step": 43090
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5269885714285714e-06,
+      "loss": 1.0316,
+      "step": 43091
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.52672e-06,
+      "loss": 1.1085,
+      "step": 43092
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5264514285714286e-06,
+      "loss": 1.1963,
+      "step": 43093
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.526182857142857e-06,
+      "loss": 1.2336,
+      "step": 43094
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.525914285714286e-06,
+      "loss": 0.8345,
+      "step": 43095
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5256457142857144e-06,
+      "loss": 0.9128,
+      "step": 43096
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.525377142857143e-06,
+      "loss": 1.2077,
+      "step": 43097
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5251085714285716e-06,
+      "loss": 1.5937,
+      "step": 43098
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.52484e-06,
+      "loss": 0.9861,
+      "step": 43099
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5245714285714284e-06,
+      "loss": 1.3265,
+      "step": 43100
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5243028571428574e-06,
+      "loss": 0.8642,
+      "step": 43101
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.524034285714286e-06,
+      "loss": 1.1329,
+      "step": 43102
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5237657142857146e-06,
+      "loss": 0.8987,
+      "step": 43103
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5234971428571428e-06,
+      "loss": 1.4603,
+      "step": 43104
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5232285714285714e-06,
+      "loss": 1.2003,
+      "step": 43105
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.52296e-06,
+      "loss": 1.1939,
+      "step": 43106
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5226914285714286e-06,
+      "loss": 1.3709,
+      "step": 43107
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5224228571428576e-06,
+      "loss": 1.4492,
+      "step": 43108
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5221542857142858e-06,
+      "loss": 0.8637,
+      "step": 43109
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5218857142857144e-06,
+      "loss": 1.1263,
+      "step": 43110
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.521617142857143e-06,
+      "loss": 1.1565,
+      "step": 43111
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5213485714285716e-06,
+      "loss": 0.6364,
+      "step": 43112
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5210799999999998e-06,
+      "loss": 1.1797,
+      "step": 43113
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5208114285714284e-06,
+      "loss": 1.1048,
+      "step": 43114
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5205428571428574e-06,
+      "loss": 1.1074,
+      "step": 43115
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.520274285714286e-06,
+      "loss": 1.0613,
+      "step": 43116
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5200057142857146e-06,
+      "loss": 1.1888,
+      "step": 43117
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5197371428571428e-06,
+      "loss": 1.5191,
+      "step": 43118
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5194685714285714e-06,
+      "loss": 1.2457,
+      "step": 43119
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5192e-06,
+      "loss": 0.926,
+      "step": 43120
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.518931428571429e-06,
+      "loss": 0.9384,
+      "step": 43121
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5186628571428576e-06,
+      "loss": 1.0043,
+      "step": 43122
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5183942857142858e-06,
+      "loss": 0.9425,
+      "step": 43123
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5181257142857144e-06,
+      "loss": 0.8988,
+      "step": 43124
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.517857142857143e-06,
+      "loss": 0.7483,
+      "step": 43125
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5175885714285716e-06,
+      "loss": 1.1098,
+      "step": 43126
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5173199999999997e-06,
+      "loss": 0.9266,
+      "step": 43127
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5170514285714288e-06,
+      "loss": 0.9899,
+      "step": 43128
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5167828571428573e-06,
+      "loss": 0.8325,
+      "step": 43129
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.516514285714286e-06,
+      "loss": 1.2065,
+      "step": 43130
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5162457142857145e-06,
+      "loss": 1.0873,
+      "step": 43131
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5159771428571427e-06,
+      "loss": 1.2124,
+      "step": 43132
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5157085714285713e-06,
+      "loss": 0.7423,
+      "step": 43133
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5154400000000003e-06,
+      "loss": 0.8293,
+      "step": 43134
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.515171428571429e-06,
+      "loss": 1.2836,
+      "step": 43135
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.514902857142857e-06,
+      "loss": 1.3519,
+      "step": 43136
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5146342857142857e-06,
+      "loss": 1.3371,
+      "step": 43137
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5143657142857143e-06,
+      "loss": 0.9017,
+      "step": 43138
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.514097142857143e-06,
+      "loss": 1.3489,
+      "step": 43139
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5138285714285715e-06,
+      "loss": 1.3654,
+      "step": 43140
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.51356e-06,
+      "loss": 1.1996,
+      "step": 43141
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5132914285714287e-06,
+      "loss": 1.5107,
+      "step": 43142
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5130228571428573e-06,
+      "loss": 0.9483,
+      "step": 43143
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.512754285714286e-06,
+      "loss": 1.2924,
+      "step": 43144
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.512485714285714e-06,
+      "loss": 0.8897,
+      "step": 43145
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5122171428571427e-06,
+      "loss": 1.1607,
+      "step": 43146
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5119485714285713e-06,
+      "loss": 1.474,
+      "step": 43147
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5116800000000003e-06,
+      "loss": 0.8291,
+      "step": 43148
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.511411428571429e-06,
+      "loss": 0.9023,
+      "step": 43149
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.511142857142857e-06,
+      "loss": 1.3011,
+      "step": 43150
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5108742857142857e-06,
+      "loss": 1.1946,
+      "step": 43151
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5106057142857143e-06,
+      "loss": 1.3097,
+      "step": 43152
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.510337142857143e-06,
+      "loss": 0.9965,
+      "step": 43153
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.510068571428572e-06,
+      "loss": 1.2095,
+      "step": 43154
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5098e-06,
+      "loss": 1.11,
+      "step": 43155
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5095314285714287e-06,
+      "loss": 0.9373,
+      "step": 43156
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5092628571428573e-06,
+      "loss": 1.2429,
+      "step": 43157
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.508994285714286e-06,
+      "loss": 1.1808,
+      "step": 43158
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.508725714285714e-06,
+      "loss": 1.2592,
+      "step": 43159
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5084571428571427e-06,
+      "loss": 1.0463,
+      "step": 43160
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5081885714285717e-06,
+      "loss": 1.2455,
+      "step": 43161
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5079200000000003e-06,
+      "loss": 0.6171,
+      "step": 43162
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.507651428571429e-06,
+      "loss": 1.2852,
+      "step": 43163
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.507382857142857e-06,
+      "loss": 1.2023,
+      "step": 43164
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5071142857142857e-06,
+      "loss": 1.1542,
+      "step": 43165
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5068457142857143e-06,
+      "loss": 1.1095,
+      "step": 43166
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.506577142857143e-06,
+      "loss": 0.9226,
+      "step": 43167
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5063085714285715e-06,
+      "loss": 1.1299,
+      "step": 43168
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.50604e-06,
+      "loss": 0.7971,
+      "step": 43169
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5057714285714287e-06,
+      "loss": 1.1868,
+      "step": 43170
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5055028571428573e-06,
+      "loss": 1.0835,
+      "step": 43171
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.505234285714286e-06,
+      "loss": 1.2669,
+      "step": 43172
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.504965714285714e-06,
+      "loss": 1.2872,
+      "step": 43173
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.504697142857143e-06,
+      "loss": 0.9546,
+      "step": 43174
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5044285714285717e-06,
+      "loss": 0.9249,
+      "step": 43175
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5041600000000003e-06,
+      "loss": 0.6072,
+      "step": 43176
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5038914285714284e-06,
+      "loss": 0.6189,
+      "step": 43177
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.503622857142857e-06,
+      "loss": 1.3606,
+      "step": 43178
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5033542857142856e-06,
+      "loss": 1.2301,
+      "step": 43179
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5030857142857142e-06,
+      "loss": 1.1337,
+      "step": 43180
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5028171428571433e-06,
+      "loss": 1.1003,
+      "step": 43181
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5025485714285714e-06,
+      "loss": 0.9039,
+      "step": 43182
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.50228e-06,
+      "loss": 0.9131,
+      "step": 43183
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5020114285714286e-06,
+      "loss": 0.9574,
+      "step": 43184
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5017428571428572e-06,
+      "loss": 0.858,
+      "step": 43185
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.501474285714286e-06,
+      "loss": 0.8398,
+      "step": 43186
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5012057142857144e-06,
+      "loss": 1.5128,
+      "step": 43187
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.500937142857143e-06,
+      "loss": 1.1926,
+      "step": 43188
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5006685714285716e-06,
+      "loss": 1.3307,
+      "step": 43189
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5004000000000002e-06,
+      "loss": 1.0755,
+      "step": 43190
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5001314285714284e-06,
+      "loss": 1.307,
+      "step": 43191
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.499862857142857e-06,
+      "loss": 0.924,
+      "step": 43192
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4995942857142856e-06,
+      "loss": 0.8995,
+      "step": 43193
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4993257142857146e-06,
+      "loss": 0.9017,
+      "step": 43194
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4990571428571432e-06,
+      "loss": 1.2358,
+      "step": 43195
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4987885714285714e-06,
+      "loss": 1.1054,
+      "step": 43196
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.49852e-06,
+      "loss": 1.0071,
+      "step": 43197
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4982514285714286e-06,
+      "loss": 0.7517,
+      "step": 43198
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.497982857142857e-06,
+      "loss": 1.2912,
+      "step": 43199
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.497714285714286e-06,
+      "loss": 1.0238,
+      "step": 43200
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4974457142857144e-06,
+      "loss": 1.3553,
+      "step": 43201
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.497177142857143e-06,
+      "loss": 0.6443,
+      "step": 43202
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4969085714285716e-06,
+      "loss": 0.751,
+      "step": 43203
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.49664e-06,
+      "loss": 1.221,
+      "step": 43204
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4963714285714284e-06,
+      "loss": 0.8416,
+      "step": 43205
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.496102857142857e-06,
+      "loss": 1.064,
+      "step": 43206
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.495834285714286e-06,
+      "loss": 1.5229,
+      "step": 43207
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4955657142857146e-06,
+      "loss": 1.3822,
+      "step": 43208
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.495297142857143e-06,
+      "loss": 0.7635,
+      "step": 43209
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4950285714285714e-06,
+      "loss": 1.1669,
+      "step": 43210
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.49476e-06,
+      "loss": 1.1267,
+      "step": 43211
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4944914285714286e-06,
+      "loss": 1.2034,
+      "step": 43212
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.494222857142857e-06,
+      "loss": 0.8231,
+      "step": 43213
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4939542857142858e-06,
+      "loss": 1.0301,
+      "step": 43214
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4936857142857144e-06,
+      "loss": 0.5088,
+      "step": 43215
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.493417142857143e-06,
+      "loss": 1.2014,
+      "step": 43216
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4931485714285716e-06,
+      "loss": 1.2414,
+      "step": 43217
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.49288e-06,
+      "loss": 0.8709,
+      "step": 43218
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4926114285714284e-06,
+      "loss": 1.2275,
+      "step": 43219
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4923428571428574e-06,
+      "loss": 1.0089,
+      "step": 43220
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.492074285714286e-06,
+      "loss": 1.6391,
+      "step": 43221
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4918057142857146e-06,
+      "loss": 1.13,
+      "step": 43222
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4915371428571428e-06,
+      "loss": 1.1365,
+      "step": 43223
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4912685714285714e-06,
+      "loss": 1.1526,
+      "step": 43224
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.491e-06,
+      "loss": 1.4288,
+      "step": 43225
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4907314285714286e-06,
+      "loss": 1.0155,
+      "step": 43226
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4904628571428576e-06,
+      "loss": 1.2958,
+      "step": 43227
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4901942857142858e-06,
+      "loss": 1.4094,
+      "step": 43228
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4899257142857144e-06,
+      "loss": 1.1858,
+      "step": 43229
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.489657142857143e-06,
+      "loss": 1.2504,
+      "step": 43230
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4893885714285715e-06,
+      "loss": 0.8678,
+      "step": 43231
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4891199999999997e-06,
+      "loss": 1.176,
+      "step": 43232
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4888514285714287e-06,
+      "loss": 1.0619,
+      "step": 43233
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4885828571428573e-06,
+      "loss": 1.0785,
+      "step": 43234
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.488314285714286e-06,
+      "loss": 0.9513,
+      "step": 43235
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4880457142857145e-06,
+      "loss": 1.3847,
+      "step": 43236
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4877771428571427e-06,
+      "loss": 0.7684,
+      "step": 43237
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4875085714285713e-06,
+      "loss": 0.8191,
+      "step": 43238
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.48724e-06,
+      "loss": 1.0283,
+      "step": 43239
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.486971428571429e-06,
+      "loss": 1.2937,
+      "step": 43240
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4867028571428575e-06,
+      "loss": 1.617,
+      "step": 43241
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4864342857142857e-06,
+      "loss": 1.2425,
+      "step": 43242
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4861657142857143e-06,
+      "loss": 0.6428,
+      "step": 43243
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.485897142857143e-06,
+      "loss": 1.0463,
+      "step": 43244
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4856285714285715e-06,
+      "loss": 1.278,
+      "step": 43245
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.48536e-06,
+      "loss": 1.0139,
+      "step": 43246
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4850914285714287e-06,
+      "loss": 1.0153,
+      "step": 43247
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4848228571428573e-06,
+      "loss": 0.9022,
+      "step": 43248
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.484554285714286e-06,
+      "loss": 0.9535,
+      "step": 43249
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4842857142857145e-06,
+      "loss": 1.0715,
+      "step": 43250
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4840171428571427e-06,
+      "loss": 0.5631,
+      "step": 43251
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4837485714285713e-06,
+      "loss": 1.1199,
+      "step": 43252
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4834800000000003e-06,
+      "loss": 0.6189,
+      "step": 43253
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.483211428571429e-06,
+      "loss": 1.1899,
+      "step": 43254
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.482942857142857e-06,
+      "loss": 0.9272,
+      "step": 43255
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4826742857142857e-06,
+      "loss": 1.444,
+      "step": 43256
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4824057142857143e-06,
+      "loss": 1.1022,
+      "step": 43257
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.482137142857143e-06,
+      "loss": 1.4747,
+      "step": 43258
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4818685714285715e-06,
+      "loss": 1.1651,
+      "step": 43259
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4816e-06,
+      "loss": 1.1707,
+      "step": 43260
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4813314285714287e-06,
+      "loss": 1.2032,
+      "step": 43261
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4810628571428573e-06,
+      "loss": 1.1523,
+      "step": 43262
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.480794285714286e-06,
+      "loss": 1.4224,
+      "step": 43263
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.480525714285714e-06,
+      "loss": 1.3987,
+      "step": 43264
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4802571428571427e-06,
+      "loss": 1.2406,
+      "step": 43265
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4799885714285717e-06,
+      "loss": 0.7939,
+      "step": 43266
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4797200000000003e-06,
+      "loss": 1.1366,
+      "step": 43267
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.479451428571429e-06,
+      "loss": 1.2323,
+      "step": 43268
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.479182857142857e-06,
+      "loss": 1.0521,
+      "step": 43269
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4789142857142857e-06,
+      "loss": 1.1588,
+      "step": 43270
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4786457142857143e-06,
+      "loss": 0.8312,
+      "step": 43271
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.478377142857143e-06,
+      "loss": 1.1495,
+      "step": 43272
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.478108571428572e-06,
+      "loss": 1.4141,
+      "step": 43273
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.47784e-06,
+      "loss": 1.4549,
+      "step": 43274
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4775714285714287e-06,
+      "loss": 1.0331,
+      "step": 43275
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4773028571428573e-06,
+      "loss": 1.3333,
+      "step": 43276
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.477034285714286e-06,
+      "loss": 1.569,
+      "step": 43277
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.476765714285714e-06,
+      "loss": 1.2605,
+      "step": 43278
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.476497142857143e-06,
+      "loss": 1.4767,
+      "step": 43279
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4762285714285717e-06,
+      "loss": 1.0822,
+      "step": 43280
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4759600000000003e-06,
+      "loss": 1.1522,
+      "step": 43281
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.475691428571429e-06,
+      "loss": 1.08,
+      "step": 43282
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.475422857142857e-06,
+      "loss": 0.9716,
+      "step": 43283
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4751542857142856e-06,
+      "loss": 1.399,
+      "step": 43284
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4748857142857142e-06,
+      "loss": 1.3828,
+      "step": 43285
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4746171428571433e-06,
+      "loss": 1.2928,
+      "step": 43286
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4743485714285714e-06,
+      "loss": 0.9619,
+      "step": 43287
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.47408e-06,
+      "loss": 0.7353,
+      "step": 43288
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4738114285714286e-06,
+      "loss": 0.9685,
+      "step": 43289
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4735428571428572e-06,
+      "loss": 1.5647,
+      "step": 43290
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.473274285714286e-06,
+      "loss": 1.3155,
+      "step": 43291
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4730057142857144e-06,
+      "loss": 1.0834,
+      "step": 43292
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.472737142857143e-06,
+      "loss": 1.3039,
+      "step": 43293
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4724685714285716e-06,
+      "loss": 1.4024,
+      "step": 43294
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4722000000000002e-06,
+      "loss": 1.3365,
+      "step": 43295
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.471931428571429e-06,
+      "loss": 1.1686,
+      "step": 43296
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.471662857142857e-06,
+      "loss": 1.065,
+      "step": 43297
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4713942857142856e-06,
+      "loss": 1.1475,
+      "step": 43298
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4711257142857146e-06,
+      "loss": 1.2015,
+      "step": 43299
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4708571428571432e-06,
+      "loss": 1.1694,
+      "step": 43300
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4705885714285714e-06,
+      "loss": 1.251,
+      "step": 43301
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.47032e-06,
+      "loss": 1.0226,
+      "step": 43302
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4700514285714286e-06,
+      "loss": 1.0192,
+      "step": 43303
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.469782857142857e-06,
+      "loss": 1.2159,
+      "step": 43304
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.469514285714286e-06,
+      "loss": 1.2977,
+      "step": 43305
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4692457142857144e-06,
+      "loss": 1.1958,
+      "step": 43306
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.468977142857143e-06,
+      "loss": 1.2794,
+      "step": 43307
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4687085714285716e-06,
+      "loss": 1.3201,
+      "step": 43308
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.46844e-06,
+      "loss": 1.2922,
+      "step": 43309
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.4681714285714284e-06,
+      "loss": 1.0662,
+      "step": 43310
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.467902857142857e-06,
+      "loss": 1.4729,
+      "step": 43311
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.467634285714286e-06,
+      "loss": 1.176,
+      "step": 43312
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4673657142857146e-06,
+      "loss": 1.2904,
+      "step": 43313
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.467097142857143e-06,
+      "loss": 0.6526,
+      "step": 43314
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4668285714285714e-06,
+      "loss": 0.8875,
+      "step": 43315
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.46656e-06,
+      "loss": 1.064,
+      "step": 43316
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4662914285714286e-06,
+      "loss": 1.294,
+      "step": 43317
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.466022857142857e-06,
+      "loss": 0.8511,
+      "step": 43318
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4657542857142858e-06,
+      "loss": 0.9592,
+      "step": 43319
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4654857142857144e-06,
+      "loss": 1.0005,
+      "step": 43320
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.465217142857143e-06,
+      "loss": 0.9647,
+      "step": 43321
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4649485714285716e-06,
+      "loss": 1.2121,
+      "step": 43322
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.46468e-06,
+      "loss": 1.4307,
+      "step": 43323
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4644114285714284e-06,
+      "loss": 1.4463,
+      "step": 43324
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4641428571428574e-06,
+      "loss": 0.9294,
+      "step": 43325
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.463874285714286e-06,
+      "loss": 1.3184,
+      "step": 43326
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4636057142857146e-06,
+      "loss": 1.1178,
+      "step": 43327
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.463337142857143e-06,
+      "loss": 0.9514,
+      "step": 43328
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4630685714285714e-06,
+      "loss": 1.3525,
+      "step": 43329
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4628e-06,
+      "loss": 1.0397,
+      "step": 43330
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4625314285714286e-06,
+      "loss": 1.2268,
+      "step": 43331
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4622628571428576e-06,
+      "loss": 1.2534,
+      "step": 43332
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4619942857142857e-06,
+      "loss": 0.8684,
+      "step": 43333
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4617257142857143e-06,
+      "loss": 1.0802,
+      "step": 43334
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.461457142857143e-06,
+      "loss": 1.2128,
+      "step": 43335
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4611885714285715e-06,
+      "loss": 1.3957,
+      "step": 43336
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.46092e-06,
+      "loss": 1.0932,
+      "step": 43337
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4606514285714283e-06,
+      "loss": 1.1128,
+      "step": 43338
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4603828571428573e-06,
+      "loss": 1.2462,
+      "step": 43339
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.460114285714286e-06,
+      "loss": 1.4348,
+      "step": 43340
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4598457142857145e-06,
+      "loss": 0.9748,
+      "step": 43341
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4595771428571427e-06,
+      "loss": 1.0,
+      "step": 43342
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4593085714285713e-06,
+      "loss": 1.1051,
+      "step": 43343
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.45904e-06,
+      "loss": 0.5565,
+      "step": 43344
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.458771428571429e-06,
+      "loss": 1.279,
+      "step": 43345
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4585028571428575e-06,
+      "loss": 1.0657,
+      "step": 43346
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4582342857142857e-06,
+      "loss": 0.7294,
+      "step": 43347
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4579657142857143e-06,
+      "loss": 1.2209,
+      "step": 43348
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.457697142857143e-06,
+      "loss": 0.9682,
+      "step": 43349
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4574285714285715e-06,
+      "loss": 1.14,
+      "step": 43350
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4571599999999997e-06,
+      "loss": 1.2199,
+      "step": 43351
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4568914285714287e-06,
+      "loss": 0.9872,
+      "step": 43352
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4566228571428573e-06,
+      "loss": 1.1474,
+      "step": 43353
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.456354285714286e-06,
+      "loss": 1.1707,
+      "step": 43354
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4560857142857145e-06,
+      "loss": 0.9513,
+      "step": 43355
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4558171428571427e-06,
+      "loss": 0.9779,
+      "step": 43356
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4555485714285713e-06,
+      "loss": 0.8171,
+      "step": 43357
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4552800000000003e-06,
+      "loss": 1.3516,
+      "step": 43358
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.455011428571429e-06,
+      "loss": 0.7727,
+      "step": 43359
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4547428571428575e-06,
+      "loss": 1.4862,
+      "step": 43360
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4544742857142857e-06,
+      "loss": 1.0315,
+      "step": 43361
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4542057142857143e-06,
+      "loss": 1.046,
+      "step": 43362
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.453937142857143e-06,
+      "loss": 1.5238,
+      "step": 43363
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4536685714285715e-06,
+      "loss": 1.0392,
+      "step": 43364
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4534e-06,
+      "loss": 1.051,
+      "step": 43365
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4531314285714287e-06,
+      "loss": 1.6321,
+      "step": 43366
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4528628571428573e-06,
+      "loss": 1.2301,
+      "step": 43367
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.452594285714286e-06,
+      "loss": 0.7302,
+      "step": 43368
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4523257142857145e-06,
+      "loss": 1.2532,
+      "step": 43369
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4520571428571427e-06,
+      "loss": 1.143,
+      "step": 43370
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4517885714285713e-06,
+      "loss": 1.2646,
+      "step": 43371
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4515200000000003e-06,
+      "loss": 0.6086,
+      "step": 43372
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.451251428571429e-06,
+      "loss": 1.0905,
+      "step": 43373
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.450982857142857e-06,
+      "loss": 1.1494,
+      "step": 43374
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4507142857142857e-06,
+      "loss": 1.186,
+      "step": 43375
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4504457142857143e-06,
+      "loss": 1.1505,
+      "step": 43376
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.450177142857143e-06,
+      "loss": 1.0738,
+      "step": 43377
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.449908571428572e-06,
+      "loss": 1.1718,
+      "step": 43378
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.44964e-06,
+      "loss": 1.2758,
+      "step": 43379
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4493714285714287e-06,
+      "loss": 1.0317,
+      "step": 43380
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4491028571428573e-06,
+      "loss": 1.2051,
+      "step": 43381
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.448834285714286e-06,
+      "loss": 1.1803,
+      "step": 43382
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4485657142857145e-06,
+      "loss": 1.3161,
+      "step": 43383
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4482971428571426e-06,
+      "loss": 1.0342,
+      "step": 43384
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4480285714285717e-06,
+      "loss": 0.8203,
+      "step": 43385
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4477600000000003e-06,
+      "loss": 0.842,
+      "step": 43386
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.447491428571429e-06,
+      "loss": 1.3027,
+      "step": 43387
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.447222857142857e-06,
+      "loss": 1.1513,
+      "step": 43388
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4469542857142856e-06,
+      "loss": 1.0612,
+      "step": 43389
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4466857142857142e-06,
+      "loss": 1.1082,
+      "step": 43390
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.446417142857143e-06,
+      "loss": 1.208,
+      "step": 43391
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.446148571428572e-06,
+      "loss": 1.4007,
+      "step": 43392
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.44588e-06,
+      "loss": 1.1115,
+      "step": 43393
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4456114285714286e-06,
+      "loss": 1.0843,
+      "step": 43394
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4453428571428572e-06,
+      "loss": 0.9224,
+      "step": 43395
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.445074285714286e-06,
+      "loss": 1.7041,
+      "step": 43396
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.444805714285714e-06,
+      "loss": 0.6457,
+      "step": 43397
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.444537142857143e-06,
+      "loss": 1.0896,
+      "step": 43398
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4442685714285716e-06,
+      "loss": 0.9788,
+      "step": 43399
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4440000000000002e-06,
+      "loss": 1.0033,
+      "step": 43400
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.443731428571429e-06,
+      "loss": 0.8879,
+      "step": 43401
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.443462857142857e-06,
+      "loss": 1.7164,
+      "step": 43402
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4431942857142856e-06,
+      "loss": 1.0086,
+      "step": 43403
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.442925714285714e-06,
+      "loss": 0.887,
+      "step": 43404
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4426571428571432e-06,
+      "loss": 0.7716,
+      "step": 43405
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4423885714285714e-06,
+      "loss": 1.0244,
+      "step": 43406
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.44212e-06,
+      "loss": 0.838,
+      "step": 43407
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4418514285714286e-06,
+      "loss": 1.001,
+      "step": 43408
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.441582857142857e-06,
+      "loss": 1.0458,
+      "step": 43409
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.441314285714286e-06,
+      "loss": 0.902,
+      "step": 43410
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4410457142857144e-06,
+      "loss": 0.9931,
+      "step": 43411
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.440777142857143e-06,
+      "loss": 1.1311,
+      "step": 43412
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4405085714285716e-06,
+      "loss": 0.814,
+      "step": 43413
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.44024e-06,
+      "loss": 1.0033,
+      "step": 43414
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.439971428571429e-06,
+      "loss": 1.1122,
+      "step": 43415
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.439702857142857e-06,
+      "loss": 1.1097,
+      "step": 43416
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4394342857142856e-06,
+      "loss": 1.0423,
+      "step": 43417
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4391657142857146e-06,
+      "loss": 1.4723,
+      "step": 43418
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.438897142857143e-06,
+      "loss": 0.9038,
+      "step": 43419
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4386285714285714e-06,
+      "loss": 1.2061,
+      "step": 43420
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.43836e-06,
+      "loss": 1.5239,
+      "step": 43421
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4380914285714286e-06,
+      "loss": 0.9504,
+      "step": 43422
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.437822857142857e-06,
+      "loss": 1.0684,
+      "step": 43423
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4375542857142858e-06,
+      "loss": 1.076,
+      "step": 43424
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4372857142857144e-06,
+      "loss": 0.9962,
+      "step": 43425
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.437017142857143e-06,
+      "loss": 0.9587,
+      "step": 43426
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4367485714285716e-06,
+      "loss": 1.2973,
+      "step": 43427
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.43648e-06,
+      "loss": 0.7848,
+      "step": 43428
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4362114285714284e-06,
+      "loss": 1.2317,
+      "step": 43429
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.435942857142857e-06,
+      "loss": 1.2345,
+      "step": 43430
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.435674285714286e-06,
+      "loss": 1.2162,
+      "step": 43431
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4354057142857146e-06,
+      "loss": 1.4551,
+      "step": 43432
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.435137142857143e-06,
+      "loss": 0.9724,
+      "step": 43433
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4348685714285714e-06,
+      "loss": 0.8861,
+      "step": 43434
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4346e-06,
+      "loss": 0.7177,
+      "step": 43435
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4343314285714285e-06,
+      "loss": 0.9503,
+      "step": 43436
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.434062857142857e-06,
+      "loss": 0.6892,
+      "step": 43437
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4337942857142857e-06,
+      "loss": 0.9745,
+      "step": 43438
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4335257142857143e-06,
+      "loss": 1.0444,
+      "step": 43439
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.433257142857143e-06,
+      "loss": 1.0874,
+      "step": 43440
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4329885714285715e-06,
+      "loss": 0.9881,
+      "step": 43441
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.43272e-06,
+      "loss": 1.4379,
+      "step": 43442
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4324514285714283e-06,
+      "loss": 1.3484,
+      "step": 43443
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4321828571428573e-06,
+      "loss": 1.7098,
+      "step": 43444
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.431914285714286e-06,
+      "loss": 1.6104,
+      "step": 43445
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4316457142857145e-06,
+      "loss": 1.2355,
+      "step": 43446
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.431377142857143e-06,
+      "loss": 0.6725,
+      "step": 43447
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4311085714285713e-06,
+      "loss": 1.1546,
+      "step": 43448
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.43084e-06,
+      "loss": 0.9564,
+      "step": 43449
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4305714285714285e-06,
+      "loss": 1.4049,
+      "step": 43450
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4303028571428575e-06,
+      "loss": 1.2554,
+      "step": 43451
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4300342857142857e-06,
+      "loss": 1.3597,
+      "step": 43452
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4297657142857143e-06,
+      "loss": 1.2321,
+      "step": 43453
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.429497142857143e-06,
+      "loss": 1.0739,
+      "step": 43454
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4292285714285715e-06,
+      "loss": 1.4607,
+      "step": 43455
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.42896e-06,
+      "loss": 1.102,
+      "step": 43456
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4286914285714287e-06,
+      "loss": 1.0121,
+      "step": 43457
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4284228571428573e-06,
+      "loss": 1.0582,
+      "step": 43458
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.428154285714286e-06,
+      "loss": 1.3906,
+      "step": 43459
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4278857142857145e-06,
+      "loss": 1.1935,
+      "step": 43460
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4276171428571427e-06,
+      "loss": 0.9142,
+      "step": 43461
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4273485714285713e-06,
+      "loss": 0.9239,
+      "step": 43462
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.42708e-06,
+      "loss": 0.8648,
+      "step": 43463
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.426811428571429e-06,
+      "loss": 1.2363,
+      "step": 43464
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4265428571428575e-06,
+      "loss": 1.0929,
+      "step": 43465
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4262742857142857e-06,
+      "loss": 0.7383,
+      "step": 43466
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4260057142857143e-06,
+      "loss": 1.3357,
+      "step": 43467
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.425737142857143e-06,
+      "loss": 0.9433,
+      "step": 43468
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4254685714285715e-06,
+      "loss": 1.3148,
+      "step": 43469
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4252e-06,
+      "loss": 1.0712,
+      "step": 43470
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4249314285714287e-06,
+      "loss": 1.2228,
+      "step": 43471
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4246628571428573e-06,
+      "loss": 0.6023,
+      "step": 43472
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.424394285714286e-06,
+      "loss": 1.4307,
+      "step": 43473
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4241257142857145e-06,
+      "loss": 0.9536,
+      "step": 43474
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4238571428571427e-06,
+      "loss": 1.6911,
+      "step": 43475
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4235885714285713e-06,
+      "loss": 1.22,
+      "step": 43476
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4233200000000003e-06,
+      "loss": 1.4034,
+      "step": 43477
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.423051428571429e-06,
+      "loss": 1.0274,
+      "step": 43478
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4227828571428575e-06,
+      "loss": 1.0873,
+      "step": 43479
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4225142857142857e-06,
+      "loss": 0.9342,
+      "step": 43480
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4222457142857143e-06,
+      "loss": 1.0132,
+      "step": 43481
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.421977142857143e-06,
+      "loss": 1.1171,
+      "step": 43482
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4217085714285715e-06,
+      "loss": 1.0743,
+      "step": 43483
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.42144e-06,
+      "loss": 1.3851,
+      "step": 43484
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4211714285714287e-06,
+      "loss": 1.1049,
+      "step": 43485
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.4209028571428573e-06,
+      "loss": 1.237,
+      "step": 43486
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.420634285714286e-06,
+      "loss": 1.1934,
+      "step": 43487
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4203657142857145e-06,
+      "loss": 1.0949,
+      "step": 43488
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4200971428571426e-06,
+      "loss": 1.0272,
+      "step": 43489
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4198285714285717e-06,
+      "loss": 1.2455,
+      "step": 43490
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4195600000000003e-06,
+      "loss": 1.0791,
+      "step": 43491
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.419291428571429e-06,
+      "loss": 1.0968,
+      "step": 43492
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.419022857142857e-06,
+      "loss": 0.8364,
+      "step": 43493
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4187542857142856e-06,
+      "loss": 1.0476,
+      "step": 43494
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4184857142857142e-06,
+      "loss": 0.8534,
+      "step": 43495
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.418217142857143e-06,
+      "loss": 1.1597,
+      "step": 43496
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.417948571428572e-06,
+      "loss": 1.3083,
+      "step": 43497
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.41768e-06,
+      "loss": 1.0671,
+      "step": 43498
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4174114285714286e-06,
+      "loss": 1.4071,
+      "step": 43499
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4171428571428572e-06,
+      "loss": 1.2935,
+      "step": 43500
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.416874285714286e-06,
+      "loss": 1.1264,
+      "step": 43501
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4166057142857144e-06,
+      "loss": 1.1795,
+      "step": 43502
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.416337142857143e-06,
+      "loss": 1.4375,
+      "step": 43503
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4160685714285716e-06,
+      "loss": 0.9957,
+      "step": 43504
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4158000000000002e-06,
+      "loss": 1.4052,
+      "step": 43505
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.415531428571429e-06,
+      "loss": 0.9416,
+      "step": 43506
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.415262857142857e-06,
+      "loss": 1.3787,
+      "step": 43507
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4149942857142856e-06,
+      "loss": 1.0663,
+      "step": 43508
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.414725714285714e-06,
+      "loss": 1.0035,
+      "step": 43509
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4144571428571432e-06,
+      "loss": 1.2757,
+      "step": 43510
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.414188571428572e-06,
+      "loss": 0.9916,
+      "step": 43511
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.41392e-06,
+      "loss": 0.921,
+      "step": 43512
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4136514285714286e-06,
+      "loss": 1.2323,
+      "step": 43513
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.413382857142857e-06,
+      "loss": 0.9247,
+      "step": 43514
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.413114285714286e-06,
+      "loss": 0.7187,
+      "step": 43515
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4128457142857144e-06,
+      "loss": 0.5342,
+      "step": 43516
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.412577142857143e-06,
+      "loss": 1.123,
+      "step": 43517
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4123085714285716e-06,
+      "loss": 1.0008,
+      "step": 43518
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.41204e-06,
+      "loss": 1.1411,
+      "step": 43519
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.411771428571429e-06,
+      "loss": 1.6929,
+      "step": 43520
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.411502857142857e-06,
+      "loss": 1.0365,
+      "step": 43521
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4112342857142856e-06,
+      "loss": 1.2564,
+      "step": 43522
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4109657142857146e-06,
+      "loss": 1.1834,
+      "step": 43523
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.410697142857143e-06,
+      "loss": 1.6138,
+      "step": 43524
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4104285714285714e-06,
+      "loss": 1.5441,
+      "step": 43525
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.41016e-06,
+      "loss": 1.1313,
+      "step": 43526
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4098914285714286e-06,
+      "loss": 1.2441,
+      "step": 43527
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.409622857142857e-06,
+      "loss": 1.1776,
+      "step": 43528
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4093542857142858e-06,
+      "loss": 0.9774,
+      "step": 43529
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4090857142857144e-06,
+      "loss": 0.5002,
+      "step": 43530
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.408817142857143e-06,
+      "loss": 1.4838,
+      "step": 43531
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4085485714285716e-06,
+      "loss": 1.1248,
+      "step": 43532
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.40828e-06,
+      "loss": 1.3394,
+      "step": 43533
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4080114285714288e-06,
+      "loss": 1.0323,
+      "step": 43534
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.407742857142857e-06,
+      "loss": 1.1487,
+      "step": 43535
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.407474285714286e-06,
+      "loss": 1.2877,
+      "step": 43536
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4072057142857146e-06,
+      "loss": 1.2136,
+      "step": 43537
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.406937142857143e-06,
+      "loss": 0.8951,
+      "step": 43538
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4066685714285713e-06,
+      "loss": 0.7835,
+      "step": 43539
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4064e-06,
+      "loss": 1.3677,
+      "step": 43540
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4061314285714285e-06,
+      "loss": 1.3865,
+      "step": 43541
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.405862857142857e-06,
+      "loss": 0.8953,
+      "step": 43542
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.405594285714286e-06,
+      "loss": 1.351,
+      "step": 43543
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4053257142857143e-06,
+      "loss": 1.004,
+      "step": 43544
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.405057142857143e-06,
+      "loss": 1.6323,
+      "step": 43545
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4047885714285715e-06,
+      "loss": 1.0593,
+      "step": 43546
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.40452e-06,
+      "loss": 1.3619,
+      "step": 43547
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4042514285714283e-06,
+      "loss": 0.9545,
+      "step": 43548
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4039828571428573e-06,
+      "loss": 0.6931,
+      "step": 43549
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.403714285714286e-06,
+      "loss": 0.9293,
+      "step": 43550
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4034457142857145e-06,
+      "loss": 1.4579,
+      "step": 43551
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.403177142857143e-06,
+      "loss": 0.952,
+      "step": 43552
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4029085714285713e-06,
+      "loss": 0.9977,
+      "step": 43553
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.40264e-06,
+      "loss": 1.426,
+      "step": 43554
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4023714285714285e-06,
+      "loss": 0.9176,
+      "step": 43555
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4021028571428575e-06,
+      "loss": 0.8133,
+      "step": 43556
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4018342857142857e-06,
+      "loss": 1.2102,
+      "step": 43557
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4015657142857143e-06,
+      "loss": 0.8218,
+      "step": 43558
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.401297142857143e-06,
+      "loss": 1.1804,
+      "step": 43559
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4010285714285715e-06,
+      "loss": 0.8835,
+      "step": 43560
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.40076e-06,
+      "loss": 1.1816,
+      "step": 43561
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4004914285714283e-06,
+      "loss": 1.0297,
+      "step": 43562
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.4002228571428573e-06,
+      "loss": 1.0108,
+      "step": 43563
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.399954285714286e-06,
+      "loss": 1.2858,
+      "step": 43564
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3996857142857145e-06,
+      "loss": 1.2975,
+      "step": 43565
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.399417142857143e-06,
+      "loss": 1.1461,
+      "step": 43566
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3991485714285713e-06,
+      "loss": 1.3184,
+      "step": 43567
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.39888e-06,
+      "loss": 1.2291,
+      "step": 43568
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.398611428571429e-06,
+      "loss": 1.0402,
+      "step": 43569
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3983428571428575e-06,
+      "loss": 1.3335,
+      "step": 43570
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3980742857142857e-06,
+      "loss": 1.0642,
+      "step": 43571
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3978057142857143e-06,
+      "loss": 1.1395,
+      "step": 43572
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.397537142857143e-06,
+      "loss": 1.3709,
+      "step": 43573
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3972685714285715e-06,
+      "loss": 1.1412,
+      "step": 43574
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.397e-06,
+      "loss": 1.299,
+      "step": 43575
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3967314285714287e-06,
+      "loss": 0.8241,
+      "step": 43576
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3964628571428573e-06,
+      "loss": 1.3347,
+      "step": 43577
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.396194285714286e-06,
+      "loss": 1.231,
+      "step": 43578
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3959257142857145e-06,
+      "loss": 1.1844,
+      "step": 43579
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3956571428571427e-06,
+      "loss": 1.166,
+      "step": 43580
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3953885714285713e-06,
+      "loss": 0.9946,
+      "step": 43581
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3951200000000003e-06,
+      "loss": 1.2201,
+      "step": 43582
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.394851428571429e-06,
+      "loss": 1.3103,
+      "step": 43583
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3945828571428575e-06,
+      "loss": 1.1659,
+      "step": 43584
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3943142857142857e-06,
+      "loss": 1.103,
+      "step": 43585
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3940457142857143e-06,
+      "loss": 1.662,
+      "step": 43586
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.393777142857143e-06,
+      "loss": 1.0195,
+      "step": 43587
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3935085714285715e-06,
+      "loss": 0.751,
+      "step": 43588
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3932400000000005e-06,
+      "loss": 0.9706,
+      "step": 43589
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3929714285714287e-06,
+      "loss": 1.2549,
+      "step": 43590
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3927028571428573e-06,
+      "loss": 0.8582,
+      "step": 43591
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.392434285714286e-06,
+      "loss": 1.091,
+      "step": 43592
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3921657142857145e-06,
+      "loss": 1.1542,
+      "step": 43593
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3918971428571426e-06,
+      "loss": 1.0191,
+      "step": 43594
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3916285714285712e-06,
+      "loss": 1.2278,
+      "step": 43595
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3913600000000003e-06,
+      "loss": 1.5045,
+      "step": 43596
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.391091428571429e-06,
+      "loss": 1.2593,
+      "step": 43597
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3908228571428575e-06,
+      "loss": 1.4403,
+      "step": 43598
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3905542857142856e-06,
+      "loss": 1.065,
+      "step": 43599
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3902857142857142e-06,
+      "loss": 1.0658,
+      "step": 43600
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.390017142857143e-06,
+      "loss": 1.2857,
+      "step": 43601
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.389748571428572e-06,
+      "loss": 1.1517,
+      "step": 43602
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.38948e-06,
+      "loss": 0.7616,
+      "step": 43603
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3892114285714286e-06,
+      "loss": 1.2086,
+      "step": 43604
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3889428571428572e-06,
+      "loss": 1.007,
+      "step": 43605
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.388674285714286e-06,
+      "loss": 1.09,
+      "step": 43606
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3884057142857144e-06,
+      "loss": 0.9107,
+      "step": 43607
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3881371428571426e-06,
+      "loss": 1.4663,
+      "step": 43608
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3878685714285716e-06,
+      "loss": 1.2244,
+      "step": 43609
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3876000000000002e-06,
+      "loss": 0.9069,
+      "step": 43610
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.387331428571429e-06,
+      "loss": 1.2169,
+      "step": 43611
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.387062857142857e-06,
+      "loss": 1.4517,
+      "step": 43612
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3867942857142856e-06,
+      "loss": 1.1533,
+      "step": 43613
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.386525714285714e-06,
+      "loss": 0.7599,
+      "step": 43614
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.386257142857143e-06,
+      "loss": 1.0412,
+      "step": 43615
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.385988571428572e-06,
+      "loss": 1.3745,
+      "step": 43616
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.38572e-06,
+      "loss": 1.4155,
+      "step": 43617
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3854514285714286e-06,
+      "loss": 1.0312,
+      "step": 43618
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.385182857142857e-06,
+      "loss": 1.4096,
+      "step": 43619
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.384914285714286e-06,
+      "loss": 1.2628,
+      "step": 43620
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3846457142857144e-06,
+      "loss": 1.0832,
+      "step": 43621
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.384377142857143e-06,
+      "loss": 0.7383,
+      "step": 43622
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3841085714285716e-06,
+      "loss": 1.0465,
+      "step": 43623
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.38384e-06,
+      "loss": 1.2959,
+      "step": 43624
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.383571428571429e-06,
+      "loss": 1.2782,
+      "step": 43625
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.383302857142857e-06,
+      "loss": 1.2157,
+      "step": 43626
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3830342857142856e-06,
+      "loss": 1.1485,
+      "step": 43627
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.382765714285714e-06,
+      "loss": 0.9678,
+      "step": 43628
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.382497142857143e-06,
+      "loss": 1.1738,
+      "step": 43629
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.382228571428572e-06,
+      "loss": 1.2342,
+      "step": 43630
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.38196e-06,
+      "loss": 0.8937,
+      "step": 43631
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3816914285714286e-06,
+      "loss": 1.225,
+      "step": 43632
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.381422857142857e-06,
+      "loss": 1.3585,
+      "step": 43633
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3811542857142858e-06,
+      "loss": 0.9498,
+      "step": 43634
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3808857142857144e-06,
+      "loss": 1.1137,
+      "step": 43635
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.380617142857143e-06,
+      "loss": 1.3254,
+      "step": 43636
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3803485714285716e-06,
+      "loss": 1.0821,
+      "step": 43637
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.38008e-06,
+      "loss": 1.6027,
+      "step": 43638
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3798114285714288e-06,
+      "loss": 1.0701,
+      "step": 43639
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.379542857142857e-06,
+      "loss": 1.3476,
+      "step": 43640
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3792742857142855e-06,
+      "loss": 1.0825,
+      "step": 43641
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3790057142857146e-06,
+      "loss": 1.1055,
+      "step": 43642
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.378737142857143e-06,
+      "loss": 1.4328,
+      "step": 43643
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3784685714285713e-06,
+      "loss": 0.7736,
+      "step": 43644
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3782e-06,
+      "loss": 1.384,
+      "step": 43645
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3779314285714285e-06,
+      "loss": 0.9172,
+      "step": 43646
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.377662857142857e-06,
+      "loss": 1.4169,
+      "step": 43647
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3773942857142857e-06,
+      "loss": 1.1145,
+      "step": 43648
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3771257142857143e-06,
+      "loss": 0.8367,
+      "step": 43649
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.376857142857143e-06,
+      "loss": 1.259,
+      "step": 43650
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3765885714285715e-06,
+      "loss": 1.3223,
+      "step": 43651
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.37632e-06,
+      "loss": 1.3976,
+      "step": 43652
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3760514285714287e-06,
+      "loss": 1.1498,
+      "step": 43653
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.375782857142857e-06,
+      "loss": 1.2875,
+      "step": 43654
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.375514285714286e-06,
+      "loss": 1.3933,
+      "step": 43655
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3752457142857145e-06,
+      "loss": 1.2265,
+      "step": 43656
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.374977142857143e-06,
+      "loss": 1.171,
+      "step": 43657
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3747085714285713e-06,
+      "loss": 1.2789,
+      "step": 43658
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.37444e-06,
+      "loss": 1.0478,
+      "step": 43659
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.3741714285714285e-06,
+      "loss": 0.9453,
+      "step": 43660
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.373902857142857e-06,
+      "loss": 0.8956,
+      "step": 43661
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.373634285714286e-06,
+      "loss": 1.3286,
+      "step": 43662
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3733657142857143e-06,
+      "loss": 0.9809,
+      "step": 43663
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.373097142857143e-06,
+      "loss": 1.119,
+      "step": 43664
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3728285714285715e-06,
+      "loss": 1.1955,
+      "step": 43665
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.37256e-06,
+      "loss": 1.5546,
+      "step": 43666
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3722914285714283e-06,
+      "loss": 1.1468,
+      "step": 43667
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3720228571428573e-06,
+      "loss": 1.5805,
+      "step": 43668
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.371754285714286e-06,
+      "loss": 1.1487,
+      "step": 43669
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3714857142857145e-06,
+      "loss": 1.3142,
+      "step": 43670
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.371217142857143e-06,
+      "loss": 1.3604,
+      "step": 43671
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3709485714285713e-06,
+      "loss": 1.2562,
+      "step": 43672
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.37068e-06,
+      "loss": 0.8903,
+      "step": 43673
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3704114285714285e-06,
+      "loss": 1.1058,
+      "step": 43674
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3701428571428575e-06,
+      "loss": 1.0729,
+      "step": 43675
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3698742857142857e-06,
+      "loss": 1.0174,
+      "step": 43676
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3696057142857143e-06,
+      "loss": 0.957,
+      "step": 43677
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.369337142857143e-06,
+      "loss": 1.1533,
+      "step": 43678
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3690685714285715e-06,
+      "loss": 1.198,
+      "step": 43679
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3688e-06,
+      "loss": 1.0111,
+      "step": 43680
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3685314285714287e-06,
+      "loss": 1.1842,
+      "step": 43681
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3682628571428573e-06,
+      "loss": 0.8763,
+      "step": 43682
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.367994285714286e-06,
+      "loss": 1.1253,
+      "step": 43683
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3677257142857145e-06,
+      "loss": 1.0239,
+      "step": 43684
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.367457142857143e-06,
+      "loss": 1.1101,
+      "step": 43685
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3671885714285713e-06,
+      "loss": 0.7657,
+      "step": 43686
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.36692e-06,
+      "loss": 1.1544,
+      "step": 43687
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.366651428571429e-06,
+      "loss": 0.8804,
+      "step": 43688
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3663828571428575e-06,
+      "loss": 1.0475,
+      "step": 43689
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3661142857142857e-06,
+      "loss": 1.2011,
+      "step": 43690
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3658457142857143e-06,
+      "loss": 1.0804,
+      "step": 43691
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.365577142857143e-06,
+      "loss": 0.7488,
+      "step": 43692
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3653085714285715e-06,
+      "loss": 1.1011,
+      "step": 43693
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.36504e-06,
+      "loss": 1.444,
+      "step": 43694
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3647714285714287e-06,
+      "loss": 1.0433,
+      "step": 43695
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3645028571428573e-06,
+      "loss": 0.8706,
+      "step": 43696
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.364234285714286e-06,
+      "loss": 1.1158,
+      "step": 43697
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3639657142857145e-06,
+      "loss": 1.4886,
+      "step": 43698
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3636971428571426e-06,
+      "loss": 0.8783,
+      "step": 43699
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3634285714285712e-06,
+      "loss": 0.9352,
+      "step": 43700
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3631600000000003e-06,
+      "loss": 1.425,
+      "step": 43701
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.362891428571429e-06,
+      "loss": 1.2289,
+      "step": 43702
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3626228571428575e-06,
+      "loss": 1.278,
+      "step": 43703
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3623542857142856e-06,
+      "loss": 1.0707,
+      "step": 43704
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3620857142857142e-06,
+      "loss": 1.395,
+      "step": 43705
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.361817142857143e-06,
+      "loss": 1.564,
+      "step": 43706
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3615485714285714e-06,
+      "loss": 1.1934,
+      "step": 43707
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3612800000000005e-06,
+      "loss": 0.8999,
+      "step": 43708
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3610114285714286e-06,
+      "loss": 1.188,
+      "step": 43709
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3607428571428572e-06,
+      "loss": 0.9109,
+      "step": 43710
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.360474285714286e-06,
+      "loss": 1.2793,
+      "step": 43711
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3602057142857144e-06,
+      "loss": 1.1844,
+      "step": 43712
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3599371428571426e-06,
+      "loss": 1.2322,
+      "step": 43713
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3596685714285716e-06,
+      "loss": 0.8463,
+      "step": 43714
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3594000000000002e-06,
+      "loss": 0.9917,
+      "step": 43715
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.359131428571429e-06,
+      "loss": 1.1813,
+      "step": 43716
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3588628571428574e-06,
+      "loss": 1.4246,
+      "step": 43717
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3585942857142856e-06,
+      "loss": 0.9748,
+      "step": 43718
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.358325714285714e-06,
+      "loss": 0.9244,
+      "step": 43719
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.358057142857143e-06,
+      "loss": 1.5347,
+      "step": 43720
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.357788571428572e-06,
+      "loss": 0.8315,
+      "step": 43721
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.35752e-06,
+      "loss": 0.9523,
+      "step": 43722
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3572514285714286e-06,
+      "loss": 1.1769,
+      "step": 43723
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.356982857142857e-06,
+      "loss": 1.5275,
+      "step": 43724
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.356714285714286e-06,
+      "loss": 1.2213,
+      "step": 43725
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3564457142857144e-06,
+      "loss": 0.8382,
+      "step": 43726
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.356177142857143e-06,
+      "loss": 1.371,
+      "step": 43727
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3559085714285716e-06,
+      "loss": 1.0239,
+      "step": 43728
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.35564e-06,
+      "loss": 1.1459,
+      "step": 43729
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.355371428571429e-06,
+      "loss": 1.2155,
+      "step": 43730
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.355102857142857e-06,
+      "loss": 0.865,
+      "step": 43731
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3548342857142856e-06,
+      "loss": 1.2218,
+      "step": 43732
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.354565714285714e-06,
+      "loss": 1.1235,
+      "step": 43733
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.354297142857143e-06,
+      "loss": 1.2467,
+      "step": 43734
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.354028571428572e-06,
+      "loss": 1.1367,
+      "step": 43735
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.35376e-06,
+      "loss": 0.6646,
+      "step": 43736
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3534914285714286e-06,
+      "loss": 0.8534,
+      "step": 43737
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.353222857142857e-06,
+      "loss": 0.9322,
+      "step": 43738
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3529542857142858e-06,
+      "loss": 1.3941,
+      "step": 43739
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3526857142857144e-06,
+      "loss": 1.5025,
+      "step": 43740
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.352417142857143e-06,
+      "loss": 1.4873,
+      "step": 43741
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3521485714285716e-06,
+      "loss": 1.0706,
+      "step": 43742
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.35188e-06,
+      "loss": 1.3184,
+      "step": 43743
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3516114285714288e-06,
+      "loss": 1.2704,
+      "step": 43744
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.351342857142857e-06,
+      "loss": 1.2153,
+      "step": 43745
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3510742857142855e-06,
+      "loss": 1.2788,
+      "step": 43746
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3508057142857146e-06,
+      "loss": 1.4198,
+      "step": 43747
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.350537142857143e-06,
+      "loss": 1.3202,
+      "step": 43748
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3502685714285718e-06,
+      "loss": 1.1557,
+      "step": 43749
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.35e-06,
+      "loss": 0.9485,
+      "step": 43750
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3497314285714285e-06,
+      "loss": 1.2432,
+      "step": 43751
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.349462857142857e-06,
+      "loss": 0.82,
+      "step": 43752
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3491942857142857e-06,
+      "loss": 1.1828,
+      "step": 43753
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3489257142857143e-06,
+      "loss": 1.0468,
+      "step": 43754
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.348657142857143e-06,
+      "loss": 1.1563,
+      "step": 43755
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3483885714285715e-06,
+      "loss": 1.2253,
+      "step": 43756
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.34812e-06,
+      "loss": 1.1946,
+      "step": 43757
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3478514285714287e-06,
+      "loss": 1.0919,
+      "step": 43758
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.347582857142857e-06,
+      "loss": 1.5471,
+      "step": 43759
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.347314285714286e-06,
+      "loss": 1.3635,
+      "step": 43760
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3470457142857145e-06,
+      "loss": 1.073,
+      "step": 43761
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.346777142857143e-06,
+      "loss": 0.6775,
+      "step": 43762
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3465085714285713e-06,
+      "loss": 1.3513,
+      "step": 43763
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.34624e-06,
+      "loss": 0.8923,
+      "step": 43764
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3459714285714285e-06,
+      "loss": 0.7985,
+      "step": 43765
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.345702857142857e-06,
+      "loss": 0.8753,
+      "step": 43766
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.345434285714286e-06,
+      "loss": 1.109,
+      "step": 43767
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3451657142857143e-06,
+      "loss": 0.8996,
+      "step": 43768
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.344897142857143e-06,
+      "loss": 0.8003,
+      "step": 43769
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3446285714285715e-06,
+      "loss": 0.6584,
+      "step": 43770
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.34436e-06,
+      "loss": 0.9114,
+      "step": 43771
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3440914285714287e-06,
+      "loss": 0.8726,
+      "step": 43772
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3438228571428573e-06,
+      "loss": 1.1265,
+      "step": 43773
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.343554285714286e-06,
+      "loss": 1.1034,
+      "step": 43774
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3432857142857145e-06,
+      "loss": 1.1392,
+      "step": 43775
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.343017142857143e-06,
+      "loss": 1.4136,
+      "step": 43776
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3427485714285713e-06,
+      "loss": 0.9298,
+      "step": 43777
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.34248e-06,
+      "loss": 1.1552,
+      "step": 43778
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3422114285714285e-06,
+      "loss": 0.8451,
+      "step": 43779
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3419428571428575e-06,
+      "loss": 1.2574,
+      "step": 43780
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.341674285714286e-06,
+      "loss": 1.3036,
+      "step": 43781
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3414057142857143e-06,
+      "loss": 0.8916,
+      "step": 43782
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.341137142857143e-06,
+      "loss": 1.3231,
+      "step": 43783
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3408685714285715e-06,
+      "loss": 1.1089,
+      "step": 43784
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3406e-06,
+      "loss": 0.7353,
+      "step": 43785
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3403314285714283e-06,
+      "loss": 1.5299,
+      "step": 43786
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3400628571428573e-06,
+      "loss": 0.6941,
+      "step": 43787
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.339794285714286e-06,
+      "loss": 1.4819,
+      "step": 43788
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3395257142857145e-06,
+      "loss": 1.1559,
+      "step": 43789
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.339257142857143e-06,
+      "loss": 1.0938,
+      "step": 43790
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3389885714285713e-06,
+      "loss": 1.1953,
+      "step": 43791
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.33872e-06,
+      "loss": 1.4576,
+      "step": 43792
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.338451428571429e-06,
+      "loss": 0.8095,
+      "step": 43793
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3381828571428575e-06,
+      "loss": 0.6752,
+      "step": 43794
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.337914285714286e-06,
+      "loss": 1.2545,
+      "step": 43795
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3376457142857143e-06,
+      "loss": 1.0374,
+      "step": 43796
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.337377142857143e-06,
+      "loss": 1.1676,
+      "step": 43797
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3371085714285715e-06,
+      "loss": 1.0656,
+      "step": 43798
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.33684e-06,
+      "loss": 1.0663,
+      "step": 43799
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3365714285714287e-06,
+      "loss": 0.9904,
+      "step": 43800
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3363028571428573e-06,
+      "loss": 1.1956,
+      "step": 43801
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.336034285714286e-06,
+      "loss": 1.0541,
+      "step": 43802
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3357657142857145e-06,
+      "loss": 1.2443,
+      "step": 43803
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.335497142857143e-06,
+      "loss": 1.0486,
+      "step": 43804
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3352285714285712e-06,
+      "loss": 1.0476,
+      "step": 43805
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3349600000000003e-06,
+      "loss": 1.2976,
+      "step": 43806
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.334691428571429e-06,
+      "loss": 1.3532,
+      "step": 43807
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3344228571428575e-06,
+      "loss": 1.0202,
+      "step": 43808
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3341542857142856e-06,
+      "loss": 1.4616,
+      "step": 43809
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3338857142857142e-06,
+      "loss": 1.0842,
+      "step": 43810
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.333617142857143e-06,
+      "loss": 1.0737,
+      "step": 43811
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3333485714285714e-06,
+      "loss": 1.1057,
+      "step": 43812
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3330800000000005e-06,
+      "loss": 0.8255,
+      "step": 43813
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3328114285714286e-06,
+      "loss": 1.1104,
+      "step": 43814
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3325428571428572e-06,
+      "loss": 1.0105,
+      "step": 43815
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.332274285714286e-06,
+      "loss": 1.0214,
+      "step": 43816
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3320057142857144e-06,
+      "loss": 1.5194,
+      "step": 43817
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3317371428571426e-06,
+      "loss": 0.8678,
+      "step": 43818
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.331468571428571e-06,
+      "loss": 1.0484,
+      "step": 43819
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3312000000000002e-06,
+      "loss": 0.8725,
+      "step": 43820
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.330931428571429e-06,
+      "loss": 1.1037,
+      "step": 43821
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3306628571428574e-06,
+      "loss": 1.1528,
+      "step": 43822
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3303942857142856e-06,
+      "loss": 1.1133,
+      "step": 43823
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.330125714285714e-06,
+      "loss": 0.9963,
+      "step": 43824
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.329857142857143e-06,
+      "loss": 1.3378,
+      "step": 43825
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.329588571428572e-06,
+      "loss": 0.9724,
+      "step": 43826
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3293200000000004e-06,
+      "loss": 1.1549,
+      "step": 43827
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3290514285714286e-06,
+      "loss": 1.2751,
+      "step": 43828
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.328782857142857e-06,
+      "loss": 1.1924,
+      "step": 43829
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.328514285714286e-06,
+      "loss": 0.8976,
+      "step": 43830
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3282457142857144e-06,
+      "loss": 1.2497,
+      "step": 43831
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3279771428571426e-06,
+      "loss": 0.9099,
+      "step": 43832
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3277085714285716e-06,
+      "loss": 0.7408,
+      "step": 43833
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.32744e-06,
+      "loss": 1.0468,
+      "step": 43834
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.327171428571429e-06,
+      "loss": 0.6927,
+      "step": 43835
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3269028571428574e-06,
+      "loss": 1.1234,
+      "step": 43836
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.3266342857142856e-06,
+      "loss": 0.7681,
+      "step": 43837
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.326365714285714e-06,
+      "loss": 1.3278,
+      "step": 43838
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3260971428571428e-06,
+      "loss": 1.1894,
+      "step": 43839
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.325828571428572e-06,
+      "loss": 0.704,
+      "step": 43840
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.32556e-06,
+      "loss": 0.9871,
+      "step": 43841
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3252914285714286e-06,
+      "loss": 1.2035,
+      "step": 43842
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.325022857142857e-06,
+      "loss": 1.1503,
+      "step": 43843
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3247542857142858e-06,
+      "loss": 0.7085,
+      "step": 43844
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3244857142857144e-06,
+      "loss": 1.5842,
+      "step": 43845
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.324217142857143e-06,
+      "loss": 1.284,
+      "step": 43846
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3239485714285716e-06,
+      "loss": 1.2918,
+      "step": 43847
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.32368e-06,
+      "loss": 1.1287,
+      "step": 43848
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3234114285714288e-06,
+      "loss": 0.8839,
+      "step": 43849
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.323142857142857e-06,
+      "loss": 1.4318,
+      "step": 43850
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3228742857142855e-06,
+      "loss": 1.0221,
+      "step": 43851
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.322605714285714e-06,
+      "loss": 1.1641,
+      "step": 43852
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.322337142857143e-06,
+      "loss": 0.9224,
+      "step": 43853
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3220685714285718e-06,
+      "loss": 0.9638,
+      "step": 43854
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3218e-06,
+      "loss": 1.1082,
+      "step": 43855
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3215314285714285e-06,
+      "loss": 1.3339,
+      "step": 43856
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.321262857142857e-06,
+      "loss": 0.7035,
+      "step": 43857
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3209942857142857e-06,
+      "loss": 1.1201,
+      "step": 43858
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3207257142857148e-06,
+      "loss": 0.7574,
+      "step": 43859
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.320457142857143e-06,
+      "loss": 0.941,
+      "step": 43860
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3201885714285715e-06,
+      "loss": 0.9356,
+      "step": 43861
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.31992e-06,
+      "loss": 0.8509,
+      "step": 43862
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3196514285714287e-06,
+      "loss": 1.5517,
+      "step": 43863
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.319382857142857e-06,
+      "loss": 1.2798,
+      "step": 43864
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3191142857142855e-06,
+      "loss": 1.2656,
+      "step": 43865
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3188457142857145e-06,
+      "loss": 0.9316,
+      "step": 43866
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.318577142857143e-06,
+      "loss": 1.4124,
+      "step": 43867
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3183085714285717e-06,
+      "loss": 1.3468,
+      "step": 43868
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.31804e-06,
+      "loss": 1.0819,
+      "step": 43869
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3177714285714285e-06,
+      "loss": 1.0267,
+      "step": 43870
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.317502857142857e-06,
+      "loss": 0.9608,
+      "step": 43871
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3172342857142857e-06,
+      "loss": 1.3678,
+      "step": 43872
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3169657142857143e-06,
+      "loss": 1.5268,
+      "step": 43873
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.316697142857143e-06,
+      "loss": 1.0633,
+      "step": 43874
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3164285714285715e-06,
+      "loss": 0.7095,
+      "step": 43875
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.31616e-06,
+      "loss": 1.2096,
+      "step": 43876
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3158914285714287e-06,
+      "loss": 0.979,
+      "step": 43877
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.315622857142857e-06,
+      "loss": 1.1317,
+      "step": 43878
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.315354285714286e-06,
+      "loss": 1.3001,
+      "step": 43879
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3150857142857145e-06,
+      "loss": 0.9369,
+      "step": 43880
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.314817142857143e-06,
+      "loss": 0.7946,
+      "step": 43881
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3145485714285713e-06,
+      "loss": 1.0719,
+      "step": 43882
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.31428e-06,
+      "loss": 1.3215,
+      "step": 43883
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3140114285714285e-06,
+      "loss": 0.9676,
+      "step": 43884
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.313742857142857e-06,
+      "loss": 1.1676,
+      "step": 43885
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.313474285714286e-06,
+      "loss": 1.0842,
+      "step": 43886
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3132057142857143e-06,
+      "loss": 1.2009,
+      "step": 43887
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.312937142857143e-06,
+      "loss": 0.9364,
+      "step": 43888
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3126685714285715e-06,
+      "loss": 1.2169,
+      "step": 43889
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3124e-06,
+      "loss": 0.9563,
+      "step": 43890
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3121314285714287e-06,
+      "loss": 0.8855,
+      "step": 43891
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3118628571428573e-06,
+      "loss": 0.9849,
+      "step": 43892
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.311594285714286e-06,
+      "loss": 1.0554,
+      "step": 43893
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3113257142857145e-06,
+      "loss": 0.9345,
+      "step": 43894
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.311057142857143e-06,
+      "loss": 0.9348,
+      "step": 43895
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3107885714285713e-06,
+      "loss": 0.7379,
+      "step": 43896
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.31052e-06,
+      "loss": 1.0379,
+      "step": 43897
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3102514285714285e-06,
+      "loss": 1.273,
+      "step": 43898
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3099828571428575e-06,
+      "loss": 1.2331,
+      "step": 43899
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.309714285714286e-06,
+      "loss": 1.2069,
+      "step": 43900
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3094457142857143e-06,
+      "loss": 1.1846,
+      "step": 43901
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.309177142857143e-06,
+      "loss": 1.4263,
+      "step": 43902
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3089085714285715e-06,
+      "loss": 1.2481,
+      "step": 43903
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.30864e-06,
+      "loss": 1.5193,
+      "step": 43904
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3083714285714287e-06,
+      "loss": 0.6207,
+      "step": 43905
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3081028571428573e-06,
+      "loss": 1.4939,
+      "step": 43906
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.307834285714286e-06,
+      "loss": 0.9791,
+      "step": 43907
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3075657142857145e-06,
+      "loss": 1.0363,
+      "step": 43908
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.307297142857143e-06,
+      "loss": 1.0955,
+      "step": 43909
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3070285714285712e-06,
+      "loss": 1.2186,
+      "step": 43910
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.30676e-06,
+      "loss": 0.9164,
+      "step": 43911
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.306491428571429e-06,
+      "loss": 1.418,
+      "step": 43912
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3062228571428575e-06,
+      "loss": 1.048,
+      "step": 43913
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.305954285714286e-06,
+      "loss": 0.7444,
+      "step": 43914
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3056857142857142e-06,
+      "loss": 1.2265,
+      "step": 43915
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.305417142857143e-06,
+      "loss": 0.912,
+      "step": 43916
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3051485714285714e-06,
+      "loss": 1.3901,
+      "step": 43917
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.30488e-06,
+      "loss": 0.6344,
+      "step": 43918
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3046114285714286e-06,
+      "loss": 0.9718,
+      "step": 43919
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3043428571428572e-06,
+      "loss": 1.1201,
+      "step": 43920
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.304074285714286e-06,
+      "loss": 0.5563,
+      "step": 43921
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3038057142857144e-06,
+      "loss": 1.4931,
+      "step": 43922
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.303537142857143e-06,
+      "loss": 1.0609,
+      "step": 43923
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.303268571428571e-06,
+      "loss": 0.9239,
+      "step": 43924
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3030000000000002e-06,
+      "loss": 0.9276,
+      "step": 43925
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.302731428571429e-06,
+      "loss": 0.9988,
+      "step": 43926
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3024628571428574e-06,
+      "loss": 1.3418,
+      "step": 43927
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3021942857142856e-06,
+      "loss": 1.0375,
+      "step": 43928
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.301925714285714e-06,
+      "loss": 1.4199,
+      "step": 43929
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.301657142857143e-06,
+      "loss": 1.0934,
+      "step": 43930
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3013885714285714e-06,
+      "loss": 1.2791,
+      "step": 43931
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3011200000000004e-06,
+      "loss": 1.1275,
+      "step": 43932
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3008514285714286e-06,
+      "loss": 1.2107,
+      "step": 43933
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.300582857142857e-06,
+      "loss": 0.9674,
+      "step": 43934
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.300314285714286e-06,
+      "loss": 0.9544,
+      "step": 43935
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.3000457142857144e-06,
+      "loss": 1.4205,
+      "step": 43936
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2997771428571426e-06,
+      "loss": 1.5144,
+      "step": 43937
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2995085714285716e-06,
+      "loss": 1.3618,
+      "step": 43938
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.29924e-06,
+      "loss": 0.7324,
+      "step": 43939
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.298971428571429e-06,
+      "loss": 1.3013,
+      "step": 43940
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2987028571428574e-06,
+      "loss": 1.2404,
+      "step": 43941
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2984342857142856e-06,
+      "loss": 1.2156,
+      "step": 43942
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.298165714285714e-06,
+      "loss": 1.1473,
+      "step": 43943
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2978971428571428e-06,
+      "loss": 0.9564,
+      "step": 43944
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.297628571428572e-06,
+      "loss": 1.2998,
+      "step": 43945
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2973600000000004e-06,
+      "loss": 1.0973,
+      "step": 43946
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2970914285714286e-06,
+      "loss": 1.0475,
+      "step": 43947
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.296822857142857e-06,
+      "loss": 1.1087,
+      "step": 43948
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2965542857142858e-06,
+      "loss": 1.1679,
+      "step": 43949
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2962857142857144e-06,
+      "loss": 1.646,
+      "step": 43950
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.296017142857143e-06,
+      "loss": 1.3496,
+      "step": 43951
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2957485714285716e-06,
+      "loss": 1.1642,
+      "step": 43952
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.29548e-06,
+      "loss": 1.076,
+      "step": 43953
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2952114285714288e-06,
+      "loss": 1.1887,
+      "step": 43954
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2949428571428574e-06,
+      "loss": 1.1351,
+      "step": 43955
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2946742857142855e-06,
+      "loss": 1.22,
+      "step": 43956
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.294405714285714e-06,
+      "loss": 1.2792,
+      "step": 43957
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.294137142857143e-06,
+      "loss": 0.9171,
+      "step": 43958
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2938685714285718e-06,
+      "loss": 1.1285,
+      "step": 43959
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2936e-06,
+      "loss": 1.1086,
+      "step": 43960
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2933314285714285e-06,
+      "loss": 1.1166,
+      "step": 43961
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.293062857142857e-06,
+      "loss": 1.342,
+      "step": 43962
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2927942857142857e-06,
+      "loss": 1.1327,
+      "step": 43963
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2925257142857143e-06,
+      "loss": 0.9376,
+      "step": 43964
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.292257142857143e-06,
+      "loss": 1.0548,
+      "step": 43965
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2919885714285715e-06,
+      "loss": 1.3149,
+      "step": 43966
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.29172e-06,
+      "loss": 1.2428,
+      "step": 43967
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2914514285714287e-06,
+      "loss": 1.0577,
+      "step": 43968
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.291182857142857e-06,
+      "loss": 1.2286,
+      "step": 43969
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2909142857142855e-06,
+      "loss": 1.0006,
+      "step": 43970
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2906457142857145e-06,
+      "loss": 0.8731,
+      "step": 43971
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.290377142857143e-06,
+      "loss": 1.3034,
+      "step": 43972
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2901085714285717e-06,
+      "loss": 1.2222,
+      "step": 43973
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.28984e-06,
+      "loss": 1.0045,
+      "step": 43974
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2895714285714285e-06,
+      "loss": 1.2451,
+      "step": 43975
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.289302857142857e-06,
+      "loss": 1.1209,
+      "step": 43976
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2890342857142857e-06,
+      "loss": 0.9681,
+      "step": 43977
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2887657142857147e-06,
+      "loss": 1.2836,
+      "step": 43978
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.288497142857143e-06,
+      "loss": 1.011,
+      "step": 43979
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2882285714285715e-06,
+      "loss": 1.0427,
+      "step": 43980
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.28796e-06,
+      "loss": 1.3207,
+      "step": 43981
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2876914285714287e-06,
+      "loss": 1.0174,
+      "step": 43982
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.287422857142857e-06,
+      "loss": 1.1227,
+      "step": 43983
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.287154285714286e-06,
+      "loss": 1.0552,
+      "step": 43984
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2868857142857145e-06,
+      "loss": 0.9226,
+      "step": 43985
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.286617142857143e-06,
+      "loss": 1.096,
+      "step": 43986
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2863485714285717e-06,
+      "loss": 0.8979,
+      "step": 43987
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.28608e-06,
+      "loss": 1.1824,
+      "step": 43988
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2858114285714285e-06,
+      "loss": 0.8555,
+      "step": 43989
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.285542857142857e-06,
+      "loss": 1.238,
+      "step": 43990
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.285274285714286e-06,
+      "loss": 1.1933,
+      "step": 43991
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2850057142857143e-06,
+      "loss": 1.2926,
+      "step": 43992
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.284737142857143e-06,
+      "loss": 1.4491,
+      "step": 43993
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2844685714285715e-06,
+      "loss": 0.8672,
+      "step": 43994
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2842e-06,
+      "loss": 0.7864,
+      "step": 43995
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2839314285714287e-06,
+      "loss": 0.669,
+      "step": 43996
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2836628571428573e-06,
+      "loss": 1.2107,
+      "step": 43997
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.283394285714286e-06,
+      "loss": 0.9299,
+      "step": 43998
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2831257142857145e-06,
+      "loss": 0.8263,
+      "step": 43999
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.282857142857143e-06,
+      "loss": 0.9518,
+      "step": 44000
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2825885714285717e-06,
+      "loss": 1.372,
+      "step": 44001
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.28232e-06,
+      "loss": 0.9045,
+      "step": 44002
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2820514285714285e-06,
+      "loss": 1.5058,
+      "step": 44003
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2817828571428575e-06,
+      "loss": 1.4912,
+      "step": 44004
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.281514285714286e-06,
+      "loss": 0.8106,
+      "step": 44005
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2812457142857143e-06,
+      "loss": 0.9408,
+      "step": 44006
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.280977142857143e-06,
+      "loss": 1.1851,
+      "step": 44007
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2807085714285715e-06,
+      "loss": 0.7576,
+      "step": 44008
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.28044e-06,
+      "loss": 0.7617,
+      "step": 44009
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2801714285714287e-06,
+      "loss": 1.3899,
+      "step": 44010
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2799028571428573e-06,
+      "loss": 1.173,
+      "step": 44011
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.279634285714286e-06,
+      "loss": 1.1885,
+      "step": 44012
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2793657142857145e-06,
+      "loss": 0.7216,
+      "step": 44013
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.279097142857143e-06,
+      "loss": 1.1847,
+      "step": 44014
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2788285714285712e-06,
+      "loss": 1.4637,
+      "step": 44015
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.27856e-06,
+      "loss": 1.2487,
+      "step": 44016
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.278291428571429e-06,
+      "loss": 1.1468,
+      "step": 44017
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2780228571428575e-06,
+      "loss": 0.7009,
+      "step": 44018
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.277754285714286e-06,
+      "loss": 0.9514,
+      "step": 44019
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2774857142857142e-06,
+      "loss": 1.2465,
+      "step": 44020
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.277217142857143e-06,
+      "loss": 1.5384,
+      "step": 44021
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2769485714285714e-06,
+      "loss": 1.1327,
+      "step": 44022
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.27668e-06,
+      "loss": 1.3169,
+      "step": 44023
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2764114285714286e-06,
+      "loss": 1.3176,
+      "step": 44024
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2761428571428572e-06,
+      "loss": 0.8616,
+      "step": 44025
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.275874285714286e-06,
+      "loss": 0.8559,
+      "step": 44026
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2756057142857144e-06,
+      "loss": 1.024,
+      "step": 44027
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.275337142857143e-06,
+      "loss": 1.1645,
+      "step": 44028
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.275068571428571e-06,
+      "loss": 1.0037,
+      "step": 44029
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2748000000000002e-06,
+      "loss": 0.9115,
+      "step": 44030
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.274531428571429e-06,
+      "loss": 0.9917,
+      "step": 44031
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2742628571428574e-06,
+      "loss": 1.2025,
+      "step": 44032
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.273994285714286e-06,
+      "loss": 1.4949,
+      "step": 44033
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.273725714285714e-06,
+      "loss": 1.5732,
+      "step": 44034
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.273457142857143e-06,
+      "loss": 0.6786,
+      "step": 44035
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2731885714285714e-06,
+      "loss": 0.9313,
+      "step": 44036
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2729200000000004e-06,
+      "loss": 1.279,
+      "step": 44037
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2726514285714286e-06,
+      "loss": 1.4029,
+      "step": 44038
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.272382857142857e-06,
+      "loss": 0.9611,
+      "step": 44039
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.272114285714286e-06,
+      "loss": 1.5759,
+      "step": 44040
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2718457142857144e-06,
+      "loss": 0.8094,
+      "step": 44041
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.271577142857143e-06,
+      "loss": 0.8764,
+      "step": 44042
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.271308571428571e-06,
+      "loss": 0.584,
+      "step": 44043
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.27104e-06,
+      "loss": 1.5385,
+      "step": 44044
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.270771428571429e-06,
+      "loss": 0.773,
+      "step": 44045
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2705028571428574e-06,
+      "loss": 1.1963,
+      "step": 44046
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2702342857142856e-06,
+      "loss": 1.1288,
+      "step": 44047
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.269965714285714e-06,
+      "loss": 1.296,
+      "step": 44048
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2696971428571428e-06,
+      "loss": 0.9132,
+      "step": 44049
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.269428571428572e-06,
+      "loss": 1.195,
+      "step": 44050
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2691600000000004e-06,
+      "loss": 0.8804,
+      "step": 44051
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2688914285714286e-06,
+      "loss": 1.7262,
+      "step": 44052
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.268622857142857e-06,
+      "loss": 1.0812,
+      "step": 44053
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2683542857142858e-06,
+      "loss": 0.9107,
+      "step": 44054
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2680857142857144e-06,
+      "loss": 1.4009,
+      "step": 44055
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2678171428571425e-06,
+      "loss": 1.2546,
+      "step": 44056
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2675485714285716e-06,
+      "loss": 1.4703,
+      "step": 44057
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.26728e-06,
+      "loss": 1.1624,
+      "step": 44058
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2670114285714288e-06,
+      "loss": 1.4131,
+      "step": 44059
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2667428571428574e-06,
+      "loss": 1.2948,
+      "step": 44060
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2664742857142855e-06,
+      "loss": 0.9858,
+      "step": 44061
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.266205714285714e-06,
+      "loss": 1.7078,
+      "step": 44062
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.265937142857143e-06,
+      "loss": 1.0994,
+      "step": 44063
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2656685714285718e-06,
+      "loss": 0.9705,
+      "step": 44064
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2654000000000004e-06,
+      "loss": 1.0649,
+      "step": 44065
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2651314285714285e-06,
+      "loss": 0.7694,
+      "step": 44066
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.264862857142857e-06,
+      "loss": 1.2423,
+      "step": 44067
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2645942857142857e-06,
+      "loss": 1.2077,
+      "step": 44068
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2643257142857143e-06,
+      "loss": 1.2947,
+      "step": 44069
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.264057142857143e-06,
+      "loss": 1.3685,
+      "step": 44070
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2637885714285715e-06,
+      "loss": 1.0759,
+      "step": 44071
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.26352e-06,
+      "loss": 0.9363,
+      "step": 44072
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2632514285714287e-06,
+      "loss": 1.4215,
+      "step": 44073
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2629828571428573e-06,
+      "loss": 1.1069,
+      "step": 44074
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2627142857142855e-06,
+      "loss": 0.8107,
+      "step": 44075
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.262445714285714e-06,
+      "loss": 1.2707,
+      "step": 44076
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.262177142857143e-06,
+      "loss": 0.7048,
+      "step": 44077
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2619085714285717e-06,
+      "loss": 1.0882,
+      "step": 44078
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.26164e-06,
+      "loss": 1.2981,
+      "step": 44079
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2613714285714285e-06,
+      "loss": 0.8705,
+      "step": 44080
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.261102857142857e-06,
+      "loss": 1.3702,
+      "step": 44081
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2608342857142857e-06,
+      "loss": 1.1978,
+      "step": 44082
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2605657142857147e-06,
+      "loss": 1.052,
+      "step": 44083
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.260297142857143e-06,
+      "loss": 0.8959,
+      "step": 44084
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2600285714285715e-06,
+      "loss": 0.736,
+      "step": 44085
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.25976e-06,
+      "loss": 1.2751,
+      "step": 44086
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2594914285714287e-06,
+      "loss": 1.4612,
+      "step": 44087
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2592228571428573e-06,
+      "loss": 1.0576,
+      "step": 44088
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2589542857142855e-06,
+      "loss": 0.9571,
+      "step": 44089
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2586857142857145e-06,
+      "loss": 1.1212,
+      "step": 44090
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.258417142857143e-06,
+      "loss": 1.0578,
+      "step": 44091
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2581485714285717e-06,
+      "loss": 1.212,
+      "step": 44092
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.25788e-06,
+      "loss": 0.8837,
+      "step": 44093
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2576114285714285e-06,
+      "loss": 1.1031,
+      "step": 44094
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.257342857142857e-06,
+      "loss": 1.2493,
+      "step": 44095
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2570742857142857e-06,
+      "loss": 0.8298,
+      "step": 44096
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2568057142857147e-06,
+      "loss": 0.6805,
+      "step": 44097
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.256537142857143e-06,
+      "loss": 0.8624,
+      "step": 44098
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2562685714285715e-06,
+      "loss": 0.6665,
+      "step": 44099
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.256e-06,
+      "loss": 0.947,
+      "step": 44100
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2557314285714287e-06,
+      "loss": 0.886,
+      "step": 44101
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.255462857142857e-06,
+      "loss": 1.1361,
+      "step": 44102
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.255194285714286e-06,
+      "loss": 1.0458,
+      "step": 44103
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2549257142857145e-06,
+      "loss": 1.2338,
+      "step": 44104
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.254657142857143e-06,
+      "loss": 0.9666,
+      "step": 44105
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2543885714285717e-06,
+      "loss": 0.9048,
+      "step": 44106
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.25412e-06,
+      "loss": 1.3934,
+      "step": 44107
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2538514285714285e-06,
+      "loss": 0.968,
+      "step": 44108
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.253582857142857e-06,
+      "loss": 0.9368,
+      "step": 44109
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.253314285714286e-06,
+      "loss": 1.0808,
+      "step": 44110
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2530457142857143e-06,
+      "loss": 1.4866,
+      "step": 44111
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.252777142857143e-06,
+      "loss": 1.0165,
+      "step": 44112
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2525085714285715e-06,
+      "loss": 1.0252,
+      "step": 44113
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.25224e-06,
+      "loss": 1.0316,
+      "step": 44114
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2519714285714287e-06,
+      "loss": 1.102,
+      "step": 44115
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2517028571428573e-06,
+      "loss": 1.3911,
+      "step": 44116
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.251434285714286e-06,
+      "loss": 0.938,
+      "step": 44117
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2511657142857145e-06,
+      "loss": 0.9928,
+      "step": 44118
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.250897142857143e-06,
+      "loss": 0.8742,
+      "step": 44119
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2506285714285717e-06,
+      "loss": 1.0719,
+      "step": 44120
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.25036e-06,
+      "loss": 0.8098,
+      "step": 44121
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2500914285714284e-06,
+      "loss": 1.3373,
+      "step": 44122
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2498228571428575e-06,
+      "loss": 1.1679,
+      "step": 44123
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.249554285714286e-06,
+      "loss": 1.0592,
+      "step": 44124
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2492857142857142e-06,
+      "loss": 0.8271,
+      "step": 44125
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.249017142857143e-06,
+      "loss": 1.3373,
+      "step": 44126
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2487485714285714e-06,
+      "loss": 1.3192,
+      "step": 44127
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.24848e-06,
+      "loss": 0.9719,
+      "step": 44128
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2482114285714286e-06,
+      "loss": 1.2635,
+      "step": 44129
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2479428571428572e-06,
+      "loss": 1.4596,
+      "step": 44130
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.247674285714286e-06,
+      "loss": 1.6539,
+      "step": 44131
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2474057142857144e-06,
+      "loss": 1.1268,
+      "step": 44132
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.247137142857143e-06,
+      "loss": 1.0314,
+      "step": 44133
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246868571428571e-06,
+      "loss": 1.0123,
+      "step": 44134
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2466e-06,
+      "loss": 0.9316,
+      "step": 44135
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246331428571429e-06,
+      "loss": 1.3044,
+      "step": 44136
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2460628571428574e-06,
+      "loss": 0.8241,
+      "step": 44137
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245794285714286e-06,
+      "loss": 1.1338,
+      "step": 44138
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245525714285714e-06,
+      "loss": 0.7635,
+      "step": 44139
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245257142857143e-06,
+      "loss": 0.8996,
+      "step": 44140
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2449885714285714e-06,
+      "loss": 1.2529,
+      "step": 44141
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.24472e-06,
+      "loss": 1.2521,
+      "step": 44142
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2444514285714286e-06,
+      "loss": 1.0592,
+      "step": 44143
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244182857142857e-06,
+      "loss": 1.0729,
+      "step": 44144
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.243914285714286e-06,
+      "loss": 1.1785,
+      "step": 44145
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2436457142857144e-06,
+      "loss": 1.0152,
+      "step": 44146
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.243377142857143e-06,
+      "loss": 1.2447,
+      "step": 44147
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.243108571428571e-06,
+      "loss": 1.1005,
+      "step": 44148
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.24284e-06,
+      "loss": 0.9993,
+      "step": 44149
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.242571428571429e-06,
+      "loss": 0.8917,
+      "step": 44150
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2423028571428574e-06,
+      "loss": 1.3754,
+      "step": 44151
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.242034285714286e-06,
+      "loss": 1.2734,
+      "step": 44152
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.241765714285714e-06,
+      "loss": 1.039,
+      "step": 44153
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2414971428571428e-06,
+      "loss": 1.2885,
+      "step": 44154
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2412285714285714e-06,
+      "loss": 1.3016,
+      "step": 44155
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2409600000000004e-06,
+      "loss": 1.0805,
+      "step": 44156
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2406914285714286e-06,
+      "loss": 1.1803,
+      "step": 44157
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.240422857142857e-06,
+      "loss": 1.4333,
+      "step": 44158
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2401542857142858e-06,
+      "loss": 1.1341,
+      "step": 44159
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2398857142857144e-06,
+      "loss": 1.0527,
+      "step": 44160
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.239617142857143e-06,
+      "loss": 1.1447,
+      "step": 44161
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2393485714285716e-06,
+      "loss": 0.8521,
+      "step": 44162
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.23908e-06,
+      "loss": 1.1493,
+      "step": 44163
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2388114285714288e-06,
+      "loss": 0.929,
+      "step": 44164
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2385428571428574e-06,
+      "loss": 0.769,
+      "step": 44165
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2382742857142855e-06,
+      "loss": 0.9381,
+      "step": 44166
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.238005714285714e-06,
+      "loss": 1.0028,
+      "step": 44167
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2377371428571427e-06,
+      "loss": 0.909,
+      "step": 44168
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2374685714285718e-06,
+      "loss": 1.5795,
+      "step": 44169
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2372000000000004e-06,
+      "loss": 1.1142,
+      "step": 44170
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2369314285714285e-06,
+      "loss": 1.4162,
+      "step": 44171
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.236662857142857e-06,
+      "loss": 1.1404,
+      "step": 44172
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2363942857142857e-06,
+      "loss": 0.7919,
+      "step": 44173
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2361257142857143e-06,
+      "loss": 1.6479,
+      "step": 44174
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.235857142857143e-06,
+      "loss": 0.7993,
+      "step": 44175
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2355885714285715e-06,
+      "loss": 0.9722,
+      "step": 44176
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.23532e-06,
+      "loss": 1.1196,
+      "step": 44177
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2350514285714287e-06,
+      "loss": 0.9394,
+      "step": 44178
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2347828571428573e-06,
+      "loss": 1.0911,
+      "step": 44179
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2345142857142855e-06,
+      "loss": 0.9613,
+      "step": 44180
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.234245714285714e-06,
+      "loss": 1.2485,
+      "step": 44181
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.233977142857143e-06,
+      "loss": 1.2023,
+      "step": 44182
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2337085714285717e-06,
+      "loss": 0.7402,
+      "step": 44183
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2334400000000003e-06,
+      "loss": 1.4922,
+      "step": 44184
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2331714285714285e-06,
+      "loss": 0.9278,
+      "step": 44185
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.232902857142857e-06,
+      "loss": 1.6491,
+      "step": 44186
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2326342857142857e-06,
+      "loss": 1.5148,
+      "step": 44187
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2323657142857143e-06,
+      "loss": 1.3796,
+      "step": 44188
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.232097142857143e-06,
+      "loss": 0.7964,
+      "step": 44189
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2318285714285715e-06,
+      "loss": 1.5209,
+      "step": 44190
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.23156e-06,
+      "loss": 1.3793,
+      "step": 44191
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2312914285714287e-06,
+      "loss": 1.4905,
+      "step": 44192
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2310228571428573e-06,
+      "loss": 1.0644,
+      "step": 44193
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2307542857142855e-06,
+      "loss": 1.2558,
+      "step": 44194
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2304857142857145e-06,
+      "loss": 0.6767,
+      "step": 44195
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.230217142857143e-06,
+      "loss": 1.4371,
+      "step": 44196
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2299485714285717e-06,
+      "loss": 1.2414,
+      "step": 44197
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.22968e-06,
+      "loss": 1.6111,
+      "step": 44198
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2294114285714285e-06,
+      "loss": 0.9453,
+      "step": 44199
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.229142857142857e-06,
+      "loss": 0.6986,
+      "step": 44200
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2288742857142857e-06,
+      "loss": 1.3234,
+      "step": 44201
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2286057142857147e-06,
+      "loss": 1.0895,
+      "step": 44202
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.228337142857143e-06,
+      "loss": 1.2379,
+      "step": 44203
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2280685714285715e-06,
+      "loss": 1.3934,
+      "step": 44204
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2278e-06,
+      "loss": 0.9404,
+      "step": 44205
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2275314285714287e-06,
+      "loss": 1.0254,
+      "step": 44206
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2272628571428573e-06,
+      "loss": 1.3237,
+      "step": 44207
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.226994285714286e-06,
+      "loss": 1.2022,
+      "step": 44208
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2267257142857145e-06,
+      "loss": 1.2667,
+      "step": 44209
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.226457142857143e-06,
+      "loss": 1.2164,
+      "step": 44210
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2261885714285717e-06,
+      "loss": 0.9311,
+      "step": 44211
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.22592e-06,
+      "loss": 1.2588,
+      "step": 44212
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2256514285714285e-06,
+      "loss": 0.7459,
+      "step": 44213
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.225382857142857e-06,
+      "loss": 0.9082,
+      "step": 44214
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.225114285714286e-06,
+      "loss": 0.8667,
+      "step": 44215
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2248457142857147e-06,
+      "loss": 1.3773,
+      "step": 44216
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.224577142857143e-06,
+      "loss": 1.0064,
+      "step": 44217
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2243085714285715e-06,
+      "loss": 1.0002,
+      "step": 44218
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.22404e-06,
+      "loss": 0.702,
+      "step": 44219
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2237714285714287e-06,
+      "loss": 1.063,
+      "step": 44220
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2235028571428573e-06,
+      "loss": 1.3082,
+      "step": 44221
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.223234285714286e-06,
+      "loss": 1.2829,
+      "step": 44222
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2229657142857145e-06,
+      "loss": 1.023,
+      "step": 44223
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.222697142857143e-06,
+      "loss": 1.1397,
+      "step": 44224
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2224285714285717e-06,
+      "loss": 0.9518,
+      "step": 44225
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.22216e-06,
+      "loss": 1.2042,
+      "step": 44226
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2218914285714284e-06,
+      "loss": 0.7899,
+      "step": 44227
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2216228571428575e-06,
+      "loss": 0.7741,
+      "step": 44228
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.221354285714286e-06,
+      "loss": 1.2893,
+      "step": 44229
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2210857142857142e-06,
+      "loss": 0.8581,
+      "step": 44230
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.220817142857143e-06,
+      "loss": 1.1923,
+      "step": 44231
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2205485714285714e-06,
+      "loss": 1.045,
+      "step": 44232
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.22028e-06,
+      "loss": 1.0785,
+      "step": 44233
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2200114285714286e-06,
+      "loss": 0.738,
+      "step": 44234
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2197428571428572e-06,
+      "loss": 0.7523,
+      "step": 44235
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.219474285714286e-06,
+      "loss": 1.1616,
+      "step": 44236
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2192057142857144e-06,
+      "loss": 1.5695,
+      "step": 44237
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.218937142857143e-06,
+      "loss": 0.7302,
+      "step": 44238
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2186685714285716e-06,
+      "loss": 0.9546,
+      "step": 44239
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2184e-06,
+      "loss": 1.2156,
+      "step": 44240
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.218131428571429e-06,
+      "loss": 0.953,
+      "step": 44241
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2178628571428574e-06,
+      "loss": 1.3664,
+      "step": 44242
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.217594285714286e-06,
+      "loss": 1.6962,
+      "step": 44243
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.217325714285714e-06,
+      "loss": 0.5902,
+      "step": 44244
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.217057142857143e-06,
+      "loss": 1.0509,
+      "step": 44245
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2167885714285714e-06,
+      "loss": 1.3127,
+      "step": 44246
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.21652e-06,
+      "loss": 1.0953,
+      "step": 44247
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.216251428571429e-06,
+      "loss": 1.1619,
+      "step": 44248
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.215982857142857e-06,
+      "loss": 1.125,
+      "step": 44249
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.215714285714286e-06,
+      "loss": 1.0332,
+      "step": 44250
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2154457142857144e-06,
+      "loss": 0.9984,
+      "step": 44251
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.215177142857143e-06,
+      "loss": 1.1844,
+      "step": 44252
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.214908571428571e-06,
+      "loss": 1.1231,
+      "step": 44253
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.21464e-06,
+      "loss": 1.021,
+      "step": 44254
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.214371428571429e-06,
+      "loss": 0.8863,
+      "step": 44255
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2141028571428574e-06,
+      "loss": 1.1208,
+      "step": 44256
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.213834285714286e-06,
+      "loss": 1.0599,
+      "step": 44257
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.213565714285714e-06,
+      "loss": 0.924,
+      "step": 44258
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2132971428571428e-06,
+      "loss": 1.6447,
+      "step": 44259
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2130285714285714e-06,
+      "loss": 1.1885,
+      "step": 44260
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2127600000000004e-06,
+      "loss": 0.9663,
+      "step": 44261
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2124914285714286e-06,
+      "loss": 0.8897,
+      "step": 44262
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.212222857142857e-06,
+      "loss": 1.1574,
+      "step": 44263
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2119542857142858e-06,
+      "loss": 1.3018,
+      "step": 44264
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2116857142857144e-06,
+      "loss": 0.9159,
+      "step": 44265
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.211417142857143e-06,
+      "loss": 1.1363,
+      "step": 44266
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.211148571428571e-06,
+      "loss": 1.3788,
+      "step": 44267
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.21088e-06,
+      "loss": 0.9668,
+      "step": 44268
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2106114285714288e-06,
+      "loss": 1.1061,
+      "step": 44269
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2103428571428574e-06,
+      "loss": 1.0823,
+      "step": 44270
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.210074285714286e-06,
+      "loss": 1.3809,
+      "step": 44271
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.209805714285714e-06,
+      "loss": 0.887,
+      "step": 44272
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2095371428571427e-06,
+      "loss": 1.6596,
+      "step": 44273
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2092685714285718e-06,
+      "loss": 0.9012,
+      "step": 44274
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2090000000000004e-06,
+      "loss": 1.2064,
+      "step": 44275
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2087314285714285e-06,
+      "loss": 1.0646,
+      "step": 44276
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.208462857142857e-06,
+      "loss": 1.306,
+      "step": 44277
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2081942857142857e-06,
+      "loss": 0.924,
+      "step": 44278
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2079257142857143e-06,
+      "loss": 0.8089,
+      "step": 44279
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.207657142857143e-06,
+      "loss": 1.2058,
+      "step": 44280
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2073885714285715e-06,
+      "loss": 0.9683,
+      "step": 44281
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.20712e-06,
+      "loss": 1.0994,
+      "step": 44282
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2068514285714287e-06,
+      "loss": 1.0946,
+      "step": 44283
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2065828571428573e-06,
+      "loss": 0.9572,
+      "step": 44284
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2063142857142855e-06,
+      "loss": 0.924,
+      "step": 44285
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.206045714285714e-06,
+      "loss": 1.1145,
+      "step": 44286
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.205777142857143e-06,
+      "loss": 1.0258,
+      "step": 44287
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2055085714285717e-06,
+      "loss": 1.444,
+      "step": 44288
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2052400000000003e-06,
+      "loss": 1.4725,
+      "step": 44289
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2049714285714285e-06,
+      "loss": 1.1852,
+      "step": 44290
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.204702857142857e-06,
+      "loss": 0.8349,
+      "step": 44291
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2044342857142857e-06,
+      "loss": 1.2684,
+      "step": 44292
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2041657142857143e-06,
+      "loss": 1.2453,
+      "step": 44293
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2038971428571433e-06,
+      "loss": 0.8595,
+      "step": 44294
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2036285714285715e-06,
+      "loss": 1.2308,
+      "step": 44295
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.20336e-06,
+      "loss": 1.1327,
+      "step": 44296
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2030914285714287e-06,
+      "loss": 0.9268,
+      "step": 44297
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2028228571428573e-06,
+      "loss": 1.053,
+      "step": 44298
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2025542857142855e-06,
+      "loss": 1.176,
+      "step": 44299
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.202285714285714e-06,
+      "loss": 1.3719,
+      "step": 44300
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.202017142857143e-06,
+      "loss": 1.1469,
+      "step": 44301
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2017485714285717e-06,
+      "loss": 1.6432,
+      "step": 44302
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2014800000000003e-06,
+      "loss": 1.1815,
+      "step": 44303
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2012114285714285e-06,
+      "loss": 1.1314,
+      "step": 44304
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.200942857142857e-06,
+      "loss": 1.15,
+      "step": 44305
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2006742857142857e-06,
+      "loss": 1.0668,
+      "step": 44306
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2004057142857147e-06,
+      "loss": 0.9524,
+      "step": 44307
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.200137142857143e-06,
+      "loss": 1.0459,
+      "step": 44308
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1998685714285715e-06,
+      "loss": 1.2111,
+      "step": 44309
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1996e-06,
+      "loss": 1.5364,
+      "step": 44310
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1993314285714287e-06,
+      "loss": 1.281,
+      "step": 44311
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1990628571428573e-06,
+      "loss": 0.8748,
+      "step": 44312
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1987942857142855e-06,
+      "loss": 1.5893,
+      "step": 44313
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1985257142857145e-06,
+      "loss": 0.609,
+      "step": 44314
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.198257142857143e-06,
+      "loss": 1.1402,
+      "step": 44315
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1979885714285717e-06,
+      "loss": 1.2642,
+      "step": 44316
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.19772e-06,
+      "loss": 1.1243,
+      "step": 44317
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1974514285714285e-06,
+      "loss": 1.5582,
+      "step": 44318
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.197182857142857e-06,
+      "loss": 1.299,
+      "step": 44319
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1969142857142857e-06,
+      "loss": 1.2489,
+      "step": 44320
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1966457142857147e-06,
+      "loss": 1.3774,
+      "step": 44321
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.196377142857143e-06,
+      "loss": 0.9238,
+      "step": 44322
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1961085714285715e-06,
+      "loss": 1.4683,
+      "step": 44323
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.19584e-06,
+      "loss": 0.9676,
+      "step": 44324
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1955714285714287e-06,
+      "loss": 0.6601,
+      "step": 44325
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1953028571428573e-06,
+      "loss": 1.5307,
+      "step": 44326
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.195034285714286e-06,
+      "loss": 0.8265,
+      "step": 44327
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1947657142857145e-06,
+      "loss": 1.0416,
+      "step": 44328
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.194497142857143e-06,
+      "loss": 0.881,
+      "step": 44329
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1942285714285717e-06,
+      "loss": 1.5748,
+      "step": 44330
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.19396e-06,
+      "loss": 0.9227,
+      "step": 44331
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1936914285714284e-06,
+      "loss": 0.9185,
+      "step": 44332
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.193422857142857e-06,
+      "loss": 0.9206,
+      "step": 44333
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.193154285714286e-06,
+      "loss": 0.471,
+      "step": 44334
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1928857142857146e-06,
+      "loss": 1.0335,
+      "step": 44335
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.192617142857143e-06,
+      "loss": 1.1341,
+      "step": 44336
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1923485714285714e-06,
+      "loss": 1.5972,
+      "step": 44337
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.19208e-06,
+      "loss": 0.6197,
+      "step": 44338
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1918114285714286e-06,
+      "loss": 0.7697,
+      "step": 44339
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1915428571428572e-06,
+      "loss": 0.8066,
+      "step": 44340
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.191274285714286e-06,
+      "loss": 1.1268,
+      "step": 44341
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1910057142857144e-06,
+      "loss": 1.639,
+      "step": 44342
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.190737142857143e-06,
+      "loss": 1.2219,
+      "step": 44343
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1904685714285716e-06,
+      "loss": 0.9235,
+      "step": 44344
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1902e-06,
+      "loss": 0.7168,
+      "step": 44345
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1899314285714284e-06,
+      "loss": 1.0189,
+      "step": 44346
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1896628571428574e-06,
+      "loss": 0.968,
+      "step": 44347
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.189394285714286e-06,
+      "loss": 1.2262,
+      "step": 44348
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.189125714285714e-06,
+      "loss": 1.0404,
+      "step": 44349
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.188857142857143e-06,
+      "loss": 0.9664,
+      "step": 44350
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1885885714285714e-06,
+      "loss": 1.511,
+      "step": 44351
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.18832e-06,
+      "loss": 1.1816,
+      "step": 44352
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1880514285714286e-06,
+      "loss": 1.5043,
+      "step": 44353
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.187782857142857e-06,
+      "loss": 1.036,
+      "step": 44354
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.187514285714286e-06,
+      "loss": 1.458,
+      "step": 44355
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1872457142857144e-06,
+      "loss": 0.7581,
+      "step": 44356
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.186977142857143e-06,
+      "loss": 1.3786,
+      "step": 44357
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1867085714285716e-06,
+      "loss": 0.971,
+      "step": 44358
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1864399999999998e-06,
+      "loss": 1.2658,
+      "step": 44359
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.186171428571429e-06,
+      "loss": 1.1009,
+      "step": 44360
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.1859028571428574e-06,
+      "loss": 1.016,
+      "step": 44361
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.185634285714286e-06,
+      "loss": 1.4856,
+      "step": 44362
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.185365714285714e-06,
+      "loss": 1.0615,
+      "step": 44363
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1850971428571428e-06,
+      "loss": 0.8745,
+      "step": 44364
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1848285714285714e-06,
+      "loss": 1.119,
+      "step": 44365
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.18456e-06,
+      "loss": 0.7658,
+      "step": 44366
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.184291428571429e-06,
+      "loss": 0.9429,
+      "step": 44367
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.184022857142857e-06,
+      "loss": 1.2607,
+      "step": 44368
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1837542857142858e-06,
+      "loss": 1.0383,
+      "step": 44369
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1834857142857144e-06,
+      "loss": 1.2828,
+      "step": 44370
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.183217142857143e-06,
+      "loss": 1.148,
+      "step": 44371
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.182948571428571e-06,
+      "loss": 1.1427,
+      "step": 44372
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.18268e-06,
+      "loss": 1.228,
+      "step": 44373
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1824114285714288e-06,
+      "loss": 1.0807,
+      "step": 44374
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1821428571428574e-06,
+      "loss": 1.2982,
+      "step": 44375
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.181874285714286e-06,
+      "loss": 1.0701,
+      "step": 44376
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.181605714285714e-06,
+      "loss": 1.0689,
+      "step": 44377
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1813371428571427e-06,
+      "loss": 1.2435,
+      "step": 44378
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1810685714285713e-06,
+      "loss": 1.0472,
+      "step": 44379
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1808000000000004e-06,
+      "loss": 1.4743,
+      "step": 44380
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1805314285714285e-06,
+      "loss": 0.7938,
+      "step": 44381
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.180262857142857e-06,
+      "loss": 1.0927,
+      "step": 44382
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1799942857142857e-06,
+      "loss": 0.997,
+      "step": 44383
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1797257142857143e-06,
+      "loss": 0.733,
+      "step": 44384
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.179457142857143e-06,
+      "loss": 0.8118,
+      "step": 44385
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1791885714285715e-06,
+      "loss": 1.2685,
+      "step": 44386
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.17892e-06,
+      "loss": 0.9366,
+      "step": 44387
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1786514285714287e-06,
+      "loss": 1.1473,
+      "step": 44388
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1783828571428573e-06,
+      "loss": 1.1495,
+      "step": 44389
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.178114285714286e-06,
+      "loss": 0.7759,
+      "step": 44390
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.177845714285714e-06,
+      "loss": 1.3651,
+      "step": 44391
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1775771428571427e-06,
+      "loss": 0.8773,
+      "step": 44392
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1773085714285717e-06,
+      "loss": 1.053,
+      "step": 44393
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1770400000000003e-06,
+      "loss": 1.2527,
+      "step": 44394
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1767714285714285e-06,
+      "loss": 0.94,
+      "step": 44395
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.176502857142857e-06,
+      "loss": 1.0729,
+      "step": 44396
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1762342857142857e-06,
+      "loss": 0.8552,
+      "step": 44397
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1759657142857143e-06,
+      "loss": 1.3669,
+      "step": 44398
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.175697142857143e-06,
+      "loss": 1.2711,
+      "step": 44399
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1754285714285715e-06,
+      "loss": 1.4155,
+      "step": 44400
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.17516e-06,
+      "loss": 1.2413,
+      "step": 44401
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1748914285714287e-06,
+      "loss": 0.9572,
+      "step": 44402
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1746228571428573e-06,
+      "loss": 1.1659,
+      "step": 44403
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1743542857142855e-06,
+      "loss": 0.9854,
+      "step": 44404
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.174085714285714e-06,
+      "loss": 0.8005,
+      "step": 44405
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.173817142857143e-06,
+      "loss": 0.8092,
+      "step": 44406
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1735485714285717e-06,
+      "loss": 1.031,
+      "step": 44407
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1732800000000003e-06,
+      "loss": 1.3078,
+      "step": 44408
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1730114285714285e-06,
+      "loss": 1.0073,
+      "step": 44409
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.172742857142857e-06,
+      "loss": 1.0446,
+      "step": 44410
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1724742857142857e-06,
+      "loss": 1.2985,
+      "step": 44411
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1722057142857143e-06,
+      "loss": 1.0949,
+      "step": 44412
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1719371428571433e-06,
+      "loss": 1.3936,
+      "step": 44413
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1716685714285715e-06,
+      "loss": 0.5982,
+      "step": 44414
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1714e-06,
+      "loss": 1.2785,
+      "step": 44415
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1711314285714287e-06,
+      "loss": 0.9879,
+      "step": 44416
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1708628571428573e-06,
+      "loss": 0.9954,
+      "step": 44417
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1705942857142855e-06,
+      "loss": 1.0923,
+      "step": 44418
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1703257142857145e-06,
+      "loss": 1.5787,
+      "step": 44419
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.170057142857143e-06,
+      "loss": 1.1424,
+      "step": 44420
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1697885714285717e-06,
+      "loss": 1.254,
+      "step": 44421
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1695200000000003e-06,
+      "loss": 1.2021,
+      "step": 44422
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1692514285714285e-06,
+      "loss": 1.2612,
+      "step": 44423
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.168982857142857e-06,
+      "loss": 1.054,
+      "step": 44424
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1687142857142857e-06,
+      "loss": 1.4259,
+      "step": 44425
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1684457142857147e-06,
+      "loss": 1.4117,
+      "step": 44426
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.168177142857143e-06,
+      "loss": 0.9906,
+      "step": 44427
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1679085714285715e-06,
+      "loss": 1.196,
+      "step": 44428
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.16764e-06,
+      "loss": 1.2099,
+      "step": 44429
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1673714285714287e-06,
+      "loss": 1.282,
+      "step": 44430
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1671028571428573e-06,
+      "loss": 0.8357,
+      "step": 44431
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.166834285714286e-06,
+      "loss": 1.0735,
+      "step": 44432
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1665657142857145e-06,
+      "loss": 1.425,
+      "step": 44433
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.166297142857143e-06,
+      "loss": 1.6097,
+      "step": 44434
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1660285714285717e-06,
+      "loss": 0.928,
+      "step": 44435
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.16576e-06,
+      "loss": 1.2342,
+      "step": 44436
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1654914285714284e-06,
+      "loss": 0.9369,
+      "step": 44437
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.165222857142857e-06,
+      "loss": 1.2254,
+      "step": 44438
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.164954285714286e-06,
+      "loss": 1.0344,
+      "step": 44439
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1646857142857146e-06,
+      "loss": 1.4541,
+      "step": 44440
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.164417142857143e-06,
+      "loss": 1.0564,
+      "step": 44441
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1641485714285714e-06,
+      "loss": 1.0012,
+      "step": 44442
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.16388e-06,
+      "loss": 1.252,
+      "step": 44443
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1636114285714286e-06,
+      "loss": 1.238,
+      "step": 44444
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1633428571428572e-06,
+      "loss": 1.4106,
+      "step": 44445
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.163074285714286e-06,
+      "loss": 0.9036,
+      "step": 44446
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1628057142857144e-06,
+      "loss": 1.3791,
+      "step": 44447
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.162537142857143e-06,
+      "loss": 1.4003,
+      "step": 44448
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1622685714285716e-06,
+      "loss": 1.3594,
+      "step": 44449
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.162e-06,
+      "loss": 1.1689,
+      "step": 44450
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1617314285714284e-06,
+      "loss": 1.1051,
+      "step": 44451
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1614628571428574e-06,
+      "loss": 0.7283,
+      "step": 44452
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.161194285714286e-06,
+      "loss": 1.0293,
+      "step": 44453
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1609257142857146e-06,
+      "loss": 0.8046,
+      "step": 44454
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.160657142857143e-06,
+      "loss": 1.0709,
+      "step": 44455
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1603885714285714e-06,
+      "loss": 1.2364,
+      "step": 44456
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.16012e-06,
+      "loss": 1.0306,
+      "step": 44457
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1598514285714286e-06,
+      "loss": 1.0448,
+      "step": 44458
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.159582857142857e-06,
+      "loss": 1.2574,
+      "step": 44459
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.159314285714286e-06,
+      "loss": 0.9751,
+      "step": 44460
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1590457142857144e-06,
+      "loss": 1.3541,
+      "step": 44461
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.158777142857143e-06,
+      "loss": 1.0564,
+      "step": 44462
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1585085714285716e-06,
+      "loss": 1.3163,
+      "step": 44463
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1582399999999998e-06,
+      "loss": 1.4667,
+      "step": 44464
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.157971428571429e-06,
+      "loss": 1.3917,
+      "step": 44465
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1577028571428574e-06,
+      "loss": 0.9373,
+      "step": 44466
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.157434285714286e-06,
+      "loss": 1.1171,
+      "step": 44467
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.157165714285714e-06,
+      "loss": 1.1743,
+      "step": 44468
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1568971428571428e-06,
+      "loss": 0.8932,
+      "step": 44469
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1566285714285714e-06,
+      "loss": 1.3302,
+      "step": 44470
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.15636e-06,
+      "loss": 1.1978,
+      "step": 44471
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.156091428571429e-06,
+      "loss": 0.8119,
+      "step": 44472
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.155822857142857e-06,
+      "loss": 0.816,
+      "step": 44473
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1555542857142858e-06,
+      "loss": 0.9622,
+      "step": 44474
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1552857142857144e-06,
+      "loss": 1.1829,
+      "step": 44475
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.155017142857143e-06,
+      "loss": 0.7075,
+      "step": 44476
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1547485714285716e-06,
+      "loss": 0.7278,
+      "step": 44477
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.15448e-06,
+      "loss": 1.2351,
+      "step": 44478
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1542114285714288e-06,
+      "loss": 1.1426,
+      "step": 44479
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1539428571428574e-06,
+      "loss": 1.1827,
+      "step": 44480
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.153674285714286e-06,
+      "loss": 1.0324,
+      "step": 44481
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.153405714285714e-06,
+      "loss": 1.188,
+      "step": 44482
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1531371428571427e-06,
+      "loss": 0.9699,
+      "step": 44483
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1528685714285713e-06,
+      "loss": 0.9452,
+      "step": 44484
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1526000000000004e-06,
+      "loss": 1.0033,
+      "step": 44485
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.152331428571429e-06,
+      "loss": 1.0088,
+      "step": 44486
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.152062857142857e-06,
+      "loss": 0.9583,
+      "step": 44487
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1517942857142857e-06,
+      "loss": 1.4553,
+      "step": 44488
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1515257142857143e-06,
+      "loss": 1.031,
+      "step": 44489
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.151257142857143e-06,
+      "loss": 0.9575,
+      "step": 44490
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.150988571428571e-06,
+      "loss": 1.5392,
+      "step": 44491
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.15072e-06,
+      "loss": 1.0919,
+      "step": 44492
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1504514285714287e-06,
+      "loss": 1.5372,
+      "step": 44493
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1501828571428573e-06,
+      "loss": 1.0491,
+      "step": 44494
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.149914285714286e-06,
+      "loss": 1.2057,
+      "step": 44495
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.149645714285714e-06,
+      "loss": 1.2756,
+      "step": 44496
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1493771428571427e-06,
+      "loss": 0.8356,
+      "step": 44497
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1491085714285717e-06,
+      "loss": 0.6619,
+      "step": 44498
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1488400000000003e-06,
+      "loss": 0.8522,
+      "step": 44499
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.148571428571429e-06,
+      "loss": 0.9297,
+      "step": 44500
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.148302857142857e-06,
+      "loss": 1.5478,
+      "step": 44501
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1480342857142857e-06,
+      "loss": 1.0524,
+      "step": 44502
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1477657142857143e-06,
+      "loss": 1.1728,
+      "step": 44503
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.147497142857143e-06,
+      "loss": 1.3418,
+      "step": 44504
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1472285714285715e-06,
+      "loss": 1.1467,
+      "step": 44505
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.14696e-06,
+      "loss": 0.8909,
+      "step": 44506
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1466914285714287e-06,
+      "loss": 1.4444,
+      "step": 44507
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1464228571428573e-06,
+      "loss": 1.1637,
+      "step": 44508
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.146154285714286e-06,
+      "loss": 0.933,
+      "step": 44509
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.145885714285714e-06,
+      "loss": 0.9063,
+      "step": 44510
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.145617142857143e-06,
+      "loss": 0.9996,
+      "step": 44511
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1453485714285717e-06,
+      "loss": 0.7693,
+      "step": 44512
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1450800000000003e-06,
+      "loss": 1.2925,
+      "step": 44513
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1448114285714285e-06,
+      "loss": 0.868,
+      "step": 44514
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.144542857142857e-06,
+      "loss": 1.1736,
+      "step": 44515
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1442742857142857e-06,
+      "loss": 1.4168,
+      "step": 44516
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1440057142857143e-06,
+      "loss": 1.0432,
+      "step": 44517
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1437371428571433e-06,
+      "loss": 0.7051,
+      "step": 44518
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1434685714285715e-06,
+      "loss": 1.1257,
+      "step": 44519
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1432e-06,
+      "loss": 0.796,
+      "step": 44520
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1429314285714287e-06,
+      "loss": 0.9713,
+      "step": 44521
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1426628571428573e-06,
+      "loss": 0.8632,
+      "step": 44522
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1423942857142855e-06,
+      "loss": 0.9421,
+      "step": 44523
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.142125714285714e-06,
+      "loss": 1.1201,
+      "step": 44524
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.141857142857143e-06,
+      "loss": 1.1803,
+      "step": 44525
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1415885714285717e-06,
+      "loss": 0.8606,
+      "step": 44526
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1413200000000003e-06,
+      "loss": 1.311,
+      "step": 44527
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1410514285714285e-06,
+      "loss": 0.9931,
+      "step": 44528
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.140782857142857e-06,
+      "loss": 1.0664,
+      "step": 44529
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1405142857142857e-06,
+      "loss": 1.2588,
+      "step": 44530
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1402457142857147e-06,
+      "loss": 1.0173,
+      "step": 44531
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1399771428571433e-06,
+      "loss": 0.9717,
+      "step": 44532
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1397085714285715e-06,
+      "loss": 1.1356,
+      "step": 44533
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.13944e-06,
+      "loss": 1.3315,
+      "step": 44534
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1391714285714287e-06,
+      "loss": 1.1452,
+      "step": 44535
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1389028571428573e-06,
+      "loss": 0.8129,
+      "step": 44536
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.1386342857142854e-06,
+      "loss": 1.1537,
+      "step": 44537
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1383657142857145e-06,
+      "loss": 0.9698,
+      "step": 44538
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.138097142857143e-06,
+      "loss": 1.0856,
+      "step": 44539
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1378285714285716e-06,
+      "loss": 1.1322,
+      "step": 44540
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1375600000000002e-06,
+      "loss": 1.003,
+      "step": 44541
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1372914285714284e-06,
+      "loss": 1.3378,
+      "step": 44542
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.137022857142857e-06,
+      "loss": 1.0851,
+      "step": 44543
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1367542857142856e-06,
+      "loss": 1.321,
+      "step": 44544
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1364857142857146e-06,
+      "loss": 1.1109,
+      "step": 44545
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.136217142857143e-06,
+      "loss": 1.3217,
+      "step": 44546
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1359485714285714e-06,
+      "loss": 0.8638,
+      "step": 44547
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.13568e-06,
+      "loss": 0.856,
+      "step": 44548
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1354114285714286e-06,
+      "loss": 1.2396,
+      "step": 44549
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1351428571428572e-06,
+      "loss": 0.9996,
+      "step": 44550
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.134874285714286e-06,
+      "loss": 1.0152,
+      "step": 44551
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1346057142857144e-06,
+      "loss": 1.3278,
+      "step": 44552
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.134337142857143e-06,
+      "loss": 0.9085,
+      "step": 44553
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1340685714285716e-06,
+      "loss": 1.1163,
+      "step": 44554
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1338e-06,
+      "loss": 1.5064,
+      "step": 44555
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1335314285714284e-06,
+      "loss": 0.7002,
+      "step": 44556
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.133262857142857e-06,
+      "loss": 1.2737,
+      "step": 44557
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.132994285714286e-06,
+      "loss": 0.9268,
+      "step": 44558
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1327257142857146e-06,
+      "loss": 0.8464,
+      "step": 44559
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.132457142857143e-06,
+      "loss": 1.2375,
+      "step": 44560
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1321885714285714e-06,
+      "loss": 1.1117,
+      "step": 44561
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.13192e-06,
+      "loss": 0.9507,
+      "step": 44562
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1316514285714286e-06,
+      "loss": 1.2304,
+      "step": 44563
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1313828571428576e-06,
+      "loss": 1.1681,
+      "step": 44564
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.131114285714286e-06,
+      "loss": 1.2509,
+      "step": 44565
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1308457142857144e-06,
+      "loss": 1.15,
+      "step": 44566
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.130577142857143e-06,
+      "loss": 1.4872,
+      "step": 44567
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1303085714285716e-06,
+      "loss": 0.6175,
+      "step": 44568
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1300399999999998e-06,
+      "loss": 1.6567,
+      "step": 44569
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1297714285714284e-06,
+      "loss": 1.0558,
+      "step": 44570
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1295028571428574e-06,
+      "loss": 1.1458,
+      "step": 44571
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.129234285714286e-06,
+      "loss": 1.3672,
+      "step": 44572
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1289657142857146e-06,
+      "loss": 1.183,
+      "step": 44573
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1286971428571428e-06,
+      "loss": 1.4734,
+      "step": 44574
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1284285714285714e-06,
+      "loss": 0.9854,
+      "step": 44575
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.12816e-06,
+      "loss": 1.3567,
+      "step": 44576
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1278914285714286e-06,
+      "loss": 0.9087,
+      "step": 44577
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.127622857142857e-06,
+      "loss": 1.0782,
+      "step": 44578
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1273542857142858e-06,
+      "loss": 0.9536,
+      "step": 44579
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1270857142857144e-06,
+      "loss": 1.3147,
+      "step": 44580
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.126817142857143e-06,
+      "loss": 1.5747,
+      "step": 44581
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1265485714285716e-06,
+      "loss": 1.1445,
+      "step": 44582
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1262799999999997e-06,
+      "loss": 0.9686,
+      "step": 44583
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1260114285714288e-06,
+      "loss": 1.1223,
+      "step": 44584
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1257428571428574e-06,
+      "loss": 0.9894,
+      "step": 44585
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.125474285714286e-06,
+      "loss": 1.1139,
+      "step": 44586
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1252057142857146e-06,
+      "loss": 0.7772,
+      "step": 44587
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1249371428571427e-06,
+      "loss": 1.039,
+      "step": 44588
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1246685714285713e-06,
+      "loss": 1.0337,
+      "step": 44589
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1244e-06,
+      "loss": 0.6986,
+      "step": 44590
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.124131428571429e-06,
+      "loss": 1.234,
+      "step": 44591
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.123862857142857e-06,
+      "loss": 1.2714,
+      "step": 44592
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1235942857142857e-06,
+      "loss": 1.1853,
+      "step": 44593
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1233257142857143e-06,
+      "loss": 1.215,
+      "step": 44594
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.123057142857143e-06,
+      "loss": 1.3222,
+      "step": 44595
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1227885714285715e-06,
+      "loss": 1.0467,
+      "step": 44596
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.12252e-06,
+      "loss": 1.0463,
+      "step": 44597
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1222514285714287e-06,
+      "loss": 0.9785,
+      "step": 44598
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1219828571428573e-06,
+      "loss": 1.0048,
+      "step": 44599
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.121714285714286e-06,
+      "loss": 1.2176,
+      "step": 44600
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.121445714285714e-06,
+      "loss": 0.9485,
+      "step": 44601
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1211771428571427e-06,
+      "loss": 0.9548,
+      "step": 44602
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1209085714285713e-06,
+      "loss": 1.242,
+      "step": 44603
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1206400000000003e-06,
+      "loss": 0.8068,
+      "step": 44604
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.120371428571429e-06,
+      "loss": 1.2196,
+      "step": 44605
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.120102857142857e-06,
+      "loss": 1.3317,
+      "step": 44606
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1198342857142857e-06,
+      "loss": 1.2552,
+      "step": 44607
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1195657142857143e-06,
+      "loss": 1.2427,
+      "step": 44608
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.119297142857143e-06,
+      "loss": 1.2369,
+      "step": 44609
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1190285714285715e-06,
+      "loss": 0.9666,
+      "step": 44610
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.11876e-06,
+      "loss": 1.472,
+      "step": 44611
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1184914285714287e-06,
+      "loss": 1.1737,
+      "step": 44612
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1182228571428573e-06,
+      "loss": 1.1583,
+      "step": 44613
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.117954285714286e-06,
+      "loss": 1.0428,
+      "step": 44614
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.117685714285714e-06,
+      "loss": 1.1944,
+      "step": 44615
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1174171428571427e-06,
+      "loss": 1.0336,
+      "step": 44616
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1171485714285717e-06,
+      "loss": 1.0186,
+      "step": 44617
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1168800000000003e-06,
+      "loss": 1.4008,
+      "step": 44618
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.116611428571429e-06,
+      "loss": 1.2892,
+      "step": 44619
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.116342857142857e-06,
+      "loss": 1.649,
+      "step": 44620
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1160742857142857e-06,
+      "loss": 0.8546,
+      "step": 44621
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1158057142857143e-06,
+      "loss": 0.9115,
+      "step": 44622
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.115537142857143e-06,
+      "loss": 1.1098,
+      "step": 44623
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1152685714285715e-06,
+      "loss": 1.0483,
+      "step": 44624
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.115e-06,
+      "loss": 1.2777,
+      "step": 44625
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1147314285714287e-06,
+      "loss": 1.3504,
+      "step": 44626
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1144628571428573e-06,
+      "loss": 1.1608,
+      "step": 44627
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.114194285714286e-06,
+      "loss": 1.0775,
+      "step": 44628
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.113925714285714e-06,
+      "loss": 1.0948,
+      "step": 44629
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.113657142857143e-06,
+      "loss": 1.1405,
+      "step": 44630
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1133885714285717e-06,
+      "loss": 1.0111,
+      "step": 44631
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1131200000000003e-06,
+      "loss": 1.1833,
+      "step": 44632
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1128514285714285e-06,
+      "loss": 1.093,
+      "step": 44633
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.112582857142857e-06,
+      "loss": 1.0254,
+      "step": 44634
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1123142857142857e-06,
+      "loss": 1.1367,
+      "step": 44635
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1120457142857143e-06,
+      "loss": 1.155,
+      "step": 44636
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1117771428571433e-06,
+      "loss": 1.271,
+      "step": 44637
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1115085714285715e-06,
+      "loss": 1.1133,
+      "step": 44638
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.11124e-06,
+      "loss": 0.9204,
+      "step": 44639
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1109714285714287e-06,
+      "loss": 1.189,
+      "step": 44640
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1107028571428572e-06,
+      "loss": 1.3713,
+      "step": 44641
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1104342857142854e-06,
+      "loss": 1.0489,
+      "step": 44642
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1101657142857144e-06,
+      "loss": 1.4654,
+      "step": 44643
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.109897142857143e-06,
+      "loss": 0.9749,
+      "step": 44644
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1096285714285716e-06,
+      "loss": 1.6055,
+      "step": 44645
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1093600000000002e-06,
+      "loss": 1.1159,
+      "step": 44646
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1090914285714284e-06,
+      "loss": 0.9792,
+      "step": 44647
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.108822857142857e-06,
+      "loss": 0.8723,
+      "step": 44648
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1085542857142856e-06,
+      "loss": 1.3068,
+      "step": 44649
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1082857142857146e-06,
+      "loss": 0.9973,
+      "step": 44650
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1080171428571432e-06,
+      "loss": 0.8712,
+      "step": 44651
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1077485714285714e-06,
+      "loss": 1.1638,
+      "step": 44652
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.10748e-06,
+      "loss": 1.0194,
+      "step": 44653
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1072114285714286e-06,
+      "loss": 0.8089,
+      "step": 44654
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1069428571428572e-06,
+      "loss": 1.3575,
+      "step": 44655
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.106674285714286e-06,
+      "loss": 1.2034,
+      "step": 44656
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1064057142857144e-06,
+      "loss": 0.9864,
+      "step": 44657
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.106137142857143e-06,
+      "loss": 0.9003,
+      "step": 44658
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1058685714285716e-06,
+      "loss": 1.3187,
+      "step": 44659
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1056000000000002e-06,
+      "loss": 1.3075,
+      "step": 44660
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1053314285714284e-06,
+      "loss": 1.1363,
+      "step": 44661
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.105062857142857e-06,
+      "loss": 1.4387,
+      "step": 44662
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.104794285714286e-06,
+      "loss": 1.0503,
+      "step": 44663
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1045257142857146e-06,
+      "loss": 0.9967,
+      "step": 44664
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.104257142857143e-06,
+      "loss": 0.7306,
+      "step": 44665
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1039885714285714e-06,
+      "loss": 1.1983,
+      "step": 44666
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.10372e-06,
+      "loss": 1.4604,
+      "step": 44667
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1034514285714286e-06,
+      "loss": 0.9705,
+      "step": 44668
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.103182857142857e-06,
+      "loss": 1.0935,
+      "step": 44669
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.102914285714286e-06,
+      "loss": 1.2407,
+      "step": 44670
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1026457142857144e-06,
+      "loss": 1.5845,
+      "step": 44671
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.102377142857143e-06,
+      "loss": 1.6074,
+      "step": 44672
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1021085714285716e-06,
+      "loss": 1.1886,
+      "step": 44673
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1018399999999998e-06,
+      "loss": 0.8217,
+      "step": 44674
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1015714285714284e-06,
+      "loss": 0.683,
+      "step": 44675
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1013028571428574e-06,
+      "loss": 1.4448,
+      "step": 44676
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.101034285714286e-06,
+      "loss": 1.0875,
+      "step": 44677
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1007657142857146e-06,
+      "loss": 1.3367,
+      "step": 44678
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1004971428571428e-06,
+      "loss": 0.9236,
+      "step": 44679
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.1002285714285714e-06,
+      "loss": 1.4076,
+      "step": 44680
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.09996e-06,
+      "loss": 1.2269,
+      "step": 44681
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0996914285714286e-06,
+      "loss": 1.1608,
+      "step": 44682
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0994228571428576e-06,
+      "loss": 0.9111,
+      "step": 44683
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0991542857142858e-06,
+      "loss": 1.3645,
+      "step": 44684
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0988857142857144e-06,
+      "loss": 1.0478,
+      "step": 44685
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.098617142857143e-06,
+      "loss": 1.4917,
+      "step": 44686
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0983485714285716e-06,
+      "loss": 0.7378,
+      "step": 44687
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0980799999999997e-06,
+      "loss": 1.2557,
+      "step": 44688
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0978114285714288e-06,
+      "loss": 1.2689,
+      "step": 44689
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0975428571428574e-06,
+      "loss": 1.3743,
+      "step": 44690
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.097274285714286e-06,
+      "loss": 1.1308,
+      "step": 44691
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0970057142857146e-06,
+      "loss": 1.0125,
+      "step": 44692
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0967371428571427e-06,
+      "loss": 1.6353,
+      "step": 44693
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0964685714285713e-06,
+      "loss": 0.7131,
+      "step": 44694
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0962e-06,
+      "loss": 0.6135,
+      "step": 44695
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.095931428571429e-06,
+      "loss": 1.3699,
+      "step": 44696
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.095662857142857e-06,
+      "loss": 1.072,
+      "step": 44697
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0953942857142857e-06,
+      "loss": 1.3667,
+      "step": 44698
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0951257142857143e-06,
+      "loss": 1.176,
+      "step": 44699
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.094857142857143e-06,
+      "loss": 0.9182,
+      "step": 44700
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0945885714285715e-06,
+      "loss": 1.0545,
+      "step": 44701
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.09432e-06,
+      "loss": 1.2135,
+      "step": 44702
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0940514285714287e-06,
+      "loss": 1.1692,
+      "step": 44703
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0937828571428573e-06,
+      "loss": 0.8073,
+      "step": 44704
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.093514285714286e-06,
+      "loss": 1.276,
+      "step": 44705
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0932457142857145e-06,
+      "loss": 1.0587,
+      "step": 44706
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0929771428571427e-06,
+      "loss": 0.9993,
+      "step": 44707
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0927085714285713e-06,
+      "loss": 1.2146,
+      "step": 44708
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0924400000000003e-06,
+      "loss": 1.2999,
+      "step": 44709
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.092171428571429e-06,
+      "loss": 1.2166,
+      "step": 44710
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.091902857142857e-06,
+      "loss": 1.453,
+      "step": 44711
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.0916342857142857e-06,
+      "loss": 1.3715,
+      "step": 44712
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0913657142857143e-06,
+      "loss": 1.365,
+      "step": 44713
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.091097142857143e-06,
+      "loss": 0.9435,
+      "step": 44714
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0908285714285715e-06,
+      "loss": 1.3079,
+      "step": 44715
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.09056e-06,
+      "loss": 1.1592,
+      "step": 44716
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0902914285714287e-06,
+      "loss": 0.924,
+      "step": 44717
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0900228571428573e-06,
+      "loss": 0.9773,
+      "step": 44718
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.089754285714286e-06,
+      "loss": 0.9997,
+      "step": 44719
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.089485714285714e-06,
+      "loss": 1.2382,
+      "step": 44720
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0892171428571427e-06,
+      "loss": 1.2933,
+      "step": 44721
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0889485714285717e-06,
+      "loss": 1.2293,
+      "step": 44722
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0886800000000003e-06,
+      "loss": 0.9875,
+      "step": 44723
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.088411428571429e-06,
+      "loss": 1.0244,
+      "step": 44724
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.088142857142857e-06,
+      "loss": 1.1222,
+      "step": 44725
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0878742857142857e-06,
+      "loss": 0.8885,
+      "step": 44726
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0876057142857143e-06,
+      "loss": 1.4387,
+      "step": 44727
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.087337142857143e-06,
+      "loss": 1.0839,
+      "step": 44728
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0870685714285715e-06,
+      "loss": 1.0387,
+      "step": 44729
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0868e-06,
+      "loss": 0.972,
+      "step": 44730
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0865314285714287e-06,
+      "loss": 1.0302,
+      "step": 44731
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0862628571428573e-06,
+      "loss": 1.1788,
+      "step": 44732
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.085994285714286e-06,
+      "loss": 1.0658,
+      "step": 44733
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.085725714285714e-06,
+      "loss": 0.9177,
+      "step": 44734
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.085457142857143e-06,
+      "loss": 1.2698,
+      "step": 44735
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0851885714285717e-06,
+      "loss": 1.0986,
+      "step": 44736
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0849200000000003e-06,
+      "loss": 1.4929,
+      "step": 44737
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.084651428571429e-06,
+      "loss": 1.3953,
+      "step": 44738
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.084382857142857e-06,
+      "loss": 1.2322,
+      "step": 44739
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0841142857142857e-06,
+      "loss": 1.1178,
+      "step": 44740
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0838457142857143e-06,
+      "loss": 1.4251,
+      "step": 44741
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0835771428571433e-06,
+      "loss": 1.4425,
+      "step": 44742
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0833085714285714e-06,
+      "loss": 0.9289,
+      "step": 44743
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.08304e-06,
+      "loss": 0.7866,
+      "step": 44744
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0827714285714286e-06,
+      "loss": 1.4262,
+      "step": 44745
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0825028571428572e-06,
+      "loss": 1.1255,
+      "step": 44746
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.082234285714286e-06,
+      "loss": 1.3104,
+      "step": 44747
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.081965714285714e-06,
+      "loss": 1.2105,
+      "step": 44748
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.081697142857143e-06,
+      "loss": 0.7844,
+      "step": 44749
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0814285714285716e-06,
+      "loss": 1.4325,
+      "step": 44750
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0811600000000002e-06,
+      "loss": 0.784,
+      "step": 44751
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0808914285714284e-06,
+      "loss": 1.0549,
+      "step": 44752
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.080622857142857e-06,
+      "loss": 1.3732,
+      "step": 44753
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0803542857142856e-06,
+      "loss": 1.0842,
+      "step": 44754
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0800857142857146e-06,
+      "loss": 1.5087,
+      "step": 44755
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0798171428571432e-06,
+      "loss": 1.2084,
+      "step": 44756
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0795485714285714e-06,
+      "loss": 0.9412,
+      "step": 44757
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.07928e-06,
+      "loss": 1.1566,
+      "step": 44758
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0790114285714286e-06,
+      "loss": 1.7353,
+      "step": 44759
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0787428571428572e-06,
+      "loss": 1.3716,
+      "step": 44760
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0784742857142854e-06,
+      "loss": 0.8303,
+      "step": 44761
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0782057142857144e-06,
+      "loss": 1.201,
+      "step": 44762
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.077937142857143e-06,
+      "loss": 1.5628,
+      "step": 44763
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0776685714285716e-06,
+      "loss": 0.8257,
+      "step": 44764
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0774000000000002e-06,
+      "loss": 1.1324,
+      "step": 44765
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0771314285714284e-06,
+      "loss": 1.1282,
+      "step": 44766
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.076862857142857e-06,
+      "loss": 1.1622,
+      "step": 44767
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.076594285714286e-06,
+      "loss": 0.9395,
+      "step": 44768
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0763257142857146e-06,
+      "loss": 1.218,
+      "step": 44769
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0760571428571432e-06,
+      "loss": 1.1835,
+      "step": 44770
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0757885714285714e-06,
+      "loss": 1.1637,
+      "step": 44771
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.07552e-06,
+      "loss": 0.8451,
+      "step": 44772
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0752514285714286e-06,
+      "loss": 1.0364,
+      "step": 44773
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.074982857142857e-06,
+      "loss": 1.5185,
+      "step": 44774
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.074714285714286e-06,
+      "loss": 1.2501,
+      "step": 44775
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0744457142857144e-06,
+      "loss": 1.5106,
+      "step": 44776
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.074177142857143e-06,
+      "loss": 1.1712,
+      "step": 44777
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0739085714285716e-06,
+      "loss": 1.0962,
+      "step": 44778
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.07364e-06,
+      "loss": 1.2499,
+      "step": 44779
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0733714285714284e-06,
+      "loss": 0.7909,
+      "step": 44780
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.073102857142857e-06,
+      "loss": 0.7562,
+      "step": 44781
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.072834285714286e-06,
+      "loss": 1.1663,
+      "step": 44782
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0725657142857146e-06,
+      "loss": 1.0795,
+      "step": 44783
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0722971428571428e-06,
+      "loss": 0.9611,
+      "step": 44784
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0720285714285714e-06,
+      "loss": 0.8794,
+      "step": 44785
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.07176e-06,
+      "loss": 1.4176,
+      "step": 44786
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0714914285714286e-06,
+      "loss": 1.2282,
+      "step": 44787
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0712228571428576e-06,
+      "loss": 1.1925,
+      "step": 44788
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0709542857142858e-06,
+      "loss": 0.707,
+      "step": 44789
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0706857142857144e-06,
+      "loss": 1.1035,
+      "step": 44790
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.070417142857143e-06,
+      "loss": 0.9064,
+      "step": 44791
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0701485714285716e-06,
+      "loss": 1.0195,
+      "step": 44792
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.06988e-06,
+      "loss": 1.3846,
+      "step": 44793
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0696114285714283e-06,
+      "loss": 0.9819,
+      "step": 44794
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0693428571428574e-06,
+      "loss": 1.3039,
+      "step": 44795
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.069074285714286e-06,
+      "loss": 1.4492,
+      "step": 44796
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0688057142857146e-06,
+      "loss": 1.0565,
+      "step": 44797
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0685371428571427e-06,
+      "loss": 0.8787,
+      "step": 44798
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0682685714285713e-06,
+      "loss": 1.2223,
+      "step": 44799
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.068e-06,
+      "loss": 0.8818,
+      "step": 44800
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0677314285714285e-06,
+      "loss": 1.0661,
+      "step": 44801
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0674628571428576e-06,
+      "loss": 1.0861,
+      "step": 44802
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0671942857142857e-06,
+      "loss": 0.9305,
+      "step": 44803
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0669257142857143e-06,
+      "loss": 0.5134,
+      "step": 44804
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.066657142857143e-06,
+      "loss": 1.408,
+      "step": 44805
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0663885714285715e-06,
+      "loss": 0.8737,
+      "step": 44806
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0661199999999997e-06,
+      "loss": 1.2035,
+      "step": 44807
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0658514285714287e-06,
+      "loss": 0.7509,
+      "step": 44808
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0655828571428573e-06,
+      "loss": 1.4351,
+      "step": 44809
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.065314285714286e-06,
+      "loss": 1.3031,
+      "step": 44810
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0650457142857145e-06,
+      "loss": 0.9296,
+      "step": 44811
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0647771428571427e-06,
+      "loss": 0.8846,
+      "step": 44812
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0645085714285713e-06,
+      "loss": 1.6071,
+      "step": 44813
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.06424e-06,
+      "loss": 1.3196,
+      "step": 44814
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.063971428571429e-06,
+      "loss": 1.2928,
+      "step": 44815
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.063702857142857e-06,
+      "loss": 0.8995,
+      "step": 44816
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0634342857142857e-06,
+      "loss": 1.1418,
+      "step": 44817
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0631657142857143e-06,
+      "loss": 1.1673,
+      "step": 44818
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.062897142857143e-06,
+      "loss": 1.1791,
+      "step": 44819
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0626285714285715e-06,
+      "loss": 1.2656,
+      "step": 44820
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.06236e-06,
+      "loss": 1.08,
+      "step": 44821
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0620914285714287e-06,
+      "loss": 1.1886,
+      "step": 44822
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0618228571428573e-06,
+      "loss": 1.4705,
+      "step": 44823
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.061554285714286e-06,
+      "loss": 1.4139,
+      "step": 44824
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0612857142857145e-06,
+      "loss": 1.4808,
+      "step": 44825
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0610171428571427e-06,
+      "loss": 0.8664,
+      "step": 44826
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0607485714285713e-06,
+      "loss": 0.8148,
+      "step": 44827
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0604800000000003e-06,
+      "loss": 1.2161,
+      "step": 44828
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.060211428571429e-06,
+      "loss": 1.0512,
+      "step": 44829
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.059942857142857e-06,
+      "loss": 0.8214,
+      "step": 44830
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0596742857142857e-06,
+      "loss": 1.0357,
+      "step": 44831
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0594057142857143e-06,
+      "loss": 1.192,
+      "step": 44832
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.059137142857143e-06,
+      "loss": 0.8621,
+      "step": 44833
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0588685714285715e-06,
+      "loss": 0.9969,
+      "step": 44834
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0586e-06,
+      "loss": 0.8393,
+      "step": 44835
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0583314285714287e-06,
+      "loss": 1.0029,
+      "step": 44836
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0580628571428573e-06,
+      "loss": 0.8741,
+      "step": 44837
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.057794285714286e-06,
+      "loss": 0.9751,
+      "step": 44838
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.057525714285714e-06,
+      "loss": 1.0862,
+      "step": 44839
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0572571428571427e-06,
+      "loss": 1.1518,
+      "step": 44840
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0569885714285717e-06,
+      "loss": 0.6164,
+      "step": 44841
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0567200000000003e-06,
+      "loss": 1.3418,
+      "step": 44842
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.056451428571429e-06,
+      "loss": 0.824,
+      "step": 44843
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.056182857142857e-06,
+      "loss": 0.8079,
+      "step": 44844
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0559142857142856e-06,
+      "loss": 0.9907,
+      "step": 44845
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0556457142857142e-06,
+      "loss": 0.7642,
+      "step": 44846
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.055377142857143e-06,
+      "loss": 1.0933,
+      "step": 44847
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0551085714285714e-06,
+      "loss": 1.1851,
+      "step": 44848
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.05484e-06,
+      "loss": 0.9839,
+      "step": 44849
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0545714285714286e-06,
+      "loss": 0.9304,
+      "step": 44850
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0543028571428572e-06,
+      "loss": 0.9133,
+      "step": 44851
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.054034285714286e-06,
+      "loss": 1.4138,
+      "step": 44852
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.053765714285714e-06,
+      "loss": 1.4699,
+      "step": 44853
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.053497142857143e-06,
+      "loss": 1.1194,
+      "step": 44854
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0532285714285716e-06,
+      "loss": 1.2166,
+      "step": 44855
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0529600000000002e-06,
+      "loss": 1.1495,
+      "step": 44856
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.052691428571429e-06,
+      "loss": 0.7651,
+      "step": 44857
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.052422857142857e-06,
+      "loss": 1.38,
+      "step": 44858
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0521542857142856e-06,
+      "loss": 0.6879,
+      "step": 44859
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0518857142857142e-06,
+      "loss": 0.9444,
+      "step": 44860
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0516171428571432e-06,
+      "loss": 1.3152,
+      "step": 44861
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0513485714285714e-06,
+      "loss": 1.0493,
+      "step": 44862
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.05108e-06,
+      "loss": 1.1374,
+      "step": 44863
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0508114285714286e-06,
+      "loss": 1.1287,
+      "step": 44864
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0505428571428572e-06,
+      "loss": 1.3444,
+      "step": 44865
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.050274285714286e-06,
+      "loss": 1.0174,
+      "step": 44866
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0500057142857144e-06,
+      "loss": 0.9459,
+      "step": 44867
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.049737142857143e-06,
+      "loss": 1.0882,
+      "step": 44868
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0494685714285716e-06,
+      "loss": 1.5398,
+      "step": 44869
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0492000000000002e-06,
+      "loss": 1.448,
+      "step": 44870
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0489314285714284e-06,
+      "loss": 1.1743,
+      "step": 44871
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.048662857142857e-06,
+      "loss": 0.9975,
+      "step": 44872
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0483942857142856e-06,
+      "loss": 1.2392,
+      "step": 44873
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0481257142857146e-06,
+      "loss": 1.0843,
+      "step": 44874
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0478571428571432e-06,
+      "loss": 0.8929,
+      "step": 44875
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0475885714285714e-06,
+      "loss": 1.0666,
+      "step": 44876
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.04732e-06,
+      "loss": 1.2932,
+      "step": 44877
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0470514285714286e-06,
+      "loss": 0.9449,
+      "step": 44878
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.046782857142857e-06,
+      "loss": 0.898,
+      "step": 44879
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.046514285714286e-06,
+      "loss": 0.9657,
+      "step": 44880
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0462457142857144e-06,
+      "loss": 1.1259,
+      "step": 44881
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.045977142857143e-06,
+      "loss": 0.9907,
+      "step": 44882
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0457085714285716e-06,
+      "loss": 1.0989,
+      "step": 44883
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.04544e-06,
+      "loss": 1.3537,
+      "step": 44884
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.0451714285714284e-06,
+      "loss": 1.2878,
+      "step": 44885
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.044902857142857e-06,
+      "loss": 1.2191,
+      "step": 44886
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.044634285714286e-06,
+      "loss": 1.0686,
+      "step": 44887
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0443657142857146e-06,
+      "loss": 1.1083,
+      "step": 44888
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.044097142857143e-06,
+      "loss": 0.7693,
+      "step": 44889
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0438285714285714e-06,
+      "loss": 1.2005,
+      "step": 44890
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.04356e-06,
+      "loss": 1.0924,
+      "step": 44891
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0432914285714286e-06,
+      "loss": 0.8936,
+      "step": 44892
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.043022857142857e-06,
+      "loss": 0.7685,
+      "step": 44893
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0427542857142858e-06,
+      "loss": 1.2865,
+      "step": 44894
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0424857142857144e-06,
+      "loss": 1.5147,
+      "step": 44895
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.042217142857143e-06,
+      "loss": 1.4028,
+      "step": 44896
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0419485714285716e-06,
+      "loss": 0.841,
+      "step": 44897
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.04168e-06,
+      "loss": 1.0759,
+      "step": 44898
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0414114285714283e-06,
+      "loss": 1.311,
+      "step": 44899
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0411428571428574e-06,
+      "loss": 0.9972,
+      "step": 44900
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.040874285714286e-06,
+      "loss": 1.2554,
+      "step": 44901
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0406057142857146e-06,
+      "loss": 1.0012,
+      "step": 44902
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0403371428571427e-06,
+      "loss": 1.0861,
+      "step": 44903
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0400685714285713e-06,
+      "loss": 1.0419,
+      "step": 44904
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0398e-06,
+      "loss": 0.8249,
+      "step": 44905
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0395314285714285e-06,
+      "loss": 0.6623,
+      "step": 44906
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0392628571428576e-06,
+      "loss": 1.1978,
+      "step": 44907
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0389942857142857e-06,
+      "loss": 1.019,
+      "step": 44908
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0387257142857143e-06,
+      "loss": 1.2231,
+      "step": 44909
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.038457142857143e-06,
+      "loss": 0.9153,
+      "step": 44910
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0381885714285715e-06,
+      "loss": 1.0802,
+      "step": 44911
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.03792e-06,
+      "loss": 1.0582,
+      "step": 44912
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0376514285714287e-06,
+      "loss": 0.9469,
+      "step": 44913
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0373828571428573e-06,
+      "loss": 0.9117,
+      "step": 44914
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.037114285714286e-06,
+      "loss": 1.1361,
+      "step": 44915
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0368457142857145e-06,
+      "loss": 1.1442,
+      "step": 44916
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0365771428571427e-06,
+      "loss": 0.9612,
+      "step": 44917
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0363085714285713e-06,
+      "loss": 0.8144,
+      "step": 44918
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.03604e-06,
+      "loss": 1.2755,
+      "step": 44919
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.035771428571429e-06,
+      "loss": 1.0499,
+      "step": 44920
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0355028571428575e-06,
+      "loss": 1.36,
+      "step": 44921
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0352342857142857e-06,
+      "loss": 1.4536,
+      "step": 44922
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0349657142857143e-06,
+      "loss": 0.8238,
+      "step": 44923
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.034697142857143e-06,
+      "loss": 1.0861,
+      "step": 44924
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0344285714285715e-06,
+      "loss": 1.1134,
+      "step": 44925
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.03416e-06,
+      "loss": 0.9887,
+      "step": 44926
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0338914285714287e-06,
+      "loss": 0.8665,
+      "step": 44927
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0336228571428573e-06,
+      "loss": 1.4824,
+      "step": 44928
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.033354285714286e-06,
+      "loss": 0.8367,
+      "step": 44929
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0330857142857145e-06,
+      "loss": 1.1732,
+      "step": 44930
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0328171428571427e-06,
+      "loss": 1.5677,
+      "step": 44931
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0325485714285713e-06,
+      "loss": 1.0737,
+      "step": 44932
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0322800000000003e-06,
+      "loss": 0.8714,
+      "step": 44933
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.032011428571429e-06,
+      "loss": 1.0715,
+      "step": 44934
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.031742857142857e-06,
+      "loss": 1.1845,
+      "step": 44935
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0314742857142857e-06,
+      "loss": 0.8682,
+      "step": 44936
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0312057142857143e-06,
+      "loss": 1.0606,
+      "step": 44937
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.030937142857143e-06,
+      "loss": 1.0258,
+      "step": 44938
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0306685714285715e-06,
+      "loss": 1.6218,
+      "step": 44939
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0304e-06,
+      "loss": 1.2232,
+      "step": 44940
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0301314285714287e-06,
+      "loss": 1.1299,
+      "step": 44941
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0298628571428573e-06,
+      "loss": 1.2421,
+      "step": 44942
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.029594285714286e-06,
+      "loss": 1.0987,
+      "step": 44943
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0293257142857145e-06,
+      "loss": 1.2583,
+      "step": 44944
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0290571428571427e-06,
+      "loss": 1.2414,
+      "step": 44945
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0287885714285717e-06,
+      "loss": 1.2123,
+      "step": 44946
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0285200000000003e-06,
+      "loss": 0.9429,
+      "step": 44947
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.028251428571429e-06,
+      "loss": 0.8074,
+      "step": 44948
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.027982857142857e-06,
+      "loss": 1.496,
+      "step": 44949
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0277142857142856e-06,
+      "loss": 1.1891,
+      "step": 44950
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0274457142857142e-06,
+      "loss": 0.94,
+      "step": 44951
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.027177142857143e-06,
+      "loss": 1.3791,
+      "step": 44952
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.026908571428572e-06,
+      "loss": 0.7882,
+      "step": 44953
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.02664e-06,
+      "loss": 0.968,
+      "step": 44954
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0263714285714286e-06,
+      "loss": 0.8505,
+      "step": 44955
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0261028571428572e-06,
+      "loss": 1.2085,
+      "step": 44956
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.025834285714286e-06,
+      "loss": 1.0647,
+      "step": 44957
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.025565714285714e-06,
+      "loss": 0.8817,
+      "step": 44958
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.025297142857143e-06,
+      "loss": 1.4923,
+      "step": 44959
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0250285714285716e-06,
+      "loss": 1.4459,
+      "step": 44960
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0247600000000002e-06,
+      "loss": 1.3489,
+      "step": 44961
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.024491428571429e-06,
+      "loss": 1.3262,
+      "step": 44962
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.024222857142857e-06,
+      "loss": 1.1456,
+      "step": 44963
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0239542857142856e-06,
+      "loss": 1.3204,
+      "step": 44964
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0236857142857142e-06,
+      "loss": 0.9756,
+      "step": 44965
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0234171428571432e-06,
+      "loss": 0.8733,
+      "step": 44966
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0231485714285714e-06,
+      "loss": 0.87,
+      "step": 44967
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.02288e-06,
+      "loss": 1.0827,
+      "step": 44968
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0226114285714286e-06,
+      "loss": 1.0268,
+      "step": 44969
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0223428571428572e-06,
+      "loss": 1.162,
+      "step": 44970
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.022074285714286e-06,
+      "loss": 0.8846,
+      "step": 44971
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.021805714285714e-06,
+      "loss": 0.7766,
+      "step": 44972
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.021537142857143e-06,
+      "loss": 0.8926,
+      "step": 44973
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0212685714285716e-06,
+      "loss": 1.1414,
+      "step": 44974
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0210000000000002e-06,
+      "loss": 0.9422,
+      "step": 44975
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.020731428571429e-06,
+      "loss": 1.2957,
+      "step": 44976
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.020462857142857e-06,
+      "loss": 1.0287,
+      "step": 44977
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0201942857142856e-06,
+      "loss": 1.2998,
+      "step": 44978
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0199257142857146e-06,
+      "loss": 0.9088,
+      "step": 44979
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0196571428571432e-06,
+      "loss": 1.3295,
+      "step": 44980
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0193885714285714e-06,
+      "loss": 0.7349,
+      "step": 44981
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.01912e-06,
+      "loss": 0.8652,
+      "step": 44982
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0188514285714286e-06,
+      "loss": 1.604,
+      "step": 44983
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.018582857142857e-06,
+      "loss": 1.1929,
+      "step": 44984
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.018314285714286e-06,
+      "loss": 1.3076,
+      "step": 44985
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0180457142857144e-06,
+      "loss": 0.8748,
+      "step": 44986
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.017777142857143e-06,
+      "loss": 1.1114,
+      "step": 44987
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0175085714285716e-06,
+      "loss": 1.2508,
+      "step": 44988
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.01724e-06,
+      "loss": 0.7904,
+      "step": 44989
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0169714285714284e-06,
+      "loss": 1.2855,
+      "step": 44990
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.016702857142857e-06,
+      "loss": 1.0429,
+      "step": 44991
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.016434285714286e-06,
+      "loss": 0.7315,
+      "step": 44992
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0161657142857146e-06,
+      "loss": 1.0525,
+      "step": 44993
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.015897142857143e-06,
+      "loss": 1.1728,
+      "step": 44994
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0156285714285714e-06,
+      "loss": 1.1264,
+      "step": 44995
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.01536e-06,
+      "loss": 1.0548,
+      "step": 44996
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0150914285714286e-06,
+      "loss": 1.0279,
+      "step": 44997
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.014822857142857e-06,
+      "loss": 0.4555,
+      "step": 44998
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.014554285714286e-06,
+      "loss": 0.9663,
+      "step": 44999
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0142857142857144e-06,
+      "loss": 0.9113,
+      "step": 45000
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.014017142857143e-06,
+      "loss": 1.4932,
+      "step": 45001
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0137485714285716e-06,
+      "loss": 1.6443,
+      "step": 45002
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.01348e-06,
+      "loss": 1.4565,
+      "step": 45003
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0132114285714283e-06,
+      "loss": 1.2885,
+      "step": 45004
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.012942857142857e-06,
+      "loss": 0.7989,
+      "step": 45005
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.012674285714286e-06,
+      "loss": 1.0266,
+      "step": 45006
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0124057142857146e-06,
+      "loss": 1.0616,
+      "step": 45007
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.012137142857143e-06,
+      "loss": 1.2218,
+      "step": 45008
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0118685714285713e-06,
+      "loss": 0.8944,
+      "step": 45009
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0116e-06,
+      "loss": 1.4621,
+      "step": 45010
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0113314285714285e-06,
+      "loss": 1.0989,
+      "step": 45011
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0110628571428576e-06,
+      "loss": 1.296,
+      "step": 45012
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0107942857142857e-06,
+      "loss": 1.2994,
+      "step": 45013
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0105257142857143e-06,
+      "loss": 1.3328,
+      "step": 45014
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.010257142857143e-06,
+      "loss": 0.9685,
+      "step": 45015
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0099885714285715e-06,
+      "loss": 1.2113,
+      "step": 45016
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.00972e-06,
+      "loss": 0.7816,
+      "step": 45017
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0094514285714283e-06,
+      "loss": 0.8483,
+      "step": 45018
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0091828571428573e-06,
+      "loss": 0.7805,
+      "step": 45019
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.008914285714286e-06,
+      "loss": 1.3839,
+      "step": 45020
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0086457142857145e-06,
+      "loss": 0.8933,
+      "step": 45021
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0083771428571427e-06,
+      "loss": 1.0664,
+      "step": 45022
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0081085714285713e-06,
+      "loss": 1.0445,
+      "step": 45023
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.00784e-06,
+      "loss": 1.0221,
+      "step": 45024
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0075714285714285e-06,
+      "loss": 1.2168,
+      "step": 45025
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0073028571428575e-06,
+      "loss": 0.9563,
+      "step": 45026
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0070342857142857e-06,
+      "loss": 0.8379,
+      "step": 45027
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0067657142857143e-06,
+      "loss": 0.9986,
+      "step": 45028
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.006497142857143e-06,
+      "loss": 0.9651,
+      "step": 45029
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0062285714285715e-06,
+      "loss": 1.0573,
+      "step": 45030
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.00596e-06,
+      "loss": 1.0656,
+      "step": 45031
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0056914285714287e-06,
+      "loss": 0.9073,
+      "step": 45032
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0054228571428573e-06,
+      "loss": 0.835,
+      "step": 45033
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.005154285714286e-06,
+      "loss": 0.9088,
+      "step": 45034
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0048857142857145e-06,
+      "loss": 0.9325,
+      "step": 45035
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0046171428571427e-06,
+      "loss": 1.352,
+      "step": 45036
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0043485714285713e-06,
+      "loss": 1.0106,
+      "step": 45037
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.00408e-06,
+      "loss": 1.0328,
+      "step": 45038
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.003811428571429e-06,
+      "loss": 1.1383,
+      "step": 45039
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0035428571428575e-06,
+      "loss": 0.8632,
+      "step": 45040
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0032742857142857e-06,
+      "loss": 1.3647,
+      "step": 45041
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0030057142857143e-06,
+      "loss": 1.4377,
+      "step": 45042
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.002737142857143e-06,
+      "loss": 0.8765,
+      "step": 45043
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0024685714285715e-06,
+      "loss": 0.9327,
+      "step": 45044
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0022e-06,
+      "loss": 1.4285,
+      "step": 45045
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0019314285714287e-06,
+      "loss": 1.4441,
+      "step": 45046
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0016628571428573e-06,
+      "loss": 1.0798,
+      "step": 45047
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.001394285714286e-06,
+      "loss": 1.168,
+      "step": 45048
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0011257142857145e-06,
+      "loss": 0.9719,
+      "step": 45049
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0008571428571426e-06,
+      "loss": 1.201,
+      "step": 45050
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0005885714285712e-06,
+      "loss": 0.9771,
+      "step": 45051
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.0003200000000003e-06,
+      "loss": 1.1618,
+      "step": 45052
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.000051428571429e-06,
+      "loss": 1.156,
+      "step": 45053
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.999782857142857e-06,
+      "loss": 0.9717,
+      "step": 45054
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.9995142857142856e-06,
+      "loss": 1.1557,
+      "step": 45055
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.9992457142857142e-06,
+      "loss": 0.8434,
+      "step": 45056
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.998977142857143e-06,
+      "loss": 1.2966,
+      "step": 45057
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.9987085714285714e-06,
+      "loss": 1.1731,
+      "step": 45058
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.99844e-06,
+      "loss": 1.2711,
+      "step": 45059
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.9981714285714286e-06,
+      "loss": 1.1658,
+      "step": 45060
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.9979028571428572e-06,
+      "loss": 0.8539,
+      "step": 45061
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.997634285714286e-06,
+      "loss": 1.0589,
+      "step": 45062
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9973657142857144e-06,
+      "loss": 0.9923,
+      "step": 45063
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9970971428571426e-06,
+      "loss": 1.1511,
+      "step": 45064
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9968285714285716e-06,
+      "loss": 1.4198,
+      "step": 45065
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9965600000000002e-06,
+      "loss": 1.1771,
+      "step": 45066
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.996291428571429e-06,
+      "loss": 1.4773,
+      "step": 45067
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.996022857142857e-06,
+      "loss": 1.2698,
+      "step": 45068
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9957542857142856e-06,
+      "loss": 0.7706,
+      "step": 45069
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9954857142857142e-06,
+      "loss": 1.1863,
+      "step": 45070
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.995217142857143e-06,
+      "loss": 1.5098,
+      "step": 45071
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.994948571428572e-06,
+      "loss": 1.3243,
+      "step": 45072
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.99468e-06,
+      "loss": 0.7196,
+      "step": 45073
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9944114285714286e-06,
+      "loss": 1.0548,
+      "step": 45074
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9941428571428572e-06,
+      "loss": 1.1354,
+      "step": 45075
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.993874285714286e-06,
+      "loss": 1.1141,
+      "step": 45076
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.993605714285714e-06,
+      "loss": 1.2951,
+      "step": 45077
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.993337142857143e-06,
+      "loss": 1.2026,
+      "step": 45078
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9930685714285716e-06,
+      "loss": 0.9984,
+      "step": 45079
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9928000000000002e-06,
+      "loss": 1.3435,
+      "step": 45080
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.992531428571429e-06,
+      "loss": 1.3057,
+      "step": 45081
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.992262857142857e-06,
+      "loss": 1.4487,
+      "step": 45082
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9919942857142856e-06,
+      "loss": 0.5555,
+      "step": 45083
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.991725714285714e-06,
+      "loss": 1.3342,
+      "step": 45084
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9914571428571432e-06,
+      "loss": 1.2353,
+      "step": 45085
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.991188571428572e-06,
+      "loss": 1.0466,
+      "step": 45086
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.99092e-06,
+      "loss": 1.0985,
+      "step": 45087
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9906514285714286e-06,
+      "loss": 0.932,
+      "step": 45088
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.990382857142857e-06,
+      "loss": 1.3792,
+      "step": 45089
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.990114285714286e-06,
+      "loss": 0.7286,
+      "step": 45090
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9898457142857144e-06,
+      "loss": 1.1633,
+      "step": 45091
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.989577142857143e-06,
+      "loss": 1.0735,
+      "step": 45092
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9893085714285716e-06,
+      "loss": 1.6782,
+      "step": 45093
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.98904e-06,
+      "loss": 1.0976,
+      "step": 45094
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9887714285714288e-06,
+      "loss": 1.4573,
+      "step": 45095
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.988502857142857e-06,
+      "loss": 1.0604,
+      "step": 45096
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9882342857142856e-06,
+      "loss": 1.0612,
+      "step": 45097
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9879657142857146e-06,
+      "loss": 0.8644,
+      "step": 45098
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.987697142857143e-06,
+      "loss": 0.9814,
+      "step": 45099
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9874285714285714e-06,
+      "loss": 1.2337,
+      "step": 45100
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.98716e-06,
+      "loss": 1.2976,
+      "step": 45101
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9868914285714286e-06,
+      "loss": 1.085,
+      "step": 45102
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.986622857142857e-06,
+      "loss": 1.1886,
+      "step": 45103
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9863542857142858e-06,
+      "loss": 1.2281,
+      "step": 45104
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9860857142857144e-06,
+      "loss": 1.1896,
+      "step": 45105
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.985817142857143e-06,
+      "loss": 0.8865,
+      "step": 45106
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9855485714285716e-06,
+      "loss": 1.0659,
+      "step": 45107
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.98528e-06,
+      "loss": 0.8634,
+      "step": 45108
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9850114285714283e-06,
+      "loss": 1.2604,
+      "step": 45109
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.984742857142857e-06,
+      "loss": 0.8176,
+      "step": 45110
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.984474285714286e-06,
+      "loss": 1.1121,
+      "step": 45111
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9842057142857146e-06,
+      "loss": 1.2662,
+      "step": 45112
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.983937142857143e-06,
+      "loss": 0.94,
+      "step": 45113
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9836685714285713e-06,
+      "loss": 0.7851,
+      "step": 45114
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9834e-06,
+      "loss": 1.2885,
+      "step": 45115
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9831314285714285e-06,
+      "loss": 1.0126,
+      "step": 45116
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.982862857142857e-06,
+      "loss": 0.9314,
+      "step": 45117
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.982594285714286e-06,
+      "loss": 1.1754,
+      "step": 45118
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9823257142857143e-06,
+      "loss": 0.9816,
+      "step": 45119
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.982057142857143e-06,
+      "loss": 1.2964,
+      "step": 45120
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9817885714285715e-06,
+      "loss": 1.2122,
+      "step": 45121
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.98152e-06,
+      "loss": 1.3107,
+      "step": 45122
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9812514285714283e-06,
+      "loss": 1.165,
+      "step": 45123
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9809828571428573e-06,
+      "loss": 0.7151,
+      "step": 45124
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.980714285714286e-06,
+      "loss": 1.4384,
+      "step": 45125
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9804457142857145e-06,
+      "loss": 0.7513,
+      "step": 45126
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.980177142857143e-06,
+      "loss": 1.2342,
+      "step": 45127
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9799085714285713e-06,
+      "loss": 1.011,
+      "step": 45128
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.97964e-06,
+      "loss": 1.1294,
+      "step": 45129
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9793714285714285e-06,
+      "loss": 0.9401,
+      "step": 45130
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9791028571428575e-06,
+      "loss": 0.8229,
+      "step": 45131
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9788342857142857e-06,
+      "loss": 1.1793,
+      "step": 45132
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9785657142857143e-06,
+      "loss": 1.395,
+      "step": 45133
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.978297142857143e-06,
+      "loss": 1.3474,
+      "step": 45134
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9780285714285715e-06,
+      "loss": 1.1194,
+      "step": 45135
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.97776e-06,
+      "loss": 0.9283,
+      "step": 45136
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9774914285714287e-06,
+      "loss": 1.1809,
+      "step": 45137
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9772228571428573e-06,
+      "loss": 0.9616,
+      "step": 45138
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.976954285714286e-06,
+      "loss": 0.9932,
+      "step": 45139
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9766857142857145e-06,
+      "loss": 1.4817,
+      "step": 45140
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9764171428571427e-06,
+      "loss": 1.0839,
+      "step": 45141
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9761485714285713e-06,
+      "loss": 1.1313,
+      "step": 45142
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.97588e-06,
+      "loss": 1.1683,
+      "step": 45143
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.975611428571429e-06,
+      "loss": 1.3093,
+      "step": 45144
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9753428571428575e-06,
+      "loss": 1.0792,
+      "step": 45145
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9750742857142857e-06,
+      "loss": 0.8983,
+      "step": 45146
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9748057142857143e-06,
+      "loss": 1.1069,
+      "step": 45147
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.974537142857143e-06,
+      "loss": 0.8904,
+      "step": 45148
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9742685714285715e-06,
+      "loss": 0.8732,
+      "step": 45149
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.974e-06,
+      "loss": 1.4319,
+      "step": 45150
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9737314285714287e-06,
+      "loss": 0.9171,
+      "step": 45151
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9734628571428573e-06,
+      "loss": 1.2363,
+      "step": 45152
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.973194285714286e-06,
+      "loss": 1.062,
+      "step": 45153
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9729257142857145e-06,
+      "loss": 1.1996,
+      "step": 45154
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9726571428571426e-06,
+      "loss": 1.1128,
+      "step": 45155
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9723885714285712e-06,
+      "loss": 1.1442,
+      "step": 45156
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9721200000000003e-06,
+      "loss": 0.9876,
+      "step": 45157
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.971851428571429e-06,
+      "loss": 1.0441,
+      "step": 45158
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9715828571428575e-06,
+      "loss": 1.0304,
+      "step": 45159
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9713142857142856e-06,
+      "loss": 1.0846,
+      "step": 45160
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9710457142857142e-06,
+      "loss": 1.0989,
+      "step": 45161
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.970777142857143e-06,
+      "loss": 1.0324,
+      "step": 45162
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9705085714285714e-06,
+      "loss": 0.8804,
+      "step": 45163
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.97024e-06,
+      "loss": 1.3185,
+      "step": 45164
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9699714285714286e-06,
+      "loss": 0.9034,
+      "step": 45165
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9697028571428572e-06,
+      "loss": 1.2125,
+      "step": 45166
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.969434285714286e-06,
+      "loss": 1.1124,
+      "step": 45167
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9691657142857144e-06,
+      "loss": 0.9266,
+      "step": 45168
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9688971428571426e-06,
+      "loss": 1.0854,
+      "step": 45169
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9686285714285716e-06,
+      "loss": 0.9307,
+      "step": 45170
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9683600000000002e-06,
+      "loss": 1.6508,
+      "step": 45171
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.968091428571429e-06,
+      "loss": 1.1479,
+      "step": 45172
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.967822857142857e-06,
+      "loss": 1.099,
+      "step": 45173
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9675542857142856e-06,
+      "loss": 0.9601,
+      "step": 45174
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9672857142857142e-06,
+      "loss": 1.2263,
+      "step": 45175
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.967017142857143e-06,
+      "loss": 1.1032,
+      "step": 45176
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.966748571428572e-06,
+      "loss": 0.8198,
+      "step": 45177
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.96648e-06,
+      "loss": 1.3568,
+      "step": 45178
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9662114285714286e-06,
+      "loss": 1.0103,
+      "step": 45179
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9659428571428572e-06,
+      "loss": 1.0544,
+      "step": 45180
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.965674285714286e-06,
+      "loss": 1.0631,
+      "step": 45181
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9654057142857144e-06,
+      "loss": 0.7438,
+      "step": 45182
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.965137142857143e-06,
+      "loss": 0.8928,
+      "step": 45183
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9648685714285716e-06,
+      "loss": 1.1525,
+      "step": 45184
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9646000000000002e-06,
+      "loss": 1.2192,
+      "step": 45185
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.964331428571429e-06,
+      "loss": 0.9935,
+      "step": 45186
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.964062857142857e-06,
+      "loss": 1.1495,
+      "step": 45187
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9637942857142856e-06,
+      "loss": 1.0989,
+      "step": 45188
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.963525714285714e-06,
+      "loss": 1.1182,
+      "step": 45189
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9632571428571432e-06,
+      "loss": 0.9286,
+      "step": 45190
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.962988571428572e-06,
+      "loss": 1.1248,
+      "step": 45191
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.96272e-06,
+      "loss": 1.2133,
+      "step": 45192
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9624514285714286e-06,
+      "loss": 1.4052,
+      "step": 45193
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.962182857142857e-06,
+      "loss": 1.2992,
+      "step": 45194
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9619142857142858e-06,
+      "loss": 0.9857,
+      "step": 45195
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.961645714285714e-06,
+      "loss": 1.1694,
+      "step": 45196
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.961377142857143e-06,
+      "loss": 1.3858,
+      "step": 45197
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9611085714285716e-06,
+      "loss": 1.3242,
+      "step": 45198
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.96084e-06,
+      "loss": 0.8697,
+      "step": 45199
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9605714285714288e-06,
+      "loss": 0.7231,
+      "step": 45200
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.960302857142857e-06,
+      "loss": 0.7606,
+      "step": 45201
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9600342857142856e-06,
+      "loss": 1.0335,
+      "step": 45202
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9597657142857146e-06,
+      "loss": 1.6178,
+      "step": 45203
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.959497142857143e-06,
+      "loss": 1.1795,
+      "step": 45204
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9592285714285718e-06,
+      "loss": 1.4425,
+      "step": 45205
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.95896e-06,
+      "loss": 1.2791,
+      "step": 45206
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9586914285714286e-06,
+      "loss": 0.9952,
+      "step": 45207
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.958422857142857e-06,
+      "loss": 1.3457,
+      "step": 45208
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9581542857142858e-06,
+      "loss": 1.503,
+      "step": 45209
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9578857142857144e-06,
+      "loss": 1.2813,
+      "step": 45210
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.957617142857143e-06,
+      "loss": 1.331,
+      "step": 45211
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9573485714285716e-06,
+      "loss": 1.0678,
+      "step": 45212
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.95708e-06,
+      "loss": 0.8657,
+      "step": 45213
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9568114285714288e-06,
+      "loss": 1.1388,
+      "step": 45214
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.956542857142857e-06,
+      "loss": 0.8933,
+      "step": 45215
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.956274285714286e-06,
+      "loss": 0.9925,
+      "step": 45216
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9560057142857146e-06,
+      "loss": 1.4403,
+      "step": 45217
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.955737142857143e-06,
+      "loss": 1.1257,
+      "step": 45218
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9554685714285713e-06,
+      "loss": 0.8785,
+      "step": 45219
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9552e-06,
+      "loss": 1.0925,
+      "step": 45220
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9549314285714285e-06,
+      "loss": 1.0558,
+      "step": 45221
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.954662857142857e-06,
+      "loss": 1.0891,
+      "step": 45222
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.954394285714286e-06,
+      "loss": 0.9607,
+      "step": 45223
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9541257142857143e-06,
+      "loss": 1.1502,
+      "step": 45224
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.953857142857143e-06,
+      "loss": 1.1862,
+      "step": 45225
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9535885714285715e-06,
+      "loss": 0.8588,
+      "step": 45226
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.95332e-06,
+      "loss": 1.1993,
+      "step": 45227
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9530514285714283e-06,
+      "loss": 1.3571,
+      "step": 45228
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.952782857142857e-06,
+      "loss": 0.8926,
+      "step": 45229
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.952514285714286e-06,
+      "loss": 1.163,
+      "step": 45230
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9522457142857145e-06,
+      "loss": 0.8158,
+      "step": 45231
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.951977142857143e-06,
+      "loss": 0.812,
+      "step": 45232
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9517085714285713e-06,
+      "loss": 1.2829,
+      "step": 45233
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.95144e-06,
+      "loss": 1.518,
+      "step": 45234
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9511714285714285e-06,
+      "loss": 0.9484,
+      "step": 45235
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.9509028571428575e-06,
+      "loss": 1.1303,
+      "step": 45236
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.950634285714286e-06,
+      "loss": 1.1843,
+      "step": 45237
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9503657142857143e-06,
+      "loss": 0.8909,
+      "step": 45238
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.950097142857143e-06,
+      "loss": 0.9666,
+      "step": 45239
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9498285714285715e-06,
+      "loss": 0.8458,
+      "step": 45240
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.94956e-06,
+      "loss": 1.0695,
+      "step": 45241
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9492914285714283e-06,
+      "loss": 0.8254,
+      "step": 45242
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9490228571428573e-06,
+      "loss": 1.2303,
+      "step": 45243
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.948754285714286e-06,
+      "loss": 1.2646,
+      "step": 45244
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9484857142857145e-06,
+      "loss": 0.8647,
+      "step": 45245
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.948217142857143e-06,
+      "loss": 1.0513,
+      "step": 45246
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9479485714285713e-06,
+      "loss": 1.2128,
+      "step": 45247
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.94768e-06,
+      "loss": 1.0138,
+      "step": 45248
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9474114285714285e-06,
+      "loss": 1.1998,
+      "step": 45249
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9471428571428575e-06,
+      "loss": 1.1764,
+      "step": 45250
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9468742857142857e-06,
+      "loss": 1.3748,
+      "step": 45251
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9466057142857143e-06,
+      "loss": 1.2404,
+      "step": 45252
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.946337142857143e-06,
+      "loss": 1.1403,
+      "step": 45253
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9460685714285715e-06,
+      "loss": 1.2559,
+      "step": 45254
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9458e-06,
+      "loss": 0.8752,
+      "step": 45255
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9455314285714287e-06,
+      "loss": 1.0786,
+      "step": 45256
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9452628571428573e-06,
+      "loss": 1.2234,
+      "step": 45257
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.944994285714286e-06,
+      "loss": 1.5848,
+      "step": 45258
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9447257142857145e-06,
+      "loss": 0.7212,
+      "step": 45259
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9444571428571426e-06,
+      "loss": 0.9575,
+      "step": 45260
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9441885714285712e-06,
+      "loss": 1.218,
+      "step": 45261
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.94392e-06,
+      "loss": 0.7307,
+      "step": 45262
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.943651428571429e-06,
+      "loss": 0.6578,
+      "step": 45263
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9433828571428575e-06,
+      "loss": 1.1385,
+      "step": 45264
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9431142857142856e-06,
+      "loss": 0.8887,
+      "step": 45265
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9428457142857142e-06,
+      "loss": 0.9975,
+      "step": 45266
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.942577142857143e-06,
+      "loss": 0.9071,
+      "step": 45267
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9423085714285714e-06,
+      "loss": 1.4045,
+      "step": 45268
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9420400000000005e-06,
+      "loss": 1.0807,
+      "step": 45269
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9417714285714286e-06,
+      "loss": 1.5094,
+      "step": 45270
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9415028571428572e-06,
+      "loss": 1.2448,
+      "step": 45271
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.941234285714286e-06,
+      "loss": 1.361,
+      "step": 45272
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9409657142857144e-06,
+      "loss": 0.8804,
+      "step": 45273
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9406971428571426e-06,
+      "loss": 0.9127,
+      "step": 45274
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9404285714285712e-06,
+      "loss": 1.2984,
+      "step": 45275
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9401600000000002e-06,
+      "loss": 1.03,
+      "step": 45276
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.939891428571429e-06,
+      "loss": 1.5785,
+      "step": 45277
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9396228571428574e-06,
+      "loss": 1.4827,
+      "step": 45278
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9393542857142856e-06,
+      "loss": 0.8778,
+      "step": 45279
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9390857142857142e-06,
+      "loss": 1.1085,
+      "step": 45280
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.938817142857143e-06,
+      "loss": 1.207,
+      "step": 45281
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9385485714285714e-06,
+      "loss": 1.3838,
+      "step": 45282
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.93828e-06,
+      "loss": 0.8843,
+      "step": 45283
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9380114285714286e-06,
+      "loss": 1.4613,
+      "step": 45284
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9377428571428572e-06,
+      "loss": 1.2611,
+      "step": 45285
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.937474285714286e-06,
+      "loss": 1.0049,
+      "step": 45286
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9372057142857144e-06,
+      "loss": 0.8509,
+      "step": 45287
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9369371428571426e-06,
+      "loss": 1.1755,
+      "step": 45288
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9366685714285716e-06,
+      "loss": 1.1332,
+      "step": 45289
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9364000000000002e-06,
+      "loss": 1.5251,
+      "step": 45290
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.936131428571429e-06,
+      "loss": 1.0413,
+      "step": 45291
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9358628571428574e-06,
+      "loss": 1.3798,
+      "step": 45292
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9355942857142856e-06,
+      "loss": 0.8167,
+      "step": 45293
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.935325714285714e-06,
+      "loss": 1.2169,
+      "step": 45294
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.935057142857143e-06,
+      "loss": 1.0634,
+      "step": 45295
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.934788571428572e-06,
+      "loss": 1.0463,
+      "step": 45296
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.93452e-06,
+      "loss": 0.9021,
+      "step": 45297
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9342514285714286e-06,
+      "loss": 1.2802,
+      "step": 45298
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.933982857142857e-06,
+      "loss": 0.9191,
+      "step": 45299
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9337142857142858e-06,
+      "loss": 1.0542,
+      "step": 45300
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9334457142857144e-06,
+      "loss": 1.0183,
+      "step": 45301
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.933177142857143e-06,
+      "loss": 0.6788,
+      "step": 45302
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9329085714285716e-06,
+      "loss": 1.0666,
+      "step": 45303
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.93264e-06,
+      "loss": 1.0649,
+      "step": 45304
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9323714285714288e-06,
+      "loss": 1.2475,
+      "step": 45305
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.932102857142857e-06,
+      "loss": 0.9403,
+      "step": 45306
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9318342857142856e-06,
+      "loss": 1.3643,
+      "step": 45307
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.931565714285714e-06,
+      "loss": 1.2857,
+      "step": 45308
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.931297142857143e-06,
+      "loss": 1.1378,
+      "step": 45309
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9310285714285718e-06,
+      "loss": 1.3093,
+      "step": 45310
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.93076e-06,
+      "loss": 1.32,
+      "step": 45311
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9304914285714286e-06,
+      "loss": 1.2563,
+      "step": 45312
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.930222857142857e-06,
+      "loss": 0.7877,
+      "step": 45313
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9299542857142858e-06,
+      "loss": 0.9348,
+      "step": 45314
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9296857142857144e-06,
+      "loss": 1.4042,
+      "step": 45315
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.929417142857143e-06,
+      "loss": 1.1203,
+      "step": 45316
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9291485714285716e-06,
+      "loss": 1.3821,
+      "step": 45317
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.92888e-06,
+      "loss": 0.9562,
+      "step": 45318
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9286114285714288e-06,
+      "loss": 1.1242,
+      "step": 45319
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.928342857142857e-06,
+      "loss": 1.5953,
+      "step": 45320
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9280742857142855e-06,
+      "loss": 1.1485,
+      "step": 45321
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9278057142857146e-06,
+      "loss": 1.2306,
+      "step": 45322
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.927537142857143e-06,
+      "loss": 1.1799,
+      "step": 45323
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9272685714285718e-06,
+      "loss": 1.152,
+      "step": 45324
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.927e-06,
+      "loss": 1.5791,
+      "step": 45325
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9267314285714285e-06,
+      "loss": 0.9425,
+      "step": 45326
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.926462857142857e-06,
+      "loss": 1.0968,
+      "step": 45327
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9261942857142857e-06,
+      "loss": 1.5343,
+      "step": 45328
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9259257142857143e-06,
+      "loss": 0.9435,
+      "step": 45329
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.925657142857143e-06,
+      "loss": 1.2263,
+      "step": 45330
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9253885714285715e-06,
+      "loss": 0.8197,
+      "step": 45331
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.92512e-06,
+      "loss": 1.2694,
+      "step": 45332
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9248514285714287e-06,
+      "loss": 1.1803,
+      "step": 45333
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.924582857142857e-06,
+      "loss": 1.1819,
+      "step": 45334
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.924314285714286e-06,
+      "loss": 0.8625,
+      "step": 45335
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9240457142857145e-06,
+      "loss": 1.3349,
+      "step": 45336
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.923777142857143e-06,
+      "loss": 1.402,
+      "step": 45337
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9235085714285713e-06,
+      "loss": 1.3682,
+      "step": 45338
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.92324e-06,
+      "loss": 1.3401,
+      "step": 45339
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9229714285714285e-06,
+      "loss": 0.8493,
+      "step": 45340
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.922702857142857e-06,
+      "loss": 0.7203,
+      "step": 45341
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.922434285714286e-06,
+      "loss": 1.4737,
+      "step": 45342
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9221657142857143e-06,
+      "loss": 1.0941,
+      "step": 45343
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.921897142857143e-06,
+      "loss": 1.4264,
+      "step": 45344
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9216285714285715e-06,
+      "loss": 1.261,
+      "step": 45345
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.92136e-06,
+      "loss": 0.9676,
+      "step": 45346
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9210914285714283e-06,
+      "loss": 0.9144,
+      "step": 45347
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9208228571428573e-06,
+      "loss": 0.9233,
+      "step": 45348
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.920554285714286e-06,
+      "loss": 0.9238,
+      "step": 45349
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9202857142857145e-06,
+      "loss": 1.0389,
+      "step": 45350
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.920017142857143e-06,
+      "loss": 1.4892,
+      "step": 45351
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9197485714285713e-06,
+      "loss": 0.703,
+      "step": 45352
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.91948e-06,
+      "loss": 1.0042,
+      "step": 45353
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9192114285714285e-06,
+      "loss": 1.3878,
+      "step": 45354
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9189428571428575e-06,
+      "loss": 1.0884,
+      "step": 45355
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.918674285714286e-06,
+      "loss": 1.2511,
+      "step": 45356
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9184057142857143e-06,
+      "loss": 0.8884,
+      "step": 45357
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.918137142857143e-06,
+      "loss": 1.2085,
+      "step": 45358
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9178685714285715e-06,
+      "loss": 0.9517,
+      "step": 45359
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9176e-06,
+      "loss": 0.9616,
+      "step": 45360
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9173314285714287e-06,
+      "loss": 0.9968,
+      "step": 45361
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9170628571428573e-06,
+      "loss": 1.0996,
+      "step": 45362
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.916794285714286e-06,
+      "loss": 1.3591,
+      "step": 45363
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9165257142857145e-06,
+      "loss": 1.162,
+      "step": 45364
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.916257142857143e-06,
+      "loss": 1.2412,
+      "step": 45365
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9159885714285712e-06,
+      "loss": 0.7998,
+      "step": 45366
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.91572e-06,
+      "loss": 0.9314,
+      "step": 45367
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.915451428571429e-06,
+      "loss": 1.5332,
+      "step": 45368
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9151828571428575e-06,
+      "loss": 1.1966,
+      "step": 45369
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9149142857142856e-06,
+      "loss": 1.6343,
+      "step": 45370
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9146457142857142e-06,
+      "loss": 1.2488,
+      "step": 45371
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.914377142857143e-06,
+      "loss": 1.1612,
+      "step": 45372
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9141085714285714e-06,
+      "loss": 1.3124,
+      "step": 45373
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.91384e-06,
+      "loss": 0.8311,
+      "step": 45374
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9135714285714286e-06,
+      "loss": 0.7262,
+      "step": 45375
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9133028571428572e-06,
+      "loss": 1.651,
+      "step": 45376
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.913034285714286e-06,
+      "loss": 0.7888,
+      "step": 45377
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9127657142857144e-06,
+      "loss": 1.1923,
+      "step": 45378
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9124971428571426e-06,
+      "loss": 0.9056,
+      "step": 45379
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9122285714285712e-06,
+      "loss": 0.8091,
+      "step": 45380
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9119600000000002e-06,
+      "loss": 1.1318,
+      "step": 45381
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.911691428571429e-06,
+      "loss": 0.9029,
+      "step": 45382
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9114228571428574e-06,
+      "loss": 0.9872,
+      "step": 45383
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9111542857142856e-06,
+      "loss": 1.2224,
+      "step": 45384
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9108857142857142e-06,
+      "loss": 1.3265,
+      "step": 45385
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.910617142857143e-06,
+      "loss": 1.067,
+      "step": 45386
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9103485714285714e-06,
+      "loss": 0.9414,
+      "step": 45387
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9100800000000004e-06,
+      "loss": 1.1351,
+      "step": 45388
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9098114285714286e-06,
+      "loss": 0.6405,
+      "step": 45389
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9095428571428572e-06,
+      "loss": 1.0253,
+      "step": 45390
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.909274285714286e-06,
+      "loss": 0.9391,
+      "step": 45391
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9090057142857144e-06,
+      "loss": 1.3812,
+      "step": 45392
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9087371428571426e-06,
+      "loss": 1.2187,
+      "step": 45393
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9084685714285716e-06,
+      "loss": 1.0155,
+      "step": 45394
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9082000000000002e-06,
+      "loss": 1.1906,
+      "step": 45395
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.907931428571429e-06,
+      "loss": 1.0397,
+      "step": 45396
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9076628571428574e-06,
+      "loss": 1.3916,
+      "step": 45397
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9073942857142856e-06,
+      "loss": 1.0145,
+      "step": 45398
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9071257142857144e-06,
+      "loss": 1.1629,
+      "step": 45399
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.906857142857143e-06,
+      "loss": 1.0173,
+      "step": 45400
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9065885714285716e-06,
+      "loss": 0.746,
+      "step": 45401
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.90632e-06,
+      "loss": 0.925,
+      "step": 45402
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9060514285714286e-06,
+      "loss": 1.2746,
+      "step": 45403
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9057828571428572e-06,
+      "loss": 1.5391,
+      "step": 45404
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9055142857142858e-06,
+      "loss": 0.9076,
+      "step": 45405
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9052457142857146e-06,
+      "loss": 1.2405,
+      "step": 45406
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9049771428571428e-06,
+      "loss": 0.8281,
+      "step": 45407
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9047085714285714e-06,
+      "loss": 1.138,
+      "step": 45408
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9044400000000002e-06,
+      "loss": 1.3093,
+      "step": 45409
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9041714285714288e-06,
+      "loss": 1.285,
+      "step": 45410
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9039028571428574e-06,
+      "loss": 0.9851,
+      "step": 45411
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9036342857142858e-06,
+      "loss": 1.3876,
+      "step": 45412
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9033657142857144e-06,
+      "loss": 1.0097,
+      "step": 45413
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.903097142857143e-06,
+      "loss": 1.3124,
+      "step": 45414
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9028285714285716e-06,
+      "loss": 0.7257,
+      "step": 45415
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.90256e-06,
+      "loss": 0.9611,
+      "step": 45416
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9022914285714286e-06,
+      "loss": 1.4802,
+      "step": 45417
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9020228571428572e-06,
+      "loss": 1.4016,
+      "step": 45418
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.901754285714286e-06,
+      "loss": 0.7436,
+      "step": 45419
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9014857142857146e-06,
+      "loss": 1.1046,
+      "step": 45420
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9012171428571427e-06,
+      "loss": 1.3486,
+      "step": 45421
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9009485714285716e-06,
+      "loss": 1.0138,
+      "step": 45422
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9006800000000002e-06,
+      "loss": 1.4405,
+      "step": 45423
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9004114285714288e-06,
+      "loss": 1.0344,
+      "step": 45424
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.9001428571428571e-06,
+      "loss": 1.0125,
+      "step": 45425
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8998742857142857e-06,
+      "loss": 0.9317,
+      "step": 45426
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8996057142857143e-06,
+      "loss": 1.1874,
+      "step": 45427
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.899337142857143e-06,
+      "loss": 0.9501,
+      "step": 45428
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8990685714285718e-06,
+      "loss": 0.9242,
+      "step": 45429
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8988e-06,
+      "loss": 1.5064,
+      "step": 45430
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8985314285714285e-06,
+      "loss": 1.3169,
+      "step": 45431
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8982628571428573e-06,
+      "loss": 1.202,
+      "step": 45432
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.897994285714286e-06,
+      "loss": 1.1558,
+      "step": 45433
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8977257142857141e-06,
+      "loss": 0.734,
+      "step": 45434
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.897457142857143e-06,
+      "loss": 1.0794,
+      "step": 45435
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8971885714285715e-06,
+      "loss": 1.3281,
+      "step": 45436
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8969200000000001e-06,
+      "loss": 0.9228,
+      "step": 45437
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8966514285714287e-06,
+      "loss": 0.757,
+      "step": 45438
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8963828571428571e-06,
+      "loss": 1.1617,
+      "step": 45439
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8961142857142857e-06,
+      "loss": 0.8559,
+      "step": 45440
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8958457142857143e-06,
+      "loss": 1.1905,
+      "step": 45441
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8955771428571431e-06,
+      "loss": 1.4802,
+      "step": 45442
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8953085714285717e-06,
+      "loss": 1.2001,
+      "step": 45443
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.89504e-06,
+      "loss": 1.3773,
+      "step": 45444
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8947714285714287e-06,
+      "loss": 1.4164,
+      "step": 45445
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8945028571428573e-06,
+      "loss": 1.239,
+      "step": 45446
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.894234285714286e-06,
+      "loss": 0.9205,
+      "step": 45447
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.893965714285714e-06,
+      "loss": 1.4204,
+      "step": 45448
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.893697142857143e-06,
+      "loss": 1.2337,
+      "step": 45449
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8934285714285715e-06,
+      "loss": 0.6689,
+      "step": 45450
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.89316e-06,
+      "loss": 1.1312,
+      "step": 45451
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.892891428571429e-06,
+      "loss": 1.3055,
+      "step": 45452
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.892622857142857e-06,
+      "loss": 1.2314,
+      "step": 45453
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8923542857142857e-06,
+      "loss": 1.1053,
+      "step": 45454
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8920857142857145e-06,
+      "loss": 1.5968,
+      "step": 45455
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.891817142857143e-06,
+      "loss": 1.156,
+      "step": 45456
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8915485714285713e-06,
+      "loss": 1.0521,
+      "step": 45457
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8912799999999999e-06,
+      "loss": 1.3224,
+      "step": 45458
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8910114285714287e-06,
+      "loss": 1.2339,
+      "step": 45459
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8907428571428573e-06,
+      "loss": 1.3108,
+      "step": 45460
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8904742857142859e-06,
+      "loss": 1.1323,
+      "step": 45461
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8902057142857143e-06,
+      "loss": 1.2565,
+      "step": 45462
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8899371428571429e-06,
+      "loss": 1.4372,
+      "step": 45463
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8896685714285715e-06,
+      "loss": 0.9825,
+      "step": 45464
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8894000000000003e-06,
+      "loss": 1.1348,
+      "step": 45465
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8891314285714285e-06,
+      "loss": 1.2049,
+      "step": 45466
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.888862857142857e-06,
+      "loss": 1.1757,
+      "step": 45467
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8885942857142857e-06,
+      "loss": 1.253,
+      "step": 45468
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8883257142857145e-06,
+      "loss": 1.2551,
+      "step": 45469
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.888057142857143e-06,
+      "loss": 0.6127,
+      "step": 45470
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8877885714285712e-06,
+      "loss": 1.0724,
+      "step": 45471
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.88752e-06,
+      "loss": 1.6588,
+      "step": 45472
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8872514285714287e-06,
+      "loss": 0.958,
+      "step": 45473
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8869828571428573e-06,
+      "loss": 1.1744,
+      "step": 45474
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.886714285714286e-06,
+      "loss": 0.7795,
+      "step": 45475
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8864457142857142e-06,
+      "loss": 0.5582,
+      "step": 45476
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8861771428571428e-06,
+      "loss": 1.1617,
+      "step": 45477
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8859085714285717e-06,
+      "loss": 1.0211,
+      "step": 45478
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8856400000000003e-06,
+      "loss": 0.8911,
+      "step": 45479
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8853714285714284e-06,
+      "loss": 0.9928,
+      "step": 45480
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.885102857142857e-06,
+      "loss": 1.0869,
+      "step": 45481
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8848342857142858e-06,
+      "loss": 1.2327,
+      "step": 45482
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8845657142857144e-06,
+      "loss": 1.4396,
+      "step": 45483
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.884297142857143e-06,
+      "loss": 1.0145,
+      "step": 45484
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8840285714285714e-06,
+      "loss": 1.0502,
+      "step": 45485
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.88376e-06,
+      "loss": 1.5585,
+      "step": 45486
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8834914285714286e-06,
+      "loss": 1.022,
+      "step": 45487
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8832228571428574e-06,
+      "loss": 0.7666,
+      "step": 45488
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8829542857142856e-06,
+      "loss": 0.8095,
+      "step": 45489
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8826857142857142e-06,
+      "loss": 0.9247,
+      "step": 45490
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8824171428571428e-06,
+      "loss": 0.7874,
+      "step": 45491
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8821485714285716e-06,
+      "loss": 0.5309,
+      "step": 45492
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8818800000000002e-06,
+      "loss": 0.8533,
+      "step": 45493
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8816114285714284e-06,
+      "loss": 0.7994,
+      "step": 45494
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8813428571428572e-06,
+      "loss": 0.9379,
+      "step": 45495
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8810742857142858e-06,
+      "loss": 1.4814,
+      "step": 45496
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8808057142857144e-06,
+      "loss": 1.3926,
+      "step": 45497
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8805371428571432e-06,
+      "loss": 1.1733,
+      "step": 45498
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8802685714285714e-06,
+      "loss": 1.1539,
+      "step": 45499
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.88e-06,
+      "loss": 0.6906,
+      "step": 45500
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8797314285714286e-06,
+      "loss": 0.9774,
+      "step": 45501
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8794628571428574e-06,
+      "loss": 1.0885,
+      "step": 45502
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8791942857142856e-06,
+      "loss": 1.0889,
+      "step": 45503
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8789257142857142e-06,
+      "loss": 0.8643,
+      "step": 45504
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.878657142857143e-06,
+      "loss": 1.1927,
+      "step": 45505
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8783885714285716e-06,
+      "loss": 1.4248,
+      "step": 45506
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8781200000000002e-06,
+      "loss": 0.766,
+      "step": 45507
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8778514285714286e-06,
+      "loss": 0.7167,
+      "step": 45508
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8775828571428572e-06,
+      "loss": 0.9526,
+      "step": 45509
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8773142857142858e-06,
+      "loss": 1.1406,
+      "step": 45510
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8770457142857144e-06,
+      "loss": 1.2971,
+      "step": 45511
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8767771428571428e-06,
+      "loss": 1.1699,
+      "step": 45512
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8765085714285714e-06,
+      "loss": 1.1639,
+      "step": 45513
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.87624e-06,
+      "loss": 1.0854,
+      "step": 45514
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8759714285714288e-06,
+      "loss": 1.2303,
+      "step": 45515
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8757028571428574e-06,
+      "loss": 0.9966,
+      "step": 45516
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8754342857142856e-06,
+      "loss": 1.2793,
+      "step": 45517
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8751657142857144e-06,
+      "loss": 1.1533,
+      "step": 45518
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.874897142857143e-06,
+      "loss": 0.6345,
+      "step": 45519
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8746285714285716e-06,
+      "loss": 1.3004,
+      "step": 45520
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.87436e-06,
+      "loss": 0.764,
+      "step": 45521
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8740914285714286e-06,
+      "loss": 0.8651,
+      "step": 45522
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8738228571428572e-06,
+      "loss": 1.2709,
+      "step": 45523
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8735542857142858e-06,
+      "loss": 1.28,
+      "step": 45524
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8732857142857146e-06,
+      "loss": 1.1049,
+      "step": 45525
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8730171428571427e-06,
+      "loss": 1.0836,
+      "step": 45526
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8727485714285713e-06,
+      "loss": 1.1673,
+      "step": 45527
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8724800000000002e-06,
+      "loss": 0.9364,
+      "step": 45528
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8722114285714288e-06,
+      "loss": 0.8578,
+      "step": 45529
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8719428571428574e-06,
+      "loss": 0.922,
+      "step": 45530
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8716742857142857e-06,
+      "loss": 1.2688,
+      "step": 45531
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8714057142857143e-06,
+      "loss": 1.4887,
+      "step": 45532
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.871137142857143e-06,
+      "loss": 0.9168,
+      "step": 45533
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8708685714285715e-06,
+      "loss": 0.6899,
+      "step": 45534
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8706e-06,
+      "loss": 1.1494,
+      "step": 45535
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8703314285714285e-06,
+      "loss": 0.6895,
+      "step": 45536
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8700628571428571e-06,
+      "loss": 1.4804,
+      "step": 45537
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.869794285714286e-06,
+      "loss": 0.8196,
+      "step": 45538
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8695257142857145e-06,
+      "loss": 1.1412,
+      "step": 45539
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8692571428571427e-06,
+      "loss": 1.2373,
+      "step": 45540
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8689885714285715e-06,
+      "loss": 0.9686,
+      "step": 45541
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8687200000000001e-06,
+      "loss": 0.9842,
+      "step": 45542
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8684514285714287e-06,
+      "loss": 0.9893,
+      "step": 45543
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8681828571428571e-06,
+      "loss": 1.2567,
+      "step": 45544
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8679142857142857e-06,
+      "loss": 1.5362,
+      "step": 45545
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8676457142857143e-06,
+      "loss": 1.0259,
+      "step": 45546
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.867377142857143e-06,
+      "loss": 1.2546,
+      "step": 45547
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8671085714285717e-06,
+      "loss": 0.7544,
+      "step": 45548
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.86684e-06,
+      "loss": 0.9213,
+      "step": 45549
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8665714285714285e-06,
+      "loss": 1.215,
+      "step": 45550
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8663028571428573e-06,
+      "loss": 0.7748,
+      "step": 45551
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.866034285714286e-06,
+      "loss": 1.2578,
+      "step": 45552
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.865765714285714e-06,
+      "loss": 1.5266,
+      "step": 45553
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.865497142857143e-06,
+      "loss": 1.4753,
+      "step": 45554
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8652285714285715e-06,
+      "loss": 1.169,
+      "step": 45555
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.86496e-06,
+      "loss": 1.0475,
+      "step": 45556
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8646914285714287e-06,
+      "loss": 0.6597,
+      "step": 45557
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.864422857142857e-06,
+      "loss": 0.8752,
+      "step": 45558
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8641542857142857e-06,
+      "loss": 1.0414,
+      "step": 45559
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8638857142857143e-06,
+      "loss": 1.1004,
+      "step": 45560
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.863617142857143e-06,
+      "loss": 1.1872,
+      "step": 45561
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8633485714285717e-06,
+      "loss": 0.6837,
+      "step": 45562
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8630799999999999e-06,
+      "loss": 0.7509,
+      "step": 45563
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8628114285714287e-06,
+      "loss": 1.1374,
+      "step": 45564
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8625428571428573e-06,
+      "loss": 1.4966,
+      "step": 45565
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8622742857142859e-06,
+      "loss": 1.135,
+      "step": 45566
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8620057142857143e-06,
+      "loss": 1.387,
+      "step": 45567
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8617371428571429e-06,
+      "loss": 0.8578,
+      "step": 45568
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8614685714285715e-06,
+      "loss": 0.6651,
+      "step": 45569
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8612e-06,
+      "loss": 0.9591,
+      "step": 45570
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8609314285714289e-06,
+      "loss": 1.1875,
+      "step": 45571
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.860662857142857e-06,
+      "loss": 0.8736,
+      "step": 45572
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8603942857142857e-06,
+      "loss": 1.7639,
+      "step": 45573
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8601257142857145e-06,
+      "loss": 1.2877,
+      "step": 45574
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.859857142857143e-06,
+      "loss": 0.8965,
+      "step": 45575
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8595885714285712e-06,
+      "loss": 1.2065,
+      "step": 45576
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.85932e-06,
+      "loss": 1.0487,
+      "step": 45577
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8590514285714287e-06,
+      "loss": 1.1317,
+      "step": 45578
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8587828571428573e-06,
+      "loss": 0.8955,
+      "step": 45579
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8585142857142859e-06,
+      "loss": 0.9766,
+      "step": 45580
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8582457142857142e-06,
+      "loss": 1.5303,
+      "step": 45581
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8579771428571428e-06,
+      "loss": 1.229,
+      "step": 45582
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8577085714285714e-06,
+      "loss": 1.5513,
+      "step": 45583
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8574400000000003e-06,
+      "loss": 1.2553,
+      "step": 45584
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8571714285714284e-06,
+      "loss": 1.1348,
+      "step": 45585
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.856902857142857e-06,
+      "loss": 0.9051,
+      "step": 45586
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.8566342857142858e-06,
+      "loss": 1.2782,
+      "step": 45587
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8563657142857144e-06,
+      "loss": 1.0412,
+      "step": 45588
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.856097142857143e-06,
+      "loss": 1.6094,
+      "step": 45589
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8558285714285714e-06,
+      "loss": 1.6403,
+      "step": 45590
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.85556e-06,
+      "loss": 1.0713,
+      "step": 45591
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8552914285714286e-06,
+      "loss": 1.1547,
+      "step": 45592
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8550228571428572e-06,
+      "loss": 1.369,
+      "step": 45593
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.854754285714286e-06,
+      "loss": 0.7489,
+      "step": 45594
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8544857142857142e-06,
+      "loss": 1.2331,
+      "step": 45595
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8542171428571428e-06,
+      "loss": 1.2294,
+      "step": 45596
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8539485714285716e-06,
+      "loss": 0.9065,
+      "step": 45597
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8536800000000002e-06,
+      "loss": 0.8934,
+      "step": 45598
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8534114285714284e-06,
+      "loss": 0.9909,
+      "step": 45599
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8531428571428572e-06,
+      "loss": 1.1531,
+      "step": 45600
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8528742857142858e-06,
+      "loss": 0.974,
+      "step": 45601
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8526057142857144e-06,
+      "loss": 0.8746,
+      "step": 45602
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.852337142857143e-06,
+      "loss": 0.7267,
+      "step": 45603
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8520685714285714e-06,
+      "loss": 1.2385,
+      "step": 45604
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8518e-06,
+      "loss": 1.5428,
+      "step": 45605
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8515314285714286e-06,
+      "loss": 0.6892,
+      "step": 45606
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8512628571428574e-06,
+      "loss": 1.1917,
+      "step": 45607
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8509942857142856e-06,
+      "loss": 0.7306,
+      "step": 45608
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8507257142857142e-06,
+      "loss": 0.7058,
+      "step": 45609
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.850457142857143e-06,
+      "loss": 0.7501,
+      "step": 45610
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8501885714285716e-06,
+      "loss": 0.8943,
+      "step": 45611
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8499200000000002e-06,
+      "loss": 0.7335,
+      "step": 45612
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8496514285714286e-06,
+      "loss": 0.9297,
+      "step": 45613
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8493828571428572e-06,
+      "loss": 1.3852,
+      "step": 45614
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8491142857142858e-06,
+      "loss": 0.995,
+      "step": 45615
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8488457142857144e-06,
+      "loss": 1.1322,
+      "step": 45616
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8485771428571432e-06,
+      "loss": 0.9025,
+      "step": 45617
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8483085714285714e-06,
+      "loss": 0.9465,
+      "step": 45618
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.84804e-06,
+      "loss": 1.006,
+      "step": 45619
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8477714285714288e-06,
+      "loss": 1.1821,
+      "step": 45620
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8475028571428574e-06,
+      "loss": 1.5475,
+      "step": 45621
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8472342857142856e-06,
+      "loss": 1.0311,
+      "step": 45622
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8469657142857144e-06,
+      "loss": 1.0791,
+      "step": 45623
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.846697142857143e-06,
+      "loss": 1.6218,
+      "step": 45624
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8464285714285716e-06,
+      "loss": 1.0657,
+      "step": 45625
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8461600000000002e-06,
+      "loss": 0.8532,
+      "step": 45626
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8458914285714286e-06,
+      "loss": 1.2372,
+      "step": 45627
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8456228571428572e-06,
+      "loss": 1.2205,
+      "step": 45628
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8453542857142858e-06,
+      "loss": 1.1409,
+      "step": 45629
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8450857142857146e-06,
+      "loss": 0.9634,
+      "step": 45630
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8448171428571427e-06,
+      "loss": 1.2508,
+      "step": 45631
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8445485714285713e-06,
+      "loss": 1.2511,
+      "step": 45632
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8442800000000002e-06,
+      "loss": 0.7983,
+      "step": 45633
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8440114285714288e-06,
+      "loss": 1.0618,
+      "step": 45634
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8437428571428574e-06,
+      "loss": 1.0773,
+      "step": 45635
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8434742857142857e-06,
+      "loss": 1.3705,
+      "step": 45636
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8432057142857143e-06,
+      "loss": 0.9388,
+      "step": 45637
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.842937142857143e-06,
+      "loss": 1.3937,
+      "step": 45638
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8426685714285715e-06,
+      "loss": 1.4273,
+      "step": 45639
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8424e-06,
+      "loss": 0.9009,
+      "step": 45640
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8421314285714285e-06,
+      "loss": 0.8283,
+      "step": 45641
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8418628571428571e-06,
+      "loss": 1.0739,
+      "step": 45642
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.841594285714286e-06,
+      "loss": 1.2426,
+      "step": 45643
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8413257142857145e-06,
+      "loss": 1.1915,
+      "step": 45644
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8410571428571427e-06,
+      "loss": 1.1773,
+      "step": 45645
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8407885714285715e-06,
+      "loss": 0.7657,
+      "step": 45646
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8405200000000001e-06,
+      "loss": 1.1451,
+      "step": 45647
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8402514285714287e-06,
+      "loss": 0.7895,
+      "step": 45648
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8399828571428573e-06,
+      "loss": 1.0189,
+      "step": 45649
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8397142857142857e-06,
+      "loss": 1.5187,
+      "step": 45650
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8394457142857143e-06,
+      "loss": 1.4005,
+      "step": 45651
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.839177142857143e-06,
+      "loss": 1.1638,
+      "step": 45652
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8389085714285717e-06,
+      "loss": 1.5239,
+      "step": 45653
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.83864e-06,
+      "loss": 0.9808,
+      "step": 45654
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8383714285714285e-06,
+      "loss": 1.0884,
+      "step": 45655
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8381028571428573e-06,
+      "loss": 0.7472,
+      "step": 45656
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.837834285714286e-06,
+      "loss": 1.0953,
+      "step": 45657
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8375657142857145e-06,
+      "loss": 1.0058,
+      "step": 45658
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.837297142857143e-06,
+      "loss": 0.9181,
+      "step": 45659
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8370285714285715e-06,
+      "loss": 1.2569,
+      "step": 45660
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.83676e-06,
+      "loss": 0.8618,
+      "step": 45661
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8364914285714287e-06,
+      "loss": 1.1537,
+      "step": 45662
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.836222857142857e-06,
+      "loss": 1.0159,
+      "step": 45663
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8359542857142857e-06,
+      "loss": 1.0823,
+      "step": 45664
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8356857142857143e-06,
+      "loss": 0.9771,
+      "step": 45665
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.835417142857143e-06,
+      "loss": 1.2248,
+      "step": 45666
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8351485714285717e-06,
+      "loss": 1.4421,
+      "step": 45667
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8348799999999999e-06,
+      "loss": 1.2974,
+      "step": 45668
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8346114285714287e-06,
+      "loss": 0.9256,
+      "step": 45669
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8343428571428573e-06,
+      "loss": 1.0221,
+      "step": 45670
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8340742857142859e-06,
+      "loss": 1.1596,
+      "step": 45671
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.833805714285714e-06,
+      "loss": 1.3021,
+      "step": 45672
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8335371428571429e-06,
+      "loss": 0.8674,
+      "step": 45673
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8332685714285715e-06,
+      "loss": 1.1369,
+      "step": 45674
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.833e-06,
+      "loss": 0.9453,
+      "step": 45675
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8327314285714289e-06,
+      "loss": 0.8762,
+      "step": 45676
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.832462857142857e-06,
+      "loss": 1.56,
+      "step": 45677
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8321942857142857e-06,
+      "loss": 1.1613,
+      "step": 45678
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8319257142857145e-06,
+      "loss": 1.3353,
+      "step": 45679
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.831657142857143e-06,
+      "loss": 0.9302,
+      "step": 45680
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8313885714285717e-06,
+      "loss": 0.9552,
+      "step": 45681
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8311199999999998e-06,
+      "loss": 1.0943,
+      "step": 45682
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8308514285714287e-06,
+      "loss": 1.0428,
+      "step": 45683
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8305828571428573e-06,
+      "loss": 1.3077,
+      "step": 45684
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8303142857142859e-06,
+      "loss": 1.0326,
+      "step": 45685
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8300457142857142e-06,
+      "loss": 1.2509,
+      "step": 45686
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8297771428571428e-06,
+      "loss": 1.0753,
+      "step": 45687
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8295085714285714e-06,
+      "loss": 1.392,
+      "step": 45688
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8292400000000003e-06,
+      "loss": 0.9964,
+      "step": 45689
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8289714285714289e-06,
+      "loss": 1.3685,
+      "step": 45690
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.828702857142857e-06,
+      "loss": 1.5399,
+      "step": 45691
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8284342857142856e-06,
+      "loss": 1.3111,
+      "step": 45692
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8281657142857144e-06,
+      "loss": 1.0998,
+      "step": 45693
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.827897142857143e-06,
+      "loss": 1.3718,
+      "step": 45694
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8276285714285712e-06,
+      "loss": 0.9742,
+      "step": 45695
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.82736e-06,
+      "loss": 1.3587,
+      "step": 45696
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8270914285714286e-06,
+      "loss": 1.2167,
+      "step": 45697
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8268228571428572e-06,
+      "loss": 0.8366,
+      "step": 45698
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.826554285714286e-06,
+      "loss": 1.3013,
+      "step": 45699
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8262857142857142e-06,
+      "loss": 1.1304,
+      "step": 45700
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8260171428571428e-06,
+      "loss": 1.1505,
+      "step": 45701
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8257485714285716e-06,
+      "loss": 1.2355,
+      "step": 45702
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8254800000000002e-06,
+      "loss": 1.2643,
+      "step": 45703
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8252114285714288e-06,
+      "loss": 1.1226,
+      "step": 45704
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.824942857142857e-06,
+      "loss": 1.0902,
+      "step": 45705
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8246742857142858e-06,
+      "loss": 0.7995,
+      "step": 45706
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8244057142857144e-06,
+      "loss": 1.2823,
+      "step": 45707
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.824137142857143e-06,
+      "loss": 1.1881,
+      "step": 45708
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8238685714285714e-06,
+      "loss": 1.3671,
+      "step": 45709
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8236e-06,
+      "loss": 1.0299,
+      "step": 45710
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8233314285714286e-06,
+      "loss": 0.9979,
+      "step": 45711
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8230628571428574e-06,
+      "loss": 0.977,
+      "step": 45712
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.822794285714286e-06,
+      "loss": 0.738,
+      "step": 45713
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8225257142857142e-06,
+      "loss": 1.4252,
+      "step": 45714
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8222571428571428e-06,
+      "loss": 1.2113,
+      "step": 45715
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8219885714285716e-06,
+      "loss": 1.3898,
+      "step": 45716
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8217200000000002e-06,
+      "loss": 0.9593,
+      "step": 45717
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8214514285714284e-06,
+      "loss": 1.4138,
+      "step": 45718
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8211828571428572e-06,
+      "loss": 1.1726,
+      "step": 45719
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8209142857142858e-06,
+      "loss": 1.1292,
+      "step": 45720
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8206457142857144e-06,
+      "loss": 1.4255,
+      "step": 45721
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8203771428571432e-06,
+      "loss": 1.4018,
+      "step": 45722
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8201085714285714e-06,
+      "loss": 0.8004,
+      "step": 45723
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.81984e-06,
+      "loss": 1.1813,
+      "step": 45724
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8195714285714286e-06,
+      "loss": 1.0091,
+      "step": 45725
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8193028571428574e-06,
+      "loss": 0.9027,
+      "step": 45726
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8190342857142856e-06,
+      "loss": 1.3441,
+      "step": 45727
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8187657142857142e-06,
+      "loss": 0.8983,
+      "step": 45728
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.818497142857143e-06,
+      "loss": 1.3201,
+      "step": 45729
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8182285714285716e-06,
+      "loss": 0.984,
+      "step": 45730
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8179600000000002e-06,
+      "loss": 1.2046,
+      "step": 45731
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8176914285714286e-06,
+      "loss": 0.888,
+      "step": 45732
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8174228571428572e-06,
+      "loss": 1.3014,
+      "step": 45733
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8171542857142858e-06,
+      "loss": 1.0618,
+      "step": 45734
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8168857142857144e-06,
+      "loss": 1.0191,
+      "step": 45735
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8166171428571432e-06,
+      "loss": 1.3392,
+      "step": 45736
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8163485714285713e-06,
+      "loss": 0.718,
+      "step": 45737
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.81608e-06,
+      "loss": 0.9823,
+      "step": 45738
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8158114285714288e-06,
+      "loss": 0.9705,
+      "step": 45739
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8155428571428574e-06,
+      "loss": 1.3683,
+      "step": 45740
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8152742857142855e-06,
+      "loss": 0.8683,
+      "step": 45741
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8150057142857143e-06,
+      "loss": 0.8412,
+      "step": 45742
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.814737142857143e-06,
+      "loss": 1.1256,
+      "step": 45743
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8144685714285715e-06,
+      "loss": 1.2934,
+      "step": 45744
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8142000000000001e-06,
+      "loss": 1.1289,
+      "step": 45745
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8139314285714285e-06,
+      "loss": 1.2386,
+      "step": 45746
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8136628571428571e-06,
+      "loss": 1.0185,
+      "step": 45747
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8133942857142857e-06,
+      "loss": 1.1854,
+      "step": 45748
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8131257142857145e-06,
+      "loss": 1.5806,
+      "step": 45749
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8128571428571427e-06,
+      "loss": 0.8062,
+      "step": 45750
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8125885714285713e-06,
+      "loss": 1.5421,
+      "step": 45751
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8123200000000001e-06,
+      "loss": 1.1405,
+      "step": 45752
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8120514285714287e-06,
+      "loss": 1.128,
+      "step": 45753
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8117828571428573e-06,
+      "loss": 0.9954,
+      "step": 45754
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8115142857142857e-06,
+      "loss": 1.0905,
+      "step": 45755
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8112457142857143e-06,
+      "loss": 1.3632,
+      "step": 45756
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.810977142857143e-06,
+      "loss": 0.7906,
+      "step": 45757
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8107085714285715e-06,
+      "loss": 0.7409,
+      "step": 45758
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.81044e-06,
+      "loss": 0.8315,
+      "step": 45759
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.8101714285714285e-06,
+      "loss": 1.1491,
+      "step": 45760
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.809902857142857e-06,
+      "loss": 1.2378,
+      "step": 45761
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.809634285714286e-06,
+      "loss": 1.3292,
+      "step": 45762
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8093657142857145e-06,
+      "loss": 0.6745,
+      "step": 45763
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8090971428571427e-06,
+      "loss": 1.4629,
+      "step": 45764
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8088285714285715e-06,
+      "loss": 1.3163,
+      "step": 45765
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.80856e-06,
+      "loss": 0.9168,
+      "step": 45766
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8082914285714287e-06,
+      "loss": 1.2687,
+      "step": 45767
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8080228571428573e-06,
+      "loss": 1.313,
+      "step": 45768
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8077542857142857e-06,
+      "loss": 1.1014,
+      "step": 45769
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8074857142857143e-06,
+      "loss": 1.0118,
+      "step": 45770
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8072171428571429e-06,
+      "loss": 1.0052,
+      "step": 45771
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8069485714285717e-06,
+      "loss": 1.1017,
+      "step": 45772
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8066799999999999e-06,
+      "loss": 0.7313,
+      "step": 45773
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8064114285714285e-06,
+      "loss": 1.0124,
+      "step": 45774
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8061428571428573e-06,
+      "loss": 1.0802,
+      "step": 45775
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8058742857142859e-06,
+      "loss": 1.1251,
+      "step": 45776
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8056057142857145e-06,
+      "loss": 0.6342,
+      "step": 45777
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8053371428571429e-06,
+      "loss": 1.2509,
+      "step": 45778
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8050685714285715e-06,
+      "loss": 1.4809,
+      "step": 45779
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8048e-06,
+      "loss": 1.1957,
+      "step": 45780
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8045314285714287e-06,
+      "loss": 1.252,
+      "step": 45781
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.804262857142857e-06,
+      "loss": 1.1004,
+      "step": 45782
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8039942857142857e-06,
+      "loss": 1.1243,
+      "step": 45783
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8037257142857143e-06,
+      "loss": 1.0898,
+      "step": 45784
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.803457142857143e-06,
+      "loss": 1.042,
+      "step": 45785
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8031885714285717e-06,
+      "loss": 0.9139,
+      "step": 45786
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8029199999999998e-06,
+      "loss": 1.0073,
+      "step": 45787
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8026514285714287e-06,
+      "loss": 1.3352,
+      "step": 45788
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8023828571428573e-06,
+      "loss": 1.145,
+      "step": 45789
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8021142857142859e-06,
+      "loss": 1.179,
+      "step": 45790
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8018457142857145e-06,
+      "loss": 1.3078,
+      "step": 45791
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8015771428571428e-06,
+      "loss": 0.8483,
+      "step": 45792
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8013085714285714e-06,
+      "loss": 1.175,
+      "step": 45793
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.80104e-06,
+      "loss": 0.7544,
+      "step": 45794
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8007714285714289e-06,
+      "loss": 0.9411,
+      "step": 45795
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.800502857142857e-06,
+      "loss": 1.1234,
+      "step": 45796
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.8002342857142856e-06,
+      "loss": 1.2848,
+      "step": 45797
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7999657142857144e-06,
+      "loss": 1.2658,
+      "step": 45798
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.799697142857143e-06,
+      "loss": 1.3195,
+      "step": 45799
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7994285714285716e-06,
+      "loss": 0.8314,
+      "step": 45800
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.79916e-06,
+      "loss": 0.7488,
+      "step": 45801
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7988914285714286e-06,
+      "loss": 1.1605,
+      "step": 45802
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7986228571428572e-06,
+      "loss": 1.0334,
+      "step": 45803
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7983542857142858e-06,
+      "loss": 0.7248,
+      "step": 45804
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7980857142857142e-06,
+      "loss": 1.1145,
+      "step": 45805
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7978171428571428e-06,
+      "loss": 0.9218,
+      "step": 45806
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7975485714285714e-06,
+      "loss": 1.0177,
+      "step": 45807
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7972800000000002e-06,
+      "loss": 1.5302,
+      "step": 45808
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7970114285714288e-06,
+      "loss": 0.9823,
+      "step": 45809
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.796742857142857e-06,
+      "loss": 1.6847,
+      "step": 45810
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7964742857142858e-06,
+      "loss": 1.4317,
+      "step": 45811
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7962057142857144e-06,
+      "loss": 0.9625,
+      "step": 45812
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.795937142857143e-06,
+      "loss": 0.9087,
+      "step": 45813
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7956685714285714e-06,
+      "loss": 0.7104,
+      "step": 45814
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7954e-06,
+      "loss": 1.0186,
+      "step": 45815
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7951314285714286e-06,
+      "loss": 1.0264,
+      "step": 45816
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7948628571428572e-06,
+      "loss": 1.2637,
+      "step": 45817
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.794594285714286e-06,
+      "loss": 1.1708,
+      "step": 45818
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7943257142857142e-06,
+      "loss": 0.8169,
+      "step": 45819
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7940571428571428e-06,
+      "loss": 0.8458,
+      "step": 45820
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7937885714285716e-06,
+      "loss": 1.4144,
+      "step": 45821
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7935200000000002e-06,
+      "loss": 0.8392,
+      "step": 45822
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7932514285714288e-06,
+      "loss": 1.0515,
+      "step": 45823
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7929828571428572e-06,
+      "loss": 1.2866,
+      "step": 45824
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7927142857142858e-06,
+      "loss": 0.7803,
+      "step": 45825
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7924457142857144e-06,
+      "loss": 0.8308,
+      "step": 45826
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.792177142857143e-06,
+      "loss": 1.3879,
+      "step": 45827
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7919085714285714e-06,
+      "loss": 0.8182,
+      "step": 45828
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.79164e-06,
+      "loss": 1.2607,
+      "step": 45829
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7913714285714286e-06,
+      "loss": 1.2049,
+      "step": 45830
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7911028571428574e-06,
+      "loss": 1.3187,
+      "step": 45831
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.790834285714286e-06,
+      "loss": 1.0638,
+      "step": 45832
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7905657142857142e-06,
+      "loss": 0.8569,
+      "step": 45833
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.790297142857143e-06,
+      "loss": 1.2295,
+      "step": 45834
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7900285714285716e-06,
+      "loss": 1.1984,
+      "step": 45835
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7897600000000002e-06,
+      "loss": 1.2262,
+      "step": 45836
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7894914285714286e-06,
+      "loss": 0.9094,
+      "step": 45837
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7892228571428572e-06,
+      "loss": 1.3203,
+      "step": 45838
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7889542857142858e-06,
+      "loss": 0.6899,
+      "step": 45839
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7886857142857144e-06,
+      "loss": 1.1595,
+      "step": 45840
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7884171428571432e-06,
+      "loss": 0.6579,
+      "step": 45841
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7881485714285713e-06,
+      "loss": 0.7136,
+      "step": 45842
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.78788e-06,
+      "loss": 1.1458,
+      "step": 45843
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7876114285714288e-06,
+      "loss": 1.0786,
+      "step": 45844
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7873428571428574e-06,
+      "loss": 1.3061,
+      "step": 45845
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7870742857142855e-06,
+      "loss": 1.0465,
+      "step": 45846
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7868057142857143e-06,
+      "loss": 0.897,
+      "step": 45847
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.786537142857143e-06,
+      "loss": 0.971,
+      "step": 45848
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7862685714285715e-06,
+      "loss": 1.2555,
+      "step": 45849
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7860000000000001e-06,
+      "loss": 0.6656,
+      "step": 45850
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7857314285714285e-06,
+      "loss": 1.2239,
+      "step": 45851
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7854628571428571e-06,
+      "loss": 1.3112,
+      "step": 45852
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7851942857142857e-06,
+      "loss": 0.8096,
+      "step": 45853
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7849257142857145e-06,
+      "loss": 1.142,
+      "step": 45854
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7846571428571431e-06,
+      "loss": 0.8524,
+      "step": 45855
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7843885714285713e-06,
+      "loss": 1.1112,
+      "step": 45856
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7841200000000001e-06,
+      "loss": 0.6428,
+      "step": 45857
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7838514285714287e-06,
+      "loss": 1.0586,
+      "step": 45858
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7835828571428573e-06,
+      "loss": 1.3769,
+      "step": 45859
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7833142857142857e-06,
+      "loss": 1.2962,
+      "step": 45860
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7830457142857143e-06,
+      "loss": 1.3161,
+      "step": 45861
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.782777142857143e-06,
+      "loss": 1.1681,
+      "step": 45862
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7825085714285715e-06,
+      "loss": 1.2577,
+      "step": 45863
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7822400000000003e-06,
+      "loss": 1.2815,
+      "step": 45864
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7819714285714285e-06,
+      "loss": 1.5603,
+      "step": 45865
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.781702857142857e-06,
+      "loss": 1.357,
+      "step": 45866
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.781434285714286e-06,
+      "loss": 1.5049,
+      "step": 45867
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7811657142857145e-06,
+      "loss": 1.1824,
+      "step": 45868
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7808971428571427e-06,
+      "loss": 0.8427,
+      "step": 45869
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7806285714285715e-06,
+      "loss": 0.9663,
+      "step": 45870
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.78036e-06,
+      "loss": 1.269,
+      "step": 45871
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7800914285714287e-06,
+      "loss": 1.1106,
+      "step": 45872
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7798228571428573e-06,
+      "loss": 1.2455,
+      "step": 45873
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7795542857142857e-06,
+      "loss": 1.2723,
+      "step": 45874
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7792857142857143e-06,
+      "loss": 1.2921,
+      "step": 45875
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7790171428571429e-06,
+      "loss": 0.9616,
+      "step": 45876
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7787485714285717e-06,
+      "loss": 1.4547,
+      "step": 45877
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7784799999999999e-06,
+      "loss": 1.6053,
+      "step": 45878
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7782114285714285e-06,
+      "loss": 1.203,
+      "step": 45879
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7779428571428573e-06,
+      "loss": 1.1115,
+      "step": 45880
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7776742857142859e-06,
+      "loss": 1.4369,
+      "step": 45881
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7774057142857145e-06,
+      "loss": 1.0497,
+      "step": 45882
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7771371428571429e-06,
+      "loss": 1.2603,
+      "step": 45883
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7768685714285715e-06,
+      "loss": 1.1745,
+      "step": 45884
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7766e-06,
+      "loss": 0.9122,
+      "step": 45885
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7763314285714287e-06,
+      "loss": 1.1473,
+      "step": 45886
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7760628571428575e-06,
+      "loss": 1.3684,
+      "step": 45887
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7757942857142857e-06,
+      "loss": 0.9404,
+      "step": 45888
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7755257142857143e-06,
+      "loss": 1.0623,
+      "step": 45889
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.775257142857143e-06,
+      "loss": 0.8777,
+      "step": 45890
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7749885714285717e-06,
+      "loss": 1.2319,
+      "step": 45891
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7747199999999998e-06,
+      "loss": 1.0824,
+      "step": 45892
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7744514285714287e-06,
+      "loss": 1.4966,
+      "step": 45893
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7741828571428573e-06,
+      "loss": 1.1751,
+      "step": 45894
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7739142857142859e-06,
+      "loss": 1.0693,
+      "step": 45895
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7736457142857145e-06,
+      "loss": 1.3512,
+      "step": 45896
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7733771428571428e-06,
+      "loss": 1.2801,
+      "step": 45897
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7731085714285714e-06,
+      "loss": 1.2002,
+      "step": 45898
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.77284e-06,
+      "loss": 0.7689,
+      "step": 45899
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7725714285714288e-06,
+      "loss": 1.1294,
+      "step": 45900
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.772302857142857e-06,
+      "loss": 0.9248,
+      "step": 45901
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7720342857142856e-06,
+      "loss": 0.8759,
+      "step": 45902
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7717657142857144e-06,
+      "loss": 1.3213,
+      "step": 45903
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.771497142857143e-06,
+      "loss": 0.8778,
+      "step": 45904
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7712285714285716e-06,
+      "loss": 0.7925,
+      "step": 45905
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7709599999999998e-06,
+      "loss": 1.202,
+      "step": 45906
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7706914285714286e-06,
+      "loss": 1.3725,
+      "step": 45907
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7704228571428572e-06,
+      "loss": 0.8952,
+      "step": 45908
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7701542857142858e-06,
+      "loss": 1.2826,
+      "step": 45909
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7698857142857146e-06,
+      "loss": 1.5229,
+      "step": 45910
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7696171428571428e-06,
+      "loss": 0.9577,
+      "step": 45911
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7693485714285714e-06,
+      "loss": 1.2777,
+      "step": 45912
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7690800000000002e-06,
+      "loss": 1.2436,
+      "step": 45913
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7688114285714288e-06,
+      "loss": 0.939,
+      "step": 45914
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.768542857142857e-06,
+      "loss": 1.0509,
+      "step": 45915
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7682742857142858e-06,
+      "loss": 1.2397,
+      "step": 45916
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7680057142857144e-06,
+      "loss": 1.1656,
+      "step": 45917
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.767737142857143e-06,
+      "loss": 1.8336,
+      "step": 45918
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7674685714285716e-06,
+      "loss": 1.1268,
+      "step": 45919
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7672e-06,
+      "loss": 1.0915,
+      "step": 45920
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7669314285714286e-06,
+      "loss": 1.1887,
+      "step": 45921
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7666628571428572e-06,
+      "loss": 1.0666,
+      "step": 45922
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.766394285714286e-06,
+      "loss": 1.541,
+      "step": 45923
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7661257142857142e-06,
+      "loss": 0.9884,
+      "step": 45924
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7658571428571428e-06,
+      "loss": 1.289,
+      "step": 45925
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7655885714285716e-06,
+      "loss": 1.3924,
+      "step": 45926
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7653200000000002e-06,
+      "loss": 0.8881,
+      "step": 45927
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7650514285714288e-06,
+      "loss": 1.2165,
+      "step": 45928
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.764782857142857e-06,
+      "loss": 1.0832,
+      "step": 45929
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7645142857142858e-06,
+      "loss": 1.4311,
+      "step": 45930
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7642457142857144e-06,
+      "loss": 0.9131,
+      "step": 45931
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.763977142857143e-06,
+      "loss": 1.1842,
+      "step": 45932
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7637085714285714e-06,
+      "loss": 1.1638,
+      "step": 45933
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.76344e-06,
+      "loss": 0.804,
+      "step": 45934
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7631714285714286e-06,
+      "loss": 1.2125,
+      "step": 45935
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.7629028571428574e-06,
+      "loss": 1.1431,
+      "step": 45936
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.762634285714286e-06,
+      "loss": 1.3949,
+      "step": 45937
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7623657142857144e-06,
+      "loss": 0.9914,
+      "step": 45938
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7620971428571428e-06,
+      "loss": 0.8226,
+      "step": 45939
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7618285714285716e-06,
+      "loss": 1.114,
+      "step": 45940
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.76156e-06,
+      "loss": 1.6726,
+      "step": 45941
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7612914285714286e-06,
+      "loss": 1.1432,
+      "step": 45942
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7610228571428574e-06,
+      "loss": 1.1055,
+      "step": 45943
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7607542857142858e-06,
+      "loss": 0.6529,
+      "step": 45944
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7604857142857144e-06,
+      "loss": 1.0236,
+      "step": 45945
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.760217142857143e-06,
+      "loss": 1.3308,
+      "step": 45946
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7599485714285716e-06,
+      "loss": 1.4282,
+      "step": 45947
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.75968e-06,
+      "loss": 1.3813,
+      "step": 45948
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7594114285714285e-06,
+      "loss": 1.0229,
+      "step": 45949
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7591428571428573e-06,
+      "loss": 1.2292,
+      "step": 45950
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7588742857142857e-06,
+      "loss": 0.867,
+      "step": 45951
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7586057142857143e-06,
+      "loss": 1.1898,
+      "step": 45952
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.758337142857143e-06,
+      "loss": 1.1793,
+      "step": 45953
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7580685714285715e-06,
+      "loss": 1.0373,
+      "step": 45954
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7578e-06,
+      "loss": 0.9273,
+      "step": 45955
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7575314285714287e-06,
+      "loss": 0.8208,
+      "step": 45956
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7572628571428571e-06,
+      "loss": 1.5441,
+      "step": 45957
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7569942857142857e-06,
+      "loss": 1.0778,
+      "step": 45958
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7567257142857143e-06,
+      "loss": 1.0383,
+      "step": 45959
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.756457142857143e-06,
+      "loss": 1.0598,
+      "step": 45960
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7561885714285715e-06,
+      "loss": 0.9912,
+      "step": 45961
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.75592e-06,
+      "loss": 0.7421,
+      "step": 45962
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7556514285714287e-06,
+      "loss": 1.028,
+      "step": 45963
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7553828571428571e-06,
+      "loss": 1.6001,
+      "step": 45964
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7551142857142857e-06,
+      "loss": 1.0947,
+      "step": 45965
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7548457142857145e-06,
+      "loss": 1.097,
+      "step": 45966
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.754577142857143e-06,
+      "loss": 1.2374,
+      "step": 45967
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7543085714285715e-06,
+      "loss": 1.6253,
+      "step": 45968
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7540400000000001e-06,
+      "loss": 1.3193,
+      "step": 45969
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7537714285714287e-06,
+      "loss": 1.1863,
+      "step": 45970
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.753502857142857e-06,
+      "loss": 1.0817,
+      "step": 45971
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7532342857142857e-06,
+      "loss": 1.4194,
+      "step": 45972
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7529657142857143e-06,
+      "loss": 1.4118,
+      "step": 45973
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.752697142857143e-06,
+      "loss": 1.4045,
+      "step": 45974
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7524285714285715e-06,
+      "loss": 1.1455,
+      "step": 45975
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.75216e-06,
+      "loss": 1.2434,
+      "step": 45976
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7518914285714287e-06,
+      "loss": 0.9628,
+      "step": 45977
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.751622857142857e-06,
+      "loss": 1.1354,
+      "step": 45978
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7513542857142859e-06,
+      "loss": 1.4829,
+      "step": 45979
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7510857142857143e-06,
+      "loss": 1.1453,
+      "step": 45980
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7508171428571429e-06,
+      "loss": 1.2578,
+      "step": 45981
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7505485714285715e-06,
+      "loss": 0.9888,
+      "step": 45982
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.75028e-06,
+      "loss": 1.2765,
+      "step": 45983
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7500114285714287e-06,
+      "loss": 1.1109,
+      "step": 45984
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.749742857142857e-06,
+      "loss": 0.8484,
+      "step": 45985
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7494742857142859e-06,
+      "loss": 1.5375,
+      "step": 45986
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7492057142857143e-06,
+      "loss": 1.3878,
+      "step": 45987
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7489371428571429e-06,
+      "loss": 1.4783,
+      "step": 45988
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7486685714285717e-06,
+      "loss": 1.0579,
+      "step": 45989
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7484e-06,
+      "loss": 1.1143,
+      "step": 45990
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7481314285714287e-06,
+      "loss": 1.1382,
+      "step": 45991
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7478628571428573e-06,
+      "loss": 1.176,
+      "step": 45992
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7475942857142859e-06,
+      "loss": 1.1999,
+      "step": 45993
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7473257142857143e-06,
+      "loss": 1.0901,
+      "step": 45994
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7470571428571429e-06,
+      "loss": 1.3913,
+      "step": 45995
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7467885714285715e-06,
+      "loss": 1.3608,
+      "step": 45996
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.74652e-06,
+      "loss": 1.4943,
+      "step": 45997
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7462514285714287e-06,
+      "loss": 1.1803,
+      "step": 45998
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7459828571428573e-06,
+      "loss": 0.78,
+      "step": 45999
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7457142857142859e-06,
+      "loss": 1.1091,
+      "step": 46000
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7454457142857142e-06,
+      "loss": 0.6981,
+      "step": 46001
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.745177142857143e-06,
+      "loss": 1.0248,
+      "step": 46002
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7449085714285714e-06,
+      "loss": 1.2493,
+      "step": 46003
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.74464e-06,
+      "loss": 1.2738,
+      "step": 46004
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7443714285714286e-06,
+      "loss": 0.8778,
+      "step": 46005
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7441028571428572e-06,
+      "loss": 1.0262,
+      "step": 46006
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7438342857142858e-06,
+      "loss": 1.161,
+      "step": 46007
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7435657142857142e-06,
+      "loss": 0.7823,
+      "step": 46008
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.743297142857143e-06,
+      "loss": 1.1165,
+      "step": 46009
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7430285714285714e-06,
+      "loss": 1.4967,
+      "step": 46010
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.74276e-06,
+      "loss": 1.0826,
+      "step": 46011
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7424914285714286e-06,
+      "loss": 0.7702,
+      "step": 46012
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7422228571428572e-06,
+      "loss": 1.1187,
+      "step": 46013
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7419542857142858e-06,
+      "loss": 1.2884,
+      "step": 46014
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7416857142857144e-06,
+      "loss": 1.0382,
+      "step": 46015
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.741417142857143e-06,
+      "loss": 1.2752,
+      "step": 46016
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7411485714285714e-06,
+      "loss": 1.2136,
+      "step": 46017
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.74088e-06,
+      "loss": 1.0239,
+      "step": 46018
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7406114285714286e-06,
+      "loss": 1.1459,
+      "step": 46019
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7403428571428572e-06,
+      "loss": 1.0219,
+      "step": 46020
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7400742857142858e-06,
+      "loss": 1.1309,
+      "step": 46021
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7398057142857144e-06,
+      "loss": 1.2445,
+      "step": 46022
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.739537142857143e-06,
+      "loss": 1.1937,
+      "step": 46023
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7392685714285714e-06,
+      "loss": 0.9977,
+      "step": 46024
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7390000000000002e-06,
+      "loss": 1.067,
+      "step": 46025
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7387314285714286e-06,
+      "loss": 1.0057,
+      "step": 46026
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7384628571428572e-06,
+      "loss": 1.146,
+      "step": 46027
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7381942857142858e-06,
+      "loss": 1.0299,
+      "step": 46028
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7379257142857144e-06,
+      "loss": 1.2215,
+      "step": 46029
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.737657142857143e-06,
+      "loss": 0.8469,
+      "step": 46030
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7373885714285714e-06,
+      "loss": 0.8627,
+      "step": 46031
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7371200000000002e-06,
+      "loss": 1.2815,
+      "step": 46032
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7368514285714286e-06,
+      "loss": 0.992,
+      "step": 46033
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7365828571428572e-06,
+      "loss": 1.1022,
+      "step": 46034
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7363142857142858e-06,
+      "loss": 1.0098,
+      "step": 46035
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7360457142857144e-06,
+      "loss": 0.9696,
+      "step": 46036
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.735777142857143e-06,
+      "loss": 0.7741,
+      "step": 46037
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7355085714285716e-06,
+      "loss": 1.3237,
+      "step": 46038
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7352400000000002e-06,
+      "loss": 1.1122,
+      "step": 46039
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7349714285714286e-06,
+      "loss": 1.1453,
+      "step": 46040
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7347028571428572e-06,
+      "loss": 0.9878,
+      "step": 46041
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7344342857142858e-06,
+      "loss": 0.9565,
+      "step": 46042
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7341657142857144e-06,
+      "loss": 1.0124,
+      "step": 46043
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7338971428571428e-06,
+      "loss": 0.7144,
+      "step": 46044
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7336285714285716e-06,
+      "loss": 0.9404,
+      "step": 46045
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7333600000000002e-06,
+      "loss": 0.8983,
+      "step": 46046
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7330914285714286e-06,
+      "loss": 0.602,
+      "step": 46047
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7328228571428574e-06,
+      "loss": 0.647,
+      "step": 46048
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7325542857142858e-06,
+      "loss": 1.1138,
+      "step": 46049
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7322857142857144e-06,
+      "loss": 0.8462,
+      "step": 46050
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7320171428571427e-06,
+      "loss": 1.439,
+      "step": 46051
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7317485714285715e-06,
+      "loss": 0.9565,
+      "step": 46052
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7314800000000001e-06,
+      "loss": 1.2675,
+      "step": 46053
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7312114285714285e-06,
+      "loss": 1.2442,
+      "step": 46054
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7309428571428573e-06,
+      "loss": 0.8642,
+      "step": 46055
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7306742857142857e-06,
+      "loss": 0.8836,
+      "step": 46056
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7304057142857143e-06,
+      "loss": 1.1952,
+      "step": 46057
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.730137142857143e-06,
+      "loss": 1.0218,
+      "step": 46058
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7298685714285715e-06,
+      "loss": 1.393,
+      "step": 46059
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7296e-06,
+      "loss": 1.3714,
+      "step": 46060
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7293314285714285e-06,
+      "loss": 0.9771,
+      "step": 46061
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7290628571428573e-06,
+      "loss": 1.2715,
+      "step": 46062
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7287942857142857e-06,
+      "loss": 1.357,
+      "step": 46063
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7285257142857143e-06,
+      "loss": 0.9938,
+      "step": 46064
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.728257142857143e-06,
+      "loss": 1.0084,
+      "step": 46065
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7279885714285715e-06,
+      "loss": 1.1635,
+      "step": 46066
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.72772e-06,
+      "loss": 0.997,
+      "step": 46067
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7274514285714287e-06,
+      "loss": 0.9481,
+      "step": 46068
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7271828571428573e-06,
+      "loss": 1.2804,
+      "step": 46069
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7269142857142857e-06,
+      "loss": 0.8445,
+      "step": 46070
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7266457142857143e-06,
+      "loss": 0.9034,
+      "step": 46071
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.726377142857143e-06,
+      "loss": 0.9261,
+      "step": 46072
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7261085714285715e-06,
+      "loss": 1.0309,
+      "step": 46073
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.72584e-06,
+      "loss": 0.9794,
+      "step": 46074
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7255714285714287e-06,
+      "loss": 1.4437,
+      "step": 46075
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.725302857142857e-06,
+      "loss": 0.9443,
+      "step": 46076
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7250342857142857e-06,
+      "loss": 0.8706,
+      "step": 46077
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7247657142857145e-06,
+      "loss": 0.9214,
+      "step": 46078
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7244971428571429e-06,
+      "loss": 1.3027,
+      "step": 46079
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7242285714285715e-06,
+      "loss": 1.115,
+      "step": 46080
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.72396e-06,
+      "loss": 0.6544,
+      "step": 46081
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7236914285714287e-06,
+      "loss": 1.2526,
+      "step": 46082
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.723422857142857e-06,
+      "loss": 0.9828,
+      "step": 46083
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7231542857142857e-06,
+      "loss": 1.1148,
+      "step": 46084
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7228857142857145e-06,
+      "loss": 1.2676,
+      "step": 46085
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7226171428571429e-06,
+      "loss": 1.3905,
+      "step": 46086
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7223485714285715e-06,
+      "loss": 1.1544,
+      "step": 46087
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.72208e-06,
+      "loss": 0.944,
+      "step": 46088
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7218114285714287e-06,
+      "loss": 1.4291,
+      "step": 46089
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.721542857142857e-06,
+      "loss": 1.0431,
+      "step": 46090
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7212742857142859e-06,
+      "loss": 1.2884,
+      "step": 46091
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7210057142857145e-06,
+      "loss": 1.4211,
+      "step": 46092
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7207371428571429e-06,
+      "loss": 0.9551,
+      "step": 46093
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7204685714285715e-06,
+      "loss": 0.9049,
+      "step": 46094
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7202e-06,
+      "loss": 1.1269,
+      "step": 46095
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7199314285714287e-06,
+      "loss": 1.1933,
+      "step": 46096
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.719662857142857e-06,
+      "loss": 1.3917,
+      "step": 46097
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7193942857142859e-06,
+      "loss": 1.0512,
+      "step": 46098
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7191257142857143e-06,
+      "loss": 1.4233,
+      "step": 46099
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7188571428571429e-06,
+      "loss": 1.3526,
+      "step": 46100
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7185885714285717e-06,
+      "loss": 1.4037,
+      "step": 46101
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.71832e-06,
+      "loss": 1.1823,
+      "step": 46102
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7180514285714286e-06,
+      "loss": 0.9505,
+      "step": 46103
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7177828571428572e-06,
+      "loss": 0.7189,
+      "step": 46104
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7175142857142858e-06,
+      "loss": 0.8214,
+      "step": 46105
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7172457142857142e-06,
+      "loss": 1.0932,
+      "step": 46106
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7169771428571428e-06,
+      "loss": 1.3216,
+      "step": 46107
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7167085714285716e-06,
+      "loss": 1.0134,
+      "step": 46108
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.71644e-06,
+      "loss": 1.1174,
+      "step": 46109
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7161714285714286e-06,
+      "loss": 1.2567,
+      "step": 46110
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7159028571428572e-06,
+      "loss": 1.152,
+      "step": 46111
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.7156342857142858e-06,
+      "loss": 1.0198,
+      "step": 46112
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7153657142857142e-06,
+      "loss": 1.328,
+      "step": 46113
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.715097142857143e-06,
+      "loss": 0.9674,
+      "step": 46114
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7148285714285714e-06,
+      "loss": 1.1311,
+      "step": 46115
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.71456e-06,
+      "loss": 1.3357,
+      "step": 46116
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7142914285714286e-06,
+      "loss": 1.3438,
+      "step": 46117
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7140228571428572e-06,
+      "loss": 1.137,
+      "step": 46118
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7137542857142858e-06,
+      "loss": 1.1383,
+      "step": 46119
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7134857142857142e-06,
+      "loss": 1.1092,
+      "step": 46120
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.713217142857143e-06,
+      "loss": 0.6058,
+      "step": 46121
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7129485714285714e-06,
+      "loss": 0.8103,
+      "step": 46122
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.71268e-06,
+      "loss": 1.1956,
+      "step": 46123
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7124114285714288e-06,
+      "loss": 1.07,
+      "step": 46124
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7121428571428572e-06,
+      "loss": 0.7916,
+      "step": 46125
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7118742857142858e-06,
+      "loss": 0.765,
+      "step": 46126
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7116057142857144e-06,
+      "loss": 1.0053,
+      "step": 46127
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.711337142857143e-06,
+      "loss": 1.2285,
+      "step": 46128
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7110685714285714e-06,
+      "loss": 1.0024,
+      "step": 46129
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7108e-06,
+      "loss": 1.0957,
+      "step": 46130
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7105314285714286e-06,
+      "loss": 1.5101,
+      "step": 46131
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7102628571428572e-06,
+      "loss": 1.2312,
+      "step": 46132
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7099942857142858e-06,
+      "loss": 0.8528,
+      "step": 46133
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7097257142857144e-06,
+      "loss": 1.2115,
+      "step": 46134
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.709457142857143e-06,
+      "loss": 1.2707,
+      "step": 46135
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7091885714285714e-06,
+      "loss": 1.1281,
+      "step": 46136
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7089200000000002e-06,
+      "loss": 1.2507,
+      "step": 46137
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7086514285714286e-06,
+      "loss": 0.9497,
+      "step": 46138
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7083828571428572e-06,
+      "loss": 0.9832,
+      "step": 46139
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7081142857142858e-06,
+      "loss": 1.3057,
+      "step": 46140
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7078457142857144e-06,
+      "loss": 1.0075,
+      "step": 46141
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.707577142857143e-06,
+      "loss": 1.4543,
+      "step": 46142
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7073085714285714e-06,
+      "loss": 1.0738,
+      "step": 46143
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7070400000000002e-06,
+      "loss": 1.2282,
+      "step": 46144
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7067714285714286e-06,
+      "loss": 0.9496,
+      "step": 46145
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7065028571428572e-06,
+      "loss": 1.1847,
+      "step": 46146
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7062342857142858e-06,
+      "loss": 1.5248,
+      "step": 46147
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7059657142857144e-06,
+      "loss": 1.2622,
+      "step": 46148
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.705697142857143e-06,
+      "loss": 1.28,
+      "step": 46149
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7054285714285716e-06,
+      "loss": 1.058,
+      "step": 46150
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7051600000000002e-06,
+      "loss": 1.0024,
+      "step": 46151
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7048914285714286e-06,
+      "loss": 1.529,
+      "step": 46152
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7046228571428572e-06,
+      "loss": 1.2096,
+      "step": 46153
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7043542857142857e-06,
+      "loss": 1.2091,
+      "step": 46154
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7040857142857143e-06,
+      "loss": 1.2505,
+      "step": 46155
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.703817142857143e-06,
+      "loss": 0.8377,
+      "step": 46156
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7035485714285715e-06,
+      "loss": 1.0611,
+      "step": 46157
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7032800000000001e-06,
+      "loss": 1.1519,
+      "step": 46158
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7030114285714285e-06,
+      "loss": 1.3005,
+      "step": 46159
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7027428571428573e-06,
+      "loss": 1.119,
+      "step": 46160
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7024742857142857e-06,
+      "loss": 0.7145,
+      "step": 46161
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7022057142857143e-06,
+      "loss": 0.7368,
+      "step": 46162
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7019371428571427e-06,
+      "loss": 0.8947,
+      "step": 46163
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7016685714285715e-06,
+      "loss": 1.1923,
+      "step": 46164
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7014000000000001e-06,
+      "loss": 1.4317,
+      "step": 46165
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7011314285714285e-06,
+      "loss": 1.2532,
+      "step": 46166
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7008628571428573e-06,
+      "loss": 1.1946,
+      "step": 46167
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7005942857142857e-06,
+      "loss": 0.6137,
+      "step": 46168
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.7003257142857143e-06,
+      "loss": 0.9759,
+      "step": 46169
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.700057142857143e-06,
+      "loss": 0.8487,
+      "step": 46170
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6997885714285715e-06,
+      "loss": 0.7916,
+      "step": 46171
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6995200000000001e-06,
+      "loss": 1.1183,
+      "step": 46172
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6992514285714285e-06,
+      "loss": 0.9273,
+      "step": 46173
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6989828571428573e-06,
+      "loss": 1.4202,
+      "step": 46174
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6987142857142857e-06,
+      "loss": 0.978,
+      "step": 46175
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6984457142857143e-06,
+      "loss": 1.054,
+      "step": 46176
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.698177142857143e-06,
+      "loss": 0.96,
+      "step": 46177
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6979085714285715e-06,
+      "loss": 0.8328,
+      "step": 46178
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6976399999999999e-06,
+      "loss": 1.3603,
+      "step": 46179
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6973714285714287e-06,
+      "loss": 1.323,
+      "step": 46180
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6971028571428573e-06,
+      "loss": 1.6251,
+      "step": 46181
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6968342857142857e-06,
+      "loss": 0.8487,
+      "step": 46182
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6965657142857145e-06,
+      "loss": 1.5541,
+      "step": 46183
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6962971428571429e-06,
+      "loss": 0.7299,
+      "step": 46184
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6960285714285715e-06,
+      "loss": 0.9963,
+      "step": 46185
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6957599999999999e-06,
+      "loss": 1.2682,
+      "step": 46186
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6954914285714287e-06,
+      "loss": 1.1575,
+      "step": 46187
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6952228571428573e-06,
+      "loss": 1.4732,
+      "step": 46188
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6949542857142857e-06,
+      "loss": 1.1063,
+      "step": 46189
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6946857142857145e-06,
+      "loss": 0.9916,
+      "step": 46190
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6944171428571429e-06,
+      "loss": 0.7982,
+      "step": 46191
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6941485714285715e-06,
+      "loss": 1.0721,
+      "step": 46192
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.69388e-06,
+      "loss": 1.5759,
+      "step": 46193
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6936114285714287e-06,
+      "loss": 1.322,
+      "step": 46194
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6933428571428573e-06,
+      "loss": 1.0356,
+      "step": 46195
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6930742857142857e-06,
+      "loss": 0.8298,
+      "step": 46196
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6928057142857145e-06,
+      "loss": 0.7205,
+      "step": 46197
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6925371428571429e-06,
+      "loss": 1.4441,
+      "step": 46198
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6922685714285715e-06,
+      "loss": 0.8731,
+      "step": 46199
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.692e-06,
+      "loss": 0.7889,
+      "step": 46200
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6917314285714287e-06,
+      "loss": 1.0435,
+      "step": 46201
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.691462857142857e-06,
+      "loss": 0.819,
+      "step": 46202
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6911942857142859e-06,
+      "loss": 1.1099,
+      "step": 46203
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6909257142857145e-06,
+      "loss": 1.1403,
+      "step": 46204
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6906571428571428e-06,
+      "loss": 1.1617,
+      "step": 46205
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6903885714285714e-06,
+      "loss": 1.568,
+      "step": 46206
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.69012e-06,
+      "loss": 0.9914,
+      "step": 46207
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6898514285714286e-06,
+      "loss": 1.2074,
+      "step": 46208
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.689582857142857e-06,
+      "loss": 1.1394,
+      "step": 46209
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6893142857142858e-06,
+      "loss": 1.0297,
+      "step": 46210
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6890457142857144e-06,
+      "loss": 1.1404,
+      "step": 46211
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6887771428571428e-06,
+      "loss": 0.8197,
+      "step": 46212
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6885085714285716e-06,
+      "loss": 1.1025,
+      "step": 46213
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.68824e-06,
+      "loss": 1.0884,
+      "step": 46214
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6879714285714286e-06,
+      "loss": 0.8189,
+      "step": 46215
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6877028571428572e-06,
+      "loss": 1.0179,
+      "step": 46216
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6874342857142858e-06,
+      "loss": 1.0949,
+      "step": 46217
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6871657142857142e-06,
+      "loss": 1.2761,
+      "step": 46218
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6868971428571428e-06,
+      "loss": 1.6089,
+      "step": 46219
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6866285714285716e-06,
+      "loss": 1.2345,
+      "step": 46220
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.68636e-06,
+      "loss": 1.3668,
+      "step": 46221
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6860914285714286e-06,
+      "loss": 0.9635,
+      "step": 46222
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6858228571428572e-06,
+      "loss": 0.9982,
+      "step": 46223
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6855542857142858e-06,
+      "loss": 1.4431,
+      "step": 46224
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6852857142857142e-06,
+      "loss": 1.1867,
+      "step": 46225
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.685017142857143e-06,
+      "loss": 1.1169,
+      "step": 46226
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6847485714285716e-06,
+      "loss": 1.2532,
+      "step": 46227
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.68448e-06,
+      "loss": 1.432,
+      "step": 46228
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6842114285714286e-06,
+      "loss": 1.4698,
+      "step": 46229
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6839428571428572e-06,
+      "loss": 0.8569,
+      "step": 46230
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6836742857142858e-06,
+      "loss": 1.1341,
+      "step": 46231
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6834057142857142e-06,
+      "loss": 1.1528,
+      "step": 46232
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.683137142857143e-06,
+      "loss": 1.1989,
+      "step": 46233
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6828685714285714e-06,
+      "loss": 1.0515,
+      "step": 46234
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6826e-06,
+      "loss": 1.2472,
+      "step": 46235
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6823314285714288e-06,
+      "loss": 0.9145,
+      "step": 46236
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6820628571428572e-06,
+      "loss": 1.1384,
+      "step": 46237
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6817942857142858e-06,
+      "loss": 1.0455,
+      "step": 46238
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6815257142857144e-06,
+      "loss": 1.3072,
+      "step": 46239
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.681257142857143e-06,
+      "loss": 1.3311,
+      "step": 46240
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6809885714285714e-06,
+      "loss": 0.9034,
+      "step": 46241
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.68072e-06,
+      "loss": 0.8677,
+      "step": 46242
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6804514285714288e-06,
+      "loss": 1.1238,
+      "step": 46243
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6801828571428572e-06,
+      "loss": 0.7741,
+      "step": 46244
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6799142857142858e-06,
+      "loss": 1.2162,
+      "step": 46245
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6796457142857144e-06,
+      "loss": 0.9121,
+      "step": 46246
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.679377142857143e-06,
+      "loss": 1.2223,
+      "step": 46247
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6791085714285714e-06,
+      "loss": 0.7158,
+      "step": 46248
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6788400000000002e-06,
+      "loss": 1.2193,
+      "step": 46249
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6785714285714286e-06,
+      "loss": 1.4663,
+      "step": 46250
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6783028571428572e-06,
+      "loss": 1.1075,
+      "step": 46251
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6780342857142858e-06,
+      "loss": 1.1652,
+      "step": 46252
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6777657142857144e-06,
+      "loss": 1.1251,
+      "step": 46253
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.677497142857143e-06,
+      "loss": 1.0857,
+      "step": 46254
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6772285714285714e-06,
+      "loss": 1.2324,
+      "step": 46255
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6769600000000002e-06,
+      "loss": 1.4098,
+      "step": 46256
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6766914285714285e-06,
+      "loss": 0.9746,
+      "step": 46257
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6764228571428571e-06,
+      "loss": 1.0308,
+      "step": 46258
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.676154285714286e-06,
+      "loss": 0.9565,
+      "step": 46259
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6758857142857143e-06,
+      "loss": 1.1772,
+      "step": 46260
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.675617142857143e-06,
+      "loss": 1.359,
+      "step": 46261
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6753485714285715e-06,
+      "loss": 1.019,
+      "step": 46262
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6750800000000001e-06,
+      "loss": 1.1941,
+      "step": 46263
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6748114285714285e-06,
+      "loss": 1.3523,
+      "step": 46264
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6745428571428571e-06,
+      "loss": 1.2597,
+      "step": 46265
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6742742857142857e-06,
+      "loss": 0.9866,
+      "step": 46266
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6740057142857143e-06,
+      "loss": 1.5844,
+      "step": 46267
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.673737142857143e-06,
+      "loss": 1.5414,
+      "step": 46268
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6734685714285715e-06,
+      "loss": 1.1459,
+      "step": 46269
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6732000000000001e-06,
+      "loss": 0.9243,
+      "step": 46270
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6729314285714285e-06,
+      "loss": 1.0154,
+      "step": 46271
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6726628571428573e-06,
+      "loss": 1.1136,
+      "step": 46272
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6723942857142857e-06,
+      "loss": 1.0242,
+      "step": 46273
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6721257142857143e-06,
+      "loss": 0.9011,
+      "step": 46274
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.671857142857143e-06,
+      "loss": 0.9863,
+      "step": 46275
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6715885714285715e-06,
+      "loss": 1.2043,
+      "step": 46276
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6713200000000001e-06,
+      "loss": 1.3276,
+      "step": 46277
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6710514285714285e-06,
+      "loss": 1.3932,
+      "step": 46278
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6707828571428573e-06,
+      "loss": 1.2421,
+      "step": 46279
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6705142857142857e-06,
+      "loss": 1.2476,
+      "step": 46280
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6702457142857143e-06,
+      "loss": 1.3653,
+      "step": 46281
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.669977142857143e-06,
+      "loss": 1.4222,
+      "step": 46282
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6697085714285715e-06,
+      "loss": 0.9941,
+      "step": 46283
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.66944e-06,
+      "loss": 1.6742,
+      "step": 46284
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6691714285714285e-06,
+      "loss": 1.1602,
+      "step": 46285
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6689028571428573e-06,
+      "loss": 0.9583,
+      "step": 46286
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.6686342857142857e-06,
+      "loss": 0.7533,
+      "step": 46287
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6683657142857143e-06,
+      "loss": 1.1601,
+      "step": 46288
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6680971428571429e-06,
+      "loss": 1.3283,
+      "step": 46289
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6678285714285715e-06,
+      "loss": 1.1737,
+      "step": 46290
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.66756e-06,
+      "loss": 1.1462,
+      "step": 46291
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6672914285714287e-06,
+      "loss": 1.1183,
+      "step": 46292
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6670228571428573e-06,
+      "loss": 0.8006,
+      "step": 46293
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6667542857142857e-06,
+      "loss": 0.594,
+      "step": 46294
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6664857142857145e-06,
+      "loss": 1.3655,
+      "step": 46295
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6662171428571429e-06,
+      "loss": 1.207,
+      "step": 46296
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6659485714285715e-06,
+      "loss": 1.4378,
+      "step": 46297
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.66568e-06,
+      "loss": 1.678,
+      "step": 46298
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6654114285714287e-06,
+      "loss": 0.9098,
+      "step": 46299
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6651428571428573e-06,
+      "loss": 0.773,
+      "step": 46300
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6648742857142857e-06,
+      "loss": 1.1184,
+      "step": 46301
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6646057142857145e-06,
+      "loss": 0.8862,
+      "step": 46302
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6643371428571429e-06,
+      "loss": 1.1651,
+      "step": 46303
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6640685714285715e-06,
+      "loss": 1.1271,
+      "step": 46304
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6638e-06,
+      "loss": 1.5146,
+      "step": 46305
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6635314285714287e-06,
+      "loss": 1.1013,
+      "step": 46306
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6632628571428573e-06,
+      "loss": 1.4293,
+      "step": 46307
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6629942857142856e-06,
+      "loss": 1.055,
+      "step": 46308
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6627257142857145e-06,
+      "loss": 1.2632,
+      "step": 46309
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6624571428571428e-06,
+      "loss": 0.9869,
+      "step": 46310
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6621885714285714e-06,
+      "loss": 0.9019,
+      "step": 46311
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.66192e-06,
+      "loss": 1.1519,
+      "step": 46312
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6616514285714286e-06,
+      "loss": 0.6547,
+      "step": 46313
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6613828571428572e-06,
+      "loss": 0.9334,
+      "step": 46314
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6611142857142858e-06,
+      "loss": 1.379,
+      "step": 46315
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6608457142857144e-06,
+      "loss": 0.9586,
+      "step": 46316
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6605771428571428e-06,
+      "loss": 1.0616,
+      "step": 46317
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6603085714285714e-06,
+      "loss": 0.9549,
+      "step": 46318
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.66004e-06,
+      "loss": 1.4043,
+      "step": 46319
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6597714285714286e-06,
+      "loss": 0.9756,
+      "step": 46320
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.659502857142857e-06,
+      "loss": 1.698,
+      "step": 46321
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6592342857142858e-06,
+      "loss": 1.7066,
+      "step": 46322
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6589657142857144e-06,
+      "loss": 0.8762,
+      "step": 46323
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6586971428571428e-06,
+      "loss": 0.8647,
+      "step": 46324
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6584285714285716e-06,
+      "loss": 0.9806,
+      "step": 46325
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.65816e-06,
+      "loss": 0.8148,
+      "step": 46326
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6578914285714286e-06,
+      "loss": 1.1324,
+      "step": 46327
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6576228571428572e-06,
+      "loss": 0.8679,
+      "step": 46328
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6573542857142858e-06,
+      "loss": 1.0133,
+      "step": 46329
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6570857142857144e-06,
+      "loss": 1.0596,
+      "step": 46330
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6568171428571428e-06,
+      "loss": 0.5438,
+      "step": 46331
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6565485714285716e-06,
+      "loss": 0.702,
+      "step": 46332
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.65628e-06,
+      "loss": 1.5001,
+      "step": 46333
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6560114285714286e-06,
+      "loss": 1.1339,
+      "step": 46334
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6557428571428572e-06,
+      "loss": 1.3085,
+      "step": 46335
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6554742857142858e-06,
+      "loss": 0.4897,
+      "step": 46336
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6552057142857142e-06,
+      "loss": 1.2151,
+      "step": 46337
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.654937142857143e-06,
+      "loss": 1.0565,
+      "step": 46338
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6546685714285716e-06,
+      "loss": 1.3183,
+      "step": 46339
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6544e-06,
+      "loss": 0.9644,
+      "step": 46340
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6541314285714286e-06,
+      "loss": 1.0726,
+      "step": 46341
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6538628571428572e-06,
+      "loss": 1.1604,
+      "step": 46342
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6535942857142858e-06,
+      "loss": 1.3652,
+      "step": 46343
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6533257142857142e-06,
+      "loss": 0.9692,
+      "step": 46344
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.653057142857143e-06,
+      "loss": 1.2585,
+      "step": 46345
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6527885714285716e-06,
+      "loss": 1.0437,
+      "step": 46346
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.65252e-06,
+      "loss": 1.3365,
+      "step": 46347
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6522514285714288e-06,
+      "loss": 1.3523,
+      "step": 46348
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6519828571428572e-06,
+      "loss": 0.9778,
+      "step": 46349
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6517142857142858e-06,
+      "loss": 1.5618,
+      "step": 46350
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6514457142857144e-06,
+      "loss": 0.9334,
+      "step": 46351
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.651177142857143e-06,
+      "loss": 1.0204,
+      "step": 46352
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6509085714285714e-06,
+      "loss": 1.1736,
+      "step": 46353
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.65064e-06,
+      "loss": 1.1414,
+      "step": 46354
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6503714285714288e-06,
+      "loss": 0.9319,
+      "step": 46355
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6501028571428572e-06,
+      "loss": 1.3558,
+      "step": 46356
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6498342857142858e-06,
+      "loss": 1.1509,
+      "step": 46357
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6495657142857144e-06,
+      "loss": 1.7097,
+      "step": 46358
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.649297142857143e-06,
+      "loss": 1.0283,
+      "step": 46359
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6490285714285713e-06,
+      "loss": 1.16,
+      "step": 46360
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6487600000000002e-06,
+      "loss": 1.0224,
+      "step": 46361
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6484914285714288e-06,
+      "loss": 1.0895,
+      "step": 46362
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6482228571428571e-06,
+      "loss": 0.922,
+      "step": 46363
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6479542857142857e-06,
+      "loss": 1.5421,
+      "step": 46364
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6476857142857143e-06,
+      "loss": 0.5069,
+      "step": 46365
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.647417142857143e-06,
+      "loss": 1.4546,
+      "step": 46366
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6471485714285713e-06,
+      "loss": 0.9534,
+      "step": 46367
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6468800000000001e-06,
+      "loss": 1.3608,
+      "step": 46368
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6466114285714285e-06,
+      "loss": 1.2109,
+      "step": 46369
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6463428571428571e-06,
+      "loss": 1.3026,
+      "step": 46370
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.646074285714286e-06,
+      "loss": 1.6457,
+      "step": 46371
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6458057142857143e-06,
+      "loss": 1.5094,
+      "step": 46372
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.645537142857143e-06,
+      "loss": 0.8228,
+      "step": 46373
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6452685714285715e-06,
+      "loss": 1.3651,
+      "step": 46374
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6450000000000001e-06,
+      "loss": 0.8076,
+      "step": 46375
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6447314285714285e-06,
+      "loss": 0.7844,
+      "step": 46376
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6444628571428571e-06,
+      "loss": 0.8287,
+      "step": 46377
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.644194285714286e-06,
+      "loss": 0.7774,
+      "step": 46378
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6439257142857143e-06,
+      "loss": 1.6257,
+      "step": 46379
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.643657142857143e-06,
+      "loss": 1.1028,
+      "step": 46380
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6433885714285715e-06,
+      "loss": 1.3052,
+      "step": 46381
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6431200000000001e-06,
+      "loss": 1.1778,
+      "step": 46382
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6428514285714285e-06,
+      "loss": 1.0542,
+      "step": 46383
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6425828571428573e-06,
+      "loss": 1.1209,
+      "step": 46384
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6423142857142857e-06,
+      "loss": 1.4425,
+      "step": 46385
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6420457142857143e-06,
+      "loss": 1.2661,
+      "step": 46386
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.641777142857143e-06,
+      "loss": 1.3646,
+      "step": 46387
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6415085714285715e-06,
+      "loss": 1.2075,
+      "step": 46388
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.64124e-06,
+      "loss": 1.0725,
+      "step": 46389
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6409714285714285e-06,
+      "loss": 1.1545,
+      "step": 46390
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6407028571428573e-06,
+      "loss": 1.0315,
+      "step": 46391
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6404342857142857e-06,
+      "loss": 0.98,
+      "step": 46392
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6401657142857143e-06,
+      "loss": 0.8687,
+      "step": 46393
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.639897142857143e-06,
+      "loss": 1.099,
+      "step": 46394
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6396285714285715e-06,
+      "loss": 1.3417,
+      "step": 46395
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.63936e-06,
+      "loss": 1.1942,
+      "step": 46396
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6390914285714285e-06,
+      "loss": 1.1864,
+      "step": 46397
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6388228571428573e-06,
+      "loss": 1.3885,
+      "step": 46398
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6385542857142857e-06,
+      "loss": 1.0804,
+      "step": 46399
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6382857142857143e-06,
+      "loss": 1.2148,
+      "step": 46400
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.638017142857143e-06,
+      "loss": 0.8013,
+      "step": 46401
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6377485714285715e-06,
+      "loss": 1.0733,
+      "step": 46402
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.63748e-06,
+      "loss": 1.2033,
+      "step": 46403
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6372114285714287e-06,
+      "loss": 1.7994,
+      "step": 46404
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6369428571428573e-06,
+      "loss": 1.0339,
+      "step": 46405
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6366742857142857e-06,
+      "loss": 0.873,
+      "step": 46406
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6364057142857145e-06,
+      "loss": 1.0188,
+      "step": 46407
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6361371428571429e-06,
+      "loss": 1.3577,
+      "step": 46408
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6358685714285715e-06,
+      "loss": 1.0061,
+      "step": 46409
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6356e-06,
+      "loss": 1.3052,
+      "step": 46410
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6353314285714287e-06,
+      "loss": 0.862,
+      "step": 46411
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6350628571428573e-06,
+      "loss": 0.6763,
+      "step": 46412
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6347942857142856e-06,
+      "loss": 1.2656,
+      "step": 46413
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6345257142857145e-06,
+      "loss": 1.3735,
+      "step": 46414
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6342571428571428e-06,
+      "loss": 0.7401,
+      "step": 46415
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6339885714285714e-06,
+      "loss": 0.9439,
+      "step": 46416
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6337200000000003e-06,
+      "loss": 1.0501,
+      "step": 46417
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6334514285714286e-06,
+      "loss": 1.5478,
+      "step": 46418
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6331828571428572e-06,
+      "loss": 0.8905,
+      "step": 46419
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6329142857142856e-06,
+      "loss": 1.4702,
+      "step": 46420
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6326457142857144e-06,
+      "loss": 1.0849,
+      "step": 46421
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6323771428571428e-06,
+      "loss": 0.7585,
+      "step": 46422
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6321085714285714e-06,
+      "loss": 0.9786,
+      "step": 46423
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.63184e-06,
+      "loss": 1.243,
+      "step": 46424
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6315714285714286e-06,
+      "loss": 1.5122,
+      "step": 46425
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6313028571428572e-06,
+      "loss": 0.7628,
+      "step": 46426
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6310342857142858e-06,
+      "loss": 1.3222,
+      "step": 46427
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6307657142857144e-06,
+      "loss": 0.9179,
+      "step": 46428
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6304971428571428e-06,
+      "loss": 1.216,
+      "step": 46429
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6302285714285714e-06,
+      "loss": 1.0401,
+      "step": 46430
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.62996e-06,
+      "loss": 1.0234,
+      "step": 46431
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6296914285714286e-06,
+      "loss": 1.4446,
+      "step": 46432
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6294228571428572e-06,
+      "loss": 1.3351,
+      "step": 46433
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6291542857142858e-06,
+      "loss": 0.665,
+      "step": 46434
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6288857142857144e-06,
+      "loss": 0.8578,
+      "step": 46435
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6286171428571428e-06,
+      "loss": 1.412,
+      "step": 46436
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6283485714285716e-06,
+      "loss": 0.9335,
+      "step": 46437
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.62808e-06,
+      "loss": 0.9141,
+      "step": 46438
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6278114285714286e-06,
+      "loss": 1.2422,
+      "step": 46439
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6275428571428572e-06,
+      "loss": 1.2937,
+      "step": 46440
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6272742857142858e-06,
+      "loss": 1.4475,
+      "step": 46441
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6270057142857144e-06,
+      "loss": 1.2588,
+      "step": 46442
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6267371428571428e-06,
+      "loss": 1.0406,
+      "step": 46443
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6264685714285716e-06,
+      "loss": 0.9705,
+      "step": 46444
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6262e-06,
+      "loss": 1.0154,
+      "step": 46445
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6259314285714286e-06,
+      "loss": 1.0784,
+      "step": 46446
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6256628571428572e-06,
+      "loss": 1.1601,
+      "step": 46447
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6253942857142858e-06,
+      "loss": 1.3876,
+      "step": 46448
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6251257142857144e-06,
+      "loss": 0.8231,
+      "step": 46449
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.624857142857143e-06,
+      "loss": 0.7518,
+      "step": 46450
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6245885714285716e-06,
+      "loss": 1.1732,
+      "step": 46451
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.62432e-06,
+      "loss": 1.3287,
+      "step": 46452
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6240514285714286e-06,
+      "loss": 0.946,
+      "step": 46453
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6237828571428572e-06,
+      "loss": 1.3507,
+      "step": 46454
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6235142857142858e-06,
+      "loss": 0.7263,
+      "step": 46455
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6232457142857142e-06,
+      "loss": 0.8861,
+      "step": 46456
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.622977142857143e-06,
+      "loss": 0.7364,
+      "step": 46457
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6227085714285716e-06,
+      "loss": 0.9909,
+      "step": 46458
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.62244e-06,
+      "loss": 1.7047,
+      "step": 46459
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6221714285714288e-06,
+      "loss": 1.0204,
+      "step": 46460
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6219028571428572e-06,
+      "loss": 1.0882,
+      "step": 46461
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.6216342857142858e-06,
+      "loss": 1.0684,
+      "step": 46462
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6213657142857144e-06,
+      "loss": 0.7726,
+      "step": 46463
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.621097142857143e-06,
+      "loss": 1.1206,
+      "step": 46464
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6208285714285716e-06,
+      "loss": 1.1278,
+      "step": 46465
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.62056e-06,
+      "loss": 1.1545,
+      "step": 46466
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6202914285714288e-06,
+      "loss": 1.0826,
+      "step": 46467
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6200228571428571e-06,
+      "loss": 1.0105,
+      "step": 46468
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6197542857142857e-06,
+      "loss": 0.829,
+      "step": 46469
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6194857142857143e-06,
+      "loss": 1.227,
+      "step": 46470
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.619217142857143e-06,
+      "loss": 0.9646,
+      "step": 46471
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6189485714285713e-06,
+      "loss": 0.814,
+      "step": 46472
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6186800000000001e-06,
+      "loss": 1.1658,
+      "step": 46473
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6184114285714287e-06,
+      "loss": 1.2406,
+      "step": 46474
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6181428571428571e-06,
+      "loss": 1.5822,
+      "step": 46475
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6178742857142857e-06,
+      "loss": 1.2767,
+      "step": 46476
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6176057142857143e-06,
+      "loss": 1.4647,
+      "step": 46477
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.617337142857143e-06,
+      "loss": 1.0938,
+      "step": 46478
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6170685714285713e-06,
+      "loss": 1.1738,
+      "step": 46479
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6168000000000001e-06,
+      "loss": 0.7959,
+      "step": 46480
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6165314285714287e-06,
+      "loss": 1.1376,
+      "step": 46481
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6162628571428571e-06,
+      "loss": 1.059,
+      "step": 46482
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.615994285714286e-06,
+      "loss": 1.5226,
+      "step": 46483
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6157257142857143e-06,
+      "loss": 1.1629,
+      "step": 46484
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.615457142857143e-06,
+      "loss": 1.2681,
+      "step": 46485
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6151885714285715e-06,
+      "loss": 1.4258,
+      "step": 46486
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6149200000000001e-06,
+      "loss": 1.1758,
+      "step": 46487
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6146514285714285e-06,
+      "loss": 0.8185,
+      "step": 46488
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.614382857142857e-06,
+      "loss": 1.066,
+      "step": 46489
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.614114285714286e-06,
+      "loss": 1.0571,
+      "step": 46490
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6138457142857143e-06,
+      "loss": 0.8951,
+      "step": 46491
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.613577142857143e-06,
+      "loss": 1.0477,
+      "step": 46492
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6133085714285715e-06,
+      "loss": 1.2187,
+      "step": 46493
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.61304e-06,
+      "loss": 1.0285,
+      "step": 46494
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6127714285714285e-06,
+      "loss": 1.3333,
+      "step": 46495
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6125028571428573e-06,
+      "loss": 0.91,
+      "step": 46496
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.612234285714286e-06,
+      "loss": 1.4401,
+      "step": 46497
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6119657142857143e-06,
+      "loss": 1.1739,
+      "step": 46498
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6116971428571429e-06,
+      "loss": 1.1743,
+      "step": 46499
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6114285714285715e-06,
+      "loss": 0.7234,
+      "step": 46500
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.61116e-06,
+      "loss": 1.2718,
+      "step": 46501
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6108914285714285e-06,
+      "loss": 0.8623,
+      "step": 46502
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6106228571428573e-06,
+      "loss": 1.3704,
+      "step": 46503
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6103542857142859e-06,
+      "loss": 0.7943,
+      "step": 46504
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6100857142857143e-06,
+      "loss": 1.7748,
+      "step": 46505
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.609817142857143e-06,
+      "loss": 1.1466,
+      "step": 46506
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6095485714285715e-06,
+      "loss": 0.97,
+      "step": 46507
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.60928e-06,
+      "loss": 1.4027,
+      "step": 46508
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6090114285714285e-06,
+      "loss": 0.945,
+      "step": 46509
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6087428571428573e-06,
+      "loss": 0.9138,
+      "step": 46510
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6084742857142857e-06,
+      "loss": 1.2524,
+      "step": 46511
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6082057142857143e-06,
+      "loss": 1.1882,
+      "step": 46512
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.607937142857143e-06,
+      "loss": 0.8747,
+      "step": 46513
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6076685714285715e-06,
+      "loss": 1.2573,
+      "step": 46514
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6074e-06,
+      "loss": 0.8908,
+      "step": 46515
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6071314285714287e-06,
+      "loss": 1.4335,
+      "step": 46516
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6068628571428573e-06,
+      "loss": 1.5929,
+      "step": 46517
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6065942857142856e-06,
+      "loss": 0.9772,
+      "step": 46518
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6063257142857145e-06,
+      "loss": 1.1392,
+      "step": 46519
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.606057142857143e-06,
+      "loss": 1.1692,
+      "step": 46520
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6057885714285714e-06,
+      "loss": 1.0873,
+      "step": 46521
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.60552e-06,
+      "loss": 1.2676,
+      "step": 46522
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6052514285714286e-06,
+      "loss": 1.3772,
+      "step": 46523
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6049828571428572e-06,
+      "loss": 0.9849,
+      "step": 46524
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6047142857142856e-06,
+      "loss": 1.0035,
+      "step": 46525
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6044457142857144e-06,
+      "loss": 0.6586,
+      "step": 46526
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6041771428571428e-06,
+      "loss": 0.4761,
+      "step": 46527
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6039085714285714e-06,
+      "loss": 1.1958,
+      "step": 46528
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6036400000000002e-06,
+      "loss": 0.6218,
+      "step": 46529
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6033714285714286e-06,
+      "loss": 1.016,
+      "step": 46530
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6031028571428572e-06,
+      "loss": 1.3763,
+      "step": 46531
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6028342857142856e-06,
+      "loss": 1.4911,
+      "step": 46532
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6025657142857144e-06,
+      "loss": 0.8484,
+      "step": 46533
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6022971428571428e-06,
+      "loss": 1.345,
+      "step": 46534
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6020285714285714e-06,
+      "loss": 0.9207,
+      "step": 46535
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6017600000000002e-06,
+      "loss": 1.0501,
+      "step": 46536
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6014914285714286e-06,
+      "loss": 1.2971,
+      "step": 46537
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6012228571428572e-06,
+      "loss": 1.3735,
+      "step": 46538
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6009542857142858e-06,
+      "loss": 1.3626,
+      "step": 46539
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6006857142857144e-06,
+      "loss": 1.5378,
+      "step": 46540
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6004171428571428e-06,
+      "loss": 1.1172,
+      "step": 46541
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.6001485714285714e-06,
+      "loss": 1.2364,
+      "step": 46542
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.59988e-06,
+      "loss": 1.4037,
+      "step": 46543
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5996114285714286e-06,
+      "loss": 1.1791,
+      "step": 46544
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5993428571428572e-06,
+      "loss": 1.1537,
+      "step": 46545
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5990742857142858e-06,
+      "loss": 1.3189,
+      "step": 46546
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5988057142857144e-06,
+      "loss": 0.8163,
+      "step": 46547
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5985371428571428e-06,
+      "loss": 1.5705,
+      "step": 46548
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5982685714285716e-06,
+      "loss": 1.0652,
+      "step": 46549
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.598e-06,
+      "loss": 1.0615,
+      "step": 46550
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5977314285714286e-06,
+      "loss": 1.2114,
+      "step": 46551
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5974628571428572e-06,
+      "loss": 0.9917,
+      "step": 46552
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5971942857142858e-06,
+      "loss": 0.9835,
+      "step": 46553
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5969257142857144e-06,
+      "loss": 1.1351,
+      "step": 46554
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5966571428571428e-06,
+      "loss": 1.0897,
+      "step": 46555
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5963885714285716e-06,
+      "loss": 0.7048,
+      "step": 46556
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.59612e-06,
+      "loss": 1.2744,
+      "step": 46557
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5958514285714286e-06,
+      "loss": 1.272,
+      "step": 46558
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5955828571428572e-06,
+      "loss": 1.3666,
+      "step": 46559
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5953142857142858e-06,
+      "loss": 0.899,
+      "step": 46560
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5950457142857144e-06,
+      "loss": 1.0196,
+      "step": 46561
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.594777142857143e-06,
+      "loss": 0.6029,
+      "step": 46562
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5945085714285716e-06,
+      "loss": 1.2895,
+      "step": 46563
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.59424e-06,
+      "loss": 0.9996,
+      "step": 46564
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5939714285714286e-06,
+      "loss": 1.2304,
+      "step": 46565
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5937028571428572e-06,
+      "loss": 1.1242,
+      "step": 46566
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5934342857142858e-06,
+      "loss": 1.0323,
+      "step": 46567
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5931657142857144e-06,
+      "loss": 1.1597,
+      "step": 46568
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.592897142857143e-06,
+      "loss": 0.9673,
+      "step": 46569
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5926285714285716e-06,
+      "loss": 1.093,
+      "step": 46570
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.59236e-06,
+      "loss": 1.4144,
+      "step": 46571
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5920914285714288e-06,
+      "loss": 1.0102,
+      "step": 46572
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5918228571428571e-06,
+      "loss": 1.0495,
+      "step": 46573
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5915542857142857e-06,
+      "loss": 1.1594,
+      "step": 46574
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5912857142857143e-06,
+      "loss": 0.8242,
+      "step": 46575
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.591017142857143e-06,
+      "loss": 1.2462,
+      "step": 46576
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5907485714285715e-06,
+      "loss": 1.1856,
+      "step": 46577
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.59048e-06,
+      "loss": 1.3366,
+      "step": 46578
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5902114285714287e-06,
+      "loss": 0.8077,
+      "step": 46579
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5899428571428571e-06,
+      "loss": 0.9698,
+      "step": 46580
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5896742857142857e-06,
+      "loss": 1.0366,
+      "step": 46581
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5894057142857143e-06,
+      "loss": 0.7784,
+      "step": 46582
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.589137142857143e-06,
+      "loss": 1.5034,
+      "step": 46583
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5888685714285715e-06,
+      "loss": 0.9901,
+      "step": 46584
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5886000000000001e-06,
+      "loss": 0.6788,
+      "step": 46585
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5883314285714287e-06,
+      "loss": 1.4187,
+      "step": 46586
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5880628571428571e-06,
+      "loss": 1.3319,
+      "step": 46587
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5877942857142857e-06,
+      "loss": 1.3357,
+      "step": 46588
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5875257142857143e-06,
+      "loss": 0.745,
+      "step": 46589
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.587257142857143e-06,
+      "loss": 1.3315,
+      "step": 46590
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5869885714285715e-06,
+      "loss": 0.8149,
+      "step": 46591
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5867200000000001e-06,
+      "loss": 1.2712,
+      "step": 46592
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5864514285714287e-06,
+      "loss": 1.1435,
+      "step": 46593
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.586182857142857e-06,
+      "loss": 0.876,
+      "step": 46594
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.585914285714286e-06,
+      "loss": 1.037,
+      "step": 46595
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5856457142857143e-06,
+      "loss": 0.658,
+      "step": 46596
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.585377142857143e-06,
+      "loss": 1.142,
+      "step": 46597
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5851085714285715e-06,
+      "loss": 1.2711,
+      "step": 46598
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.58484e-06,
+      "loss": 1.1294,
+      "step": 46599
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5845714285714287e-06,
+      "loss": 1.2913,
+      "step": 46600
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.584302857142857e-06,
+      "loss": 1.1462,
+      "step": 46601
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.584034285714286e-06,
+      "loss": 1.0739,
+      "step": 46602
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5837657142857143e-06,
+      "loss": 1.0308,
+      "step": 46603
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5834971428571429e-06,
+      "loss": 1.1199,
+      "step": 46604
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5832285714285715e-06,
+      "loss": 1.0824,
+      "step": 46605
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.58296e-06,
+      "loss": 1.0675,
+      "step": 46606
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5826914285714287e-06,
+      "loss": 1.0666,
+      "step": 46607
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5824228571428573e-06,
+      "loss": 0.8838,
+      "step": 46608
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5821542857142859e-06,
+      "loss": 1.2317,
+      "step": 46609
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5818857142857143e-06,
+      "loss": 0.8084,
+      "step": 46610
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5816171428571429e-06,
+      "loss": 1.0831,
+      "step": 46611
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5813485714285715e-06,
+      "loss": 1.2018,
+      "step": 46612
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.58108e-06,
+      "loss": 0.7826,
+      "step": 46613
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5808114285714285e-06,
+      "loss": 1.3049,
+      "step": 46614
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5805428571428573e-06,
+      "loss": 1.318,
+      "step": 46615
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5802742857142859e-06,
+      "loss": 0.7986,
+      "step": 46616
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5800057142857143e-06,
+      "loss": 1.4978,
+      "step": 46617
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.579737142857143e-06,
+      "loss": 0.9827,
+      "step": 46618
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5794685714285715e-06,
+      "loss": 0.8268,
+      "step": 46619
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5792e-06,
+      "loss": 0.9624,
+      "step": 46620
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5789314285714287e-06,
+      "loss": 0.8847,
+      "step": 46621
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5786628571428573e-06,
+      "loss": 1.027,
+      "step": 46622
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5783942857142859e-06,
+      "loss": 1.1291,
+      "step": 46623
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5781257142857142e-06,
+      "loss": 1.3005,
+      "step": 46624
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.577857142857143e-06,
+      "loss": 1.2652,
+      "step": 46625
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5775885714285714e-06,
+      "loss": 1.3037,
+      "step": 46626
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.57732e-06,
+      "loss": 0.7112,
+      "step": 46627
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5770514285714286e-06,
+      "loss": 1.2401,
+      "step": 46628
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5767828571428572e-06,
+      "loss": 0.9574,
+      "step": 46629
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5765142857142856e-06,
+      "loss": 0.9544,
+      "step": 46630
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5762457142857144e-06,
+      "loss": 1.0191,
+      "step": 46631
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.575977142857143e-06,
+      "loss": 1.2719,
+      "step": 46632
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5757085714285714e-06,
+      "loss": 0.564,
+      "step": 46633
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.57544e-06,
+      "loss": 1.3722,
+      "step": 46634
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5751714285714286e-06,
+      "loss": 1.2678,
+      "step": 46635
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5749028571428572e-06,
+      "loss": 1.4834,
+      "step": 46636
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.5746342857142856e-06,
+      "loss": 0.9299,
+      "step": 46637
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5743657142857144e-06,
+      "loss": 1.0981,
+      "step": 46638
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.574097142857143e-06,
+      "loss": 1.0723,
+      "step": 46639
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5738285714285714e-06,
+      "loss": 1.4506,
+      "step": 46640
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5735600000000002e-06,
+      "loss": 0.7321,
+      "step": 46641
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5732914285714286e-06,
+      "loss": 1.0373,
+      "step": 46642
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5730228571428572e-06,
+      "loss": 1.0786,
+      "step": 46643
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5727542857142856e-06,
+      "loss": 1.129,
+      "step": 46644
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5724857142857144e-06,
+      "loss": 1.1625,
+      "step": 46645
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5722171428571428e-06,
+      "loss": 1.1732,
+      "step": 46646
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5719485714285714e-06,
+      "loss": 1.0929,
+      "step": 46647
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5716800000000002e-06,
+      "loss": 1.5667,
+      "step": 46648
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5714114285714286e-06,
+      "loss": 1.1101,
+      "step": 46649
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5711428571428572e-06,
+      "loss": 0.9426,
+      "step": 46650
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5708742857142858e-06,
+      "loss": 1.0897,
+      "step": 46651
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5706057142857144e-06,
+      "loss": 0.8953,
+      "step": 46652
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5703371428571428e-06,
+      "loss": 1.4319,
+      "step": 46653
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5700685714285714e-06,
+      "loss": 1.2457,
+      "step": 46654
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5698000000000002e-06,
+      "loss": 1.0533,
+      "step": 46655
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5695314285714286e-06,
+      "loss": 0.606,
+      "step": 46656
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5692628571428572e-06,
+      "loss": 1.654,
+      "step": 46657
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5689942857142858e-06,
+      "loss": 1.1256,
+      "step": 46658
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5687257142857144e-06,
+      "loss": 1.2463,
+      "step": 46659
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5684571428571428e-06,
+      "loss": 1.2306,
+      "step": 46660
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5681885714285716e-06,
+      "loss": 0.9793,
+      "step": 46661
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.56792e-06,
+      "loss": 1.0328,
+      "step": 46662
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5676514285714286e-06,
+      "loss": 0.9729,
+      "step": 46663
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5673828571428572e-06,
+      "loss": 1.1819,
+      "step": 46664
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5671142857142858e-06,
+      "loss": 1.2441,
+      "step": 46665
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5668457142857144e-06,
+      "loss": 1.3246,
+      "step": 46666
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5665771428571428e-06,
+      "loss": 1.1372,
+      "step": 46667
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5663085714285716e-06,
+      "loss": 1.0095,
+      "step": 46668
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.56604e-06,
+      "loss": 1.3028,
+      "step": 46669
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5657714285714286e-06,
+      "loss": 1.1701,
+      "step": 46670
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5655028571428574e-06,
+      "loss": 1.2412,
+      "step": 46671
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5652342857142858e-06,
+      "loss": 1.0239,
+      "step": 46672
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5649657142857144e-06,
+      "loss": 0.8414,
+      "step": 46673
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.564697142857143e-06,
+      "loss": 0.8119,
+      "step": 46674
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5644285714285716e-06,
+      "loss": 1.3648,
+      "step": 46675
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.56416e-06,
+      "loss": 1.3535,
+      "step": 46676
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5638914285714285e-06,
+      "loss": 1.0947,
+      "step": 46677
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5636228571428571e-06,
+      "loss": 1.3625,
+      "step": 46678
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5633542857142857e-06,
+      "loss": 1.4791,
+      "step": 46679
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5630857142857143e-06,
+      "loss": 1.0475,
+      "step": 46680
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.562817142857143e-06,
+      "loss": 1.1484,
+      "step": 46681
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5625485714285715e-06,
+      "loss": 1.1428,
+      "step": 46682
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.56228e-06,
+      "loss": 1.0663,
+      "step": 46683
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5620114285714287e-06,
+      "loss": 1.5705,
+      "step": 46684
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5617428571428571e-06,
+      "loss": 1.1362,
+      "step": 46685
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5614742857142857e-06,
+      "loss": 0.8334,
+      "step": 46686
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5612057142857143e-06,
+      "loss": 0.8385,
+      "step": 46687
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.560937142857143e-06,
+      "loss": 1.083,
+      "step": 46688
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5606685714285715e-06,
+      "loss": 0.8046,
+      "step": 46689
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5604e-06,
+      "loss": 1.4029,
+      "step": 46690
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5601314285714287e-06,
+      "loss": 1.233,
+      "step": 46691
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5598628571428571e-06,
+      "loss": 1.123,
+      "step": 46692
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5595942857142857e-06,
+      "loss": 1.3544,
+      "step": 46693
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5593257142857145e-06,
+      "loss": 1.0086,
+      "step": 46694
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.559057142857143e-06,
+      "loss": 1.1216,
+      "step": 46695
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5587885714285715e-06,
+      "loss": 0.9629,
+      "step": 46696
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5585200000000001e-06,
+      "loss": 1.2844,
+      "step": 46697
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5582514285714287e-06,
+      "loss": 1.3514,
+      "step": 46698
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.557982857142857e-06,
+      "loss": 1.1223,
+      "step": 46699
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5577142857142857e-06,
+      "loss": 0.9206,
+      "step": 46700
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5574457142857143e-06,
+      "loss": 1.1138,
+      "step": 46701
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.557177142857143e-06,
+      "loss": 1.0674,
+      "step": 46702
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5569085714285715e-06,
+      "loss": 1.3081,
+      "step": 46703
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.55664e-06,
+      "loss": 0.5853,
+      "step": 46704
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5563714285714287e-06,
+      "loss": 0.945,
+      "step": 46705
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.556102857142857e-06,
+      "loss": 0.9955,
+      "step": 46706
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.555834285714286e-06,
+      "loss": 1.3028,
+      "step": 46707
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5555657142857143e-06,
+      "loss": 1.3235,
+      "step": 46708
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5552971428571429e-06,
+      "loss": 1.2734,
+      "step": 46709
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5550285714285715e-06,
+      "loss": 1.6246,
+      "step": 46710
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.55476e-06,
+      "loss": 1.4289,
+      "step": 46711
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5544914285714287e-06,
+      "loss": 0.657,
+      "step": 46712
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.554222857142857e-06,
+      "loss": 1.2127,
+      "step": 46713
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5539542857142859e-06,
+      "loss": 1.617,
+      "step": 46714
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5536857142857143e-06,
+      "loss": 1.112,
+      "step": 46715
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5534171428571429e-06,
+      "loss": 0.88,
+      "step": 46716
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5531485714285715e-06,
+      "loss": 1.0365,
+      "step": 46717
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.55288e-06,
+      "loss": 1.1699,
+      "step": 46718
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5526114285714287e-06,
+      "loss": 0.9004,
+      "step": 46719
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5523428571428573e-06,
+      "loss": 0.949,
+      "step": 46720
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5520742857142859e-06,
+      "loss": 1.1622,
+      "step": 46721
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5518057142857143e-06,
+      "loss": 1.1142,
+      "step": 46722
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5515371428571429e-06,
+      "loss": 1.1205,
+      "step": 46723
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5512685714285715e-06,
+      "loss": 1.2169,
+      "step": 46724
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.551e-06,
+      "loss": 1.3897,
+      "step": 46725
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5507314285714287e-06,
+      "loss": 1.1343,
+      "step": 46726
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5504628571428573e-06,
+      "loss": 1.3991,
+      "step": 46727
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5501942857142859e-06,
+      "loss": 1.3173,
+      "step": 46728
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5499257142857142e-06,
+      "loss": 1.6388,
+      "step": 46729
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.549657142857143e-06,
+      "loss": 0.9559,
+      "step": 46730
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5493885714285714e-06,
+      "loss": 1.1624,
+      "step": 46731
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.54912e-06,
+      "loss": 1.0319,
+      "step": 46732
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5488514285714286e-06,
+      "loss": 1.0152,
+      "step": 46733
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5485828571428572e-06,
+      "loss": 1.095,
+      "step": 46734
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5483142857142858e-06,
+      "loss": 1.5018,
+      "step": 46735
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5480457142857142e-06,
+      "loss": 1.3747,
+      "step": 46736
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.547777142857143e-06,
+      "loss": 0.9973,
+      "step": 46737
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5475085714285714e-06,
+      "loss": 0.8583,
+      "step": 46738
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.54724e-06,
+      "loss": 0.737,
+      "step": 46739
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5469714285714286e-06,
+      "loss": 0.6634,
+      "step": 46740
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5467028571428572e-06,
+      "loss": 1.0727,
+      "step": 46741
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5464342857142858e-06,
+      "loss": 1.2385,
+      "step": 46742
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5461657142857144e-06,
+      "loss": 1.35,
+      "step": 46743
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.545897142857143e-06,
+      "loss": 1.034,
+      "step": 46744
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5456285714285714e-06,
+      "loss": 0.8811,
+      "step": 46745
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.54536e-06,
+      "loss": 0.8376,
+      "step": 46746
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5450914285714286e-06,
+      "loss": 1.2792,
+      "step": 46747
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5448228571428572e-06,
+      "loss": 0.9975,
+      "step": 46748
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5445542857142856e-06,
+      "loss": 0.6395,
+      "step": 46749
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5442857142857144e-06,
+      "loss": 1.2023,
+      "step": 46750
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.544017142857143e-06,
+      "loss": 1.2382,
+      "step": 46751
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5437485714285714e-06,
+      "loss": 1.2675,
+      "step": 46752
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5434800000000002e-06,
+      "loss": 0.787,
+      "step": 46753
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5432114285714286e-06,
+      "loss": 0.904,
+      "step": 46754
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5429428571428572e-06,
+      "loss": 1.2982,
+      "step": 46755
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5426742857142856e-06,
+      "loss": 1.3364,
+      "step": 46756
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5424057142857144e-06,
+      "loss": 1.1275,
+      "step": 46757
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.542137142857143e-06,
+      "loss": 1.2142,
+      "step": 46758
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5418685714285714e-06,
+      "loss": 1.101,
+      "step": 46759
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5416000000000002e-06,
+      "loss": 0.93,
+      "step": 46760
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5413314285714286e-06,
+      "loss": 1.2864,
+      "step": 46761
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5410628571428572e-06,
+      "loss": 1.2637,
+      "step": 46762
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5407942857142858e-06,
+      "loss": 0.9666,
+      "step": 46763
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5405257142857144e-06,
+      "loss": 0.7555,
+      "step": 46764
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5402571428571428e-06,
+      "loss": 1.1826,
+      "step": 46765
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5399885714285714e-06,
+      "loss": 1.1222,
+      "step": 46766
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5397200000000002e-06,
+      "loss": 1.0873,
+      "step": 46767
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5394514285714286e-06,
+      "loss": 0.9193,
+      "step": 46768
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5391828571428572e-06,
+      "loss": 1.275,
+      "step": 46769
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5389142857142858e-06,
+      "loss": 1.2561,
+      "step": 46770
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5386457142857144e-06,
+      "loss": 0.9351,
+      "step": 46771
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5383771428571428e-06,
+      "loss": 1.5467,
+      "step": 46772
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5381085714285716e-06,
+      "loss": 0.9709,
+      "step": 46773
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5378400000000002e-06,
+      "loss": 1.2233,
+      "step": 46774
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5375714285714286e-06,
+      "loss": 0.7917,
+      "step": 46775
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5373028571428572e-06,
+      "loss": 1.4551,
+      "step": 46776
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5370342857142858e-06,
+      "loss": 0.9507,
+      "step": 46777
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5367657142857144e-06,
+      "loss": 0.95,
+      "step": 46778
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5364971428571427e-06,
+      "loss": 0.9754,
+      "step": 46779
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5362285714285716e-06,
+      "loss": 0.8891,
+      "step": 46780
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.53596e-06,
+      "loss": 0.9875,
+      "step": 46781
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5356914285714285e-06,
+      "loss": 0.6555,
+      "step": 46782
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5354228571428574e-06,
+      "loss": 1.2027,
+      "step": 46783
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5351542857142857e-06,
+      "loss": 1.0481,
+      "step": 46784
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5348857142857143e-06,
+      "loss": 1.1837,
+      "step": 46785
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.534617142857143e-06,
+      "loss": 1.4665,
+      "step": 46786
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5343485714285715e-06,
+      "loss": 0.6802,
+      "step": 46787
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.53408e-06,
+      "loss": 1.2303,
+      "step": 46788
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5338114285714285e-06,
+      "loss": 0.8696,
+      "step": 46789
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5335428571428573e-06,
+      "loss": 1.2226,
+      "step": 46790
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5332742857142857e-06,
+      "loss": 1.269,
+      "step": 46791
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5330057142857143e-06,
+      "loss": 0.8537,
+      "step": 46792
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.532737142857143e-06,
+      "loss": 1.0156,
+      "step": 46793
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5324685714285715e-06,
+      "loss": 1.039,
+      "step": 46794
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5322e-06,
+      "loss": 1.2608,
+      "step": 46795
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5319314285714287e-06,
+      "loss": 0.8985,
+      "step": 46796
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5316628571428573e-06,
+      "loss": 1.2243,
+      "step": 46797
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5313942857142857e-06,
+      "loss": 0.9428,
+      "step": 46798
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5311257142857143e-06,
+      "loss": 0.9862,
+      "step": 46799
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.530857142857143e-06,
+      "loss": 1.1234,
+      "step": 46800
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5305885714285715e-06,
+      "loss": 1.2007,
+      "step": 46801
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.53032e-06,
+      "loss": 1.167,
+      "step": 46802
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5300514285714287e-06,
+      "loss": 0.7234,
+      "step": 46803
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.529782857142857e-06,
+      "loss": 1.0138,
+      "step": 46804
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5295142857142857e-06,
+      "loss": 1.3803,
+      "step": 46805
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5292457142857145e-06,
+      "loss": 0.7799,
+      "step": 46806
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.528977142857143e-06,
+      "loss": 1.6836,
+      "step": 46807
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5287085714285715e-06,
+      "loss": 0.7847,
+      "step": 46808
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.52844e-06,
+      "loss": 1.0418,
+      "step": 46809
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5281714285714287e-06,
+      "loss": 1.0985,
+      "step": 46810
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.527902857142857e-06,
+      "loss": 1.2347,
+      "step": 46811
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.5276342857142857e-06,
+      "loss": 0.5984,
+      "step": 46812
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5273657142857145e-06,
+      "loss": 1.0386,
+      "step": 46813
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5270971428571429e-06,
+      "loss": 1.1686,
+      "step": 46814
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5268285714285715e-06,
+      "loss": 1.1449,
+      "step": 46815
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.52656e-06,
+      "loss": 1.3355,
+      "step": 46816
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5262914285714287e-06,
+      "loss": 0.9233,
+      "step": 46817
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.526022857142857e-06,
+      "loss": 1.0127,
+      "step": 46818
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5257542857142859e-06,
+      "loss": 1.2644,
+      "step": 46819
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5254857142857143e-06,
+      "loss": 1.3633,
+      "step": 46820
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5252171428571429e-06,
+      "loss": 1.4603,
+      "step": 46821
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5249485714285715e-06,
+      "loss": 1.5605,
+      "step": 46822
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.52468e-06,
+      "loss": 1.268,
+      "step": 46823
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5244114285714287e-06,
+      "loss": 1.0552,
+      "step": 46824
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.524142857142857e-06,
+      "loss": 0.7281,
+      "step": 46825
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5238742857142859e-06,
+      "loss": 1.2239,
+      "step": 46826
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5236057142857143e-06,
+      "loss": 0.5871,
+      "step": 46827
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5233371428571429e-06,
+      "loss": 1.0934,
+      "step": 46828
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5230685714285717e-06,
+      "loss": 0.8363,
+      "step": 46829
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5228e-06,
+      "loss": 0.9515,
+      "step": 46830
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5225314285714287e-06,
+      "loss": 0.7859,
+      "step": 46831
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5222628571428573e-06,
+      "loss": 1.3191,
+      "step": 46832
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5219942857142859e-06,
+      "loss": 1.3141,
+      "step": 46833
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5217257142857142e-06,
+      "loss": 0.902,
+      "step": 46834
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5214571428571428e-06,
+      "loss": 1.1886,
+      "step": 46835
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5211885714285714e-06,
+      "loss": 0.7614,
+      "step": 46836
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.52092e-06,
+      "loss": 0.9603,
+      "step": 46837
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5206514285714286e-06,
+      "loss": 1.0438,
+      "step": 46838
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5203828571428572e-06,
+      "loss": 0.734,
+      "step": 46839
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5201142857142858e-06,
+      "loss": 1.0548,
+      "step": 46840
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5198457142857142e-06,
+      "loss": 1.0858,
+      "step": 46841
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.519577142857143e-06,
+      "loss": 1.1351,
+      "step": 46842
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5193085714285714e-06,
+      "loss": 1.1006,
+      "step": 46843
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.51904e-06,
+      "loss": 0.927,
+      "step": 46844
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5187714285714286e-06,
+      "loss": 1.2836,
+      "step": 46845
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5185028571428572e-06,
+      "loss": 1.0113,
+      "step": 46846
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5182342857142858e-06,
+      "loss": 1.182,
+      "step": 46847
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5179657142857142e-06,
+      "loss": 1.541,
+      "step": 46848
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.517697142857143e-06,
+      "loss": 0.8324,
+      "step": 46849
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5174285714285714e-06,
+      "loss": 0.9109,
+      "step": 46850
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.51716e-06,
+      "loss": 1.5854,
+      "step": 46851
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5168914285714286e-06,
+      "loss": 1.0761,
+      "step": 46852
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5166228571428572e-06,
+      "loss": 1.0635,
+      "step": 46853
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5163542857142858e-06,
+      "loss": 1.4271,
+      "step": 46854
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5160857142857144e-06,
+      "loss": 1.1974,
+      "step": 46855
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.515817142857143e-06,
+      "loss": 1.092,
+      "step": 46856
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5155485714285714e-06,
+      "loss": 0.9377,
+      "step": 46857
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.51528e-06,
+      "loss": 0.9774,
+      "step": 46858
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5150114285714286e-06,
+      "loss": 1.239,
+      "step": 46859
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5147428571428572e-06,
+      "loss": 0.9976,
+      "step": 46860
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5144742857142858e-06,
+      "loss": 1.4013,
+      "step": 46861
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5142057142857144e-06,
+      "loss": 0.9987,
+      "step": 46862
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.513937142857143e-06,
+      "loss": 1.3321,
+      "step": 46863
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5136685714285714e-06,
+      "loss": 0.8171,
+      "step": 46864
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5134000000000002e-06,
+      "loss": 1.0945,
+      "step": 46865
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5131314285714286e-06,
+      "loss": 1.1967,
+      "step": 46866
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5128628571428572e-06,
+      "loss": 1.0618,
+      "step": 46867
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5125942857142856e-06,
+      "loss": 0.7847,
+      "step": 46868
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5123257142857144e-06,
+      "loss": 1.1013,
+      "step": 46869
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.512057142857143e-06,
+      "loss": 1.3875,
+      "step": 46870
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5117885714285714e-06,
+      "loss": 1.2404,
+      "step": 46871
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5115200000000002e-06,
+      "loss": 0.9289,
+      "step": 46872
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5112514285714286e-06,
+      "loss": 0.8502,
+      "step": 46873
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5109828571428572e-06,
+      "loss": 0.4679,
+      "step": 46874
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5107142857142858e-06,
+      "loss": 1.1166,
+      "step": 46875
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5104457142857144e-06,
+      "loss": 1.2352,
+      "step": 46876
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.510177142857143e-06,
+      "loss": 0.9101,
+      "step": 46877
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5099085714285714e-06,
+      "loss": 1.3742,
+      "step": 46878
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5096400000000002e-06,
+      "loss": 0.9267,
+      "step": 46879
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5093714285714286e-06,
+      "loss": 1.329,
+      "step": 46880
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5091028571428572e-06,
+      "loss": 0.832,
+      "step": 46881
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5088342857142858e-06,
+      "loss": 0.9151,
+      "step": 46882
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5085657142857144e-06,
+      "loss": 1.3331,
+      "step": 46883
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5082971428571427e-06,
+      "loss": 1.0863,
+      "step": 46884
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5080285714285716e-06,
+      "loss": 0.9986,
+      "step": 46885
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5077600000000002e-06,
+      "loss": 0.7803,
+      "step": 46886
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5074914285714285e-06,
+      "loss": 1.1844,
+      "step": 46887
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5072228571428571e-06,
+      "loss": 0.6401,
+      "step": 46888
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5069542857142857e-06,
+      "loss": 1.0321,
+      "step": 46889
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5066857142857143e-06,
+      "loss": 0.7809,
+      "step": 46890
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5064171428571427e-06,
+      "loss": 1.1766,
+      "step": 46891
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5061485714285715e-06,
+      "loss": 1.3864,
+      "step": 46892
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5058800000000001e-06,
+      "loss": 0.9051,
+      "step": 46893
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5056114285714285e-06,
+      "loss": 1.0642,
+      "step": 46894
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5053428571428573e-06,
+      "loss": 1.3542,
+      "step": 46895
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5050742857142857e-06,
+      "loss": 1.3676,
+      "step": 46896
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5048057142857143e-06,
+      "loss": 1.1493,
+      "step": 46897
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.504537142857143e-06,
+      "loss": 1.3947,
+      "step": 46898
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5042685714285715e-06,
+      "loss": 0.9936,
+      "step": 46899
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5040000000000001e-06,
+      "loss": 1.3648,
+      "step": 46900
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5037314285714285e-06,
+      "loss": 1.1343,
+      "step": 46901
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5034628571428573e-06,
+      "loss": 1.0913,
+      "step": 46902
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5031942857142857e-06,
+      "loss": 0.8311,
+      "step": 46903
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5029257142857143e-06,
+      "loss": 0.9861,
+      "step": 46904
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.502657142857143e-06,
+      "loss": 0.9022,
+      "step": 46905
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5023885714285715e-06,
+      "loss": 1.133,
+      "step": 46906
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.50212e-06,
+      "loss": 0.8344,
+      "step": 46907
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5018514285714287e-06,
+      "loss": 1.1152,
+      "step": 46908
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5015828571428573e-06,
+      "loss": 0.901,
+      "step": 46909
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5013142857142857e-06,
+      "loss": 0.8575,
+      "step": 46910
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5010457142857143e-06,
+      "loss": 1.3945,
+      "step": 46911
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.500777142857143e-06,
+      "loss": 0.8764,
+      "step": 46912
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5005085714285715e-06,
+      "loss": 1.0995,
+      "step": 46913
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.5002399999999999e-06,
+      "loss": 0.885,
+      "step": 46914
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4999714285714287e-06,
+      "loss": 1.0332,
+      "step": 46915
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4997028571428573e-06,
+      "loss": 0.8649,
+      "step": 46916
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4994342857142857e-06,
+      "loss": 0.9308,
+      "step": 46917
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4991657142857145e-06,
+      "loss": 1.081,
+      "step": 46918
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4988971428571429e-06,
+      "loss": 1.0832,
+      "step": 46919
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4986285714285715e-06,
+      "loss": 1.3451,
+      "step": 46920
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.49836e-06,
+      "loss": 0.7043,
+      "step": 46921
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4980914285714287e-06,
+      "loss": 1.5653,
+      "step": 46922
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.497822857142857e-06,
+      "loss": 1.1461,
+      "step": 46923
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4975542857142857e-06,
+      "loss": 1.347,
+      "step": 46924
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4972857142857145e-06,
+      "loss": 1.4831,
+      "step": 46925
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4970171428571429e-06,
+      "loss": 1.2974,
+      "step": 46926
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4967485714285715e-06,
+      "loss": 1.2368,
+      "step": 46927
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.49648e-06,
+      "loss": 0.9778,
+      "step": 46928
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4962114285714287e-06,
+      "loss": 1.0152,
+      "step": 46929
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.495942857142857e-06,
+      "loss": 1.0616,
+      "step": 46930
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4956742857142859e-06,
+      "loss": 1.2852,
+      "step": 46931
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4954057142857145e-06,
+      "loss": 1.2012,
+      "step": 46932
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4951371428571429e-06,
+      "loss": 1.6754,
+      "step": 46933
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4948685714285715e-06,
+      "loss": 1.2156,
+      "step": 46934
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4946e-06,
+      "loss": 1.0826,
+      "step": 46935
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4943314285714287e-06,
+      "loss": 0.8215,
+      "step": 46936
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.494062857142857e-06,
+      "loss": 0.7727,
+      "step": 46937
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4937942857142859e-06,
+      "loss": 1.0011,
+      "step": 46938
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4935257142857142e-06,
+      "loss": 0.7611,
+      "step": 46939
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4932571428571428e-06,
+      "loss": 0.8717,
+      "step": 46940
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4929885714285717e-06,
+      "loss": 1.4333,
+      "step": 46941
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.49272e-06,
+      "loss": 1.24,
+      "step": 46942
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4924514285714286e-06,
+      "loss": 1.0196,
+      "step": 46943
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4921828571428572e-06,
+      "loss": 1.3216,
+      "step": 46944
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4919142857142858e-06,
+      "loss": 1.3313,
+      "step": 46945
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4916457142857142e-06,
+      "loss": 1.4921,
+      "step": 46946
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4913771428571428e-06,
+      "loss": 0.4716,
+      "step": 46947
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4911085714285716e-06,
+      "loss": 1.4765,
+      "step": 46948
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.49084e-06,
+      "loss": 0.9769,
+      "step": 46949
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4905714285714286e-06,
+      "loss": 1.1247,
+      "step": 46950
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4903028571428572e-06,
+      "loss": 0.9873,
+      "step": 46951
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4900342857142858e-06,
+      "loss": 0.708,
+      "step": 46952
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4897657142857142e-06,
+      "loss": 1.4007,
+      "step": 46953
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.489497142857143e-06,
+      "loss": 1.2221,
+      "step": 46954
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4892285714285714e-06,
+      "loss": 0.866,
+      "step": 46955
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.48896e-06,
+      "loss": 1.5013,
+      "step": 46956
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4886914285714286e-06,
+      "loss": 1.0812,
+      "step": 46957
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4884228571428572e-06,
+      "loss": 1.2735,
+      "step": 46958
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4881542857142858e-06,
+      "loss": 1.2065,
+      "step": 46959
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4878857142857142e-06,
+      "loss": 0.8335,
+      "step": 46960
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.487617142857143e-06,
+      "loss": 0.7453,
+      "step": 46961
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4873485714285714e-06,
+      "loss": 1.0569,
+      "step": 46962
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.48708e-06,
+      "loss": 0.8423,
+      "step": 46963
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4868114285714288e-06,
+      "loss": 0.9175,
+      "step": 46964
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4865428571428572e-06,
+      "loss": 0.8711,
+      "step": 46965
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4862742857142858e-06,
+      "loss": 1.4209,
+      "step": 46966
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4860057142857144e-06,
+      "loss": 1.4753,
+      "step": 46967
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.485737142857143e-06,
+      "loss": 1.3465,
+      "step": 46968
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4854685714285714e-06,
+      "loss": 0.8605,
+      "step": 46969
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4852e-06,
+      "loss": 0.7476,
+      "step": 46970
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4849314285714286e-06,
+      "loss": 1.0955,
+      "step": 46971
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4846628571428572e-06,
+      "loss": 1.2012,
+      "step": 46972
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4843942857142858e-06,
+      "loss": 1.1484,
+      "step": 46973
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4841257142857144e-06,
+      "loss": 0.8248,
+      "step": 46974
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.483857142857143e-06,
+      "loss": 0.873,
+      "step": 46975
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4835885714285714e-06,
+      "loss": 0.9839,
+      "step": 46976
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4833200000000002e-06,
+      "loss": 1.1981,
+      "step": 46977
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4830514285714286e-06,
+      "loss": 1.0315,
+      "step": 46978
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4827828571428572e-06,
+      "loss": 1.1949,
+      "step": 46979
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4825142857142858e-06,
+      "loss": 1.0407,
+      "step": 46980
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4822457142857144e-06,
+      "loss": 1.0752,
+      "step": 46981
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.481977142857143e-06,
+      "loss": 1.3153,
+      "step": 46982
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4817085714285714e-06,
+      "loss": 1.0132,
+      "step": 46983
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4814400000000002e-06,
+      "loss": 1.1157,
+      "step": 46984
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4811714285714286e-06,
+      "loss": 1.5941,
+      "step": 46985
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4809028571428572e-06,
+      "loss": 0.9882,
+      "step": 46986
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4806342857142858e-06,
+      "loss": 1.0093,
+      "step": 46987
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4803657142857144e-06,
+      "loss": 0.9107,
+      "step": 46988
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.480097142857143e-06,
+      "loss": 0.8589,
+      "step": 46989
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4798285714285713e-06,
+      "loss": 1.4369,
+      "step": 46990
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4795600000000002e-06,
+      "loss": 1.2717,
+      "step": 46991
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4792914285714285e-06,
+      "loss": 0.9335,
+      "step": 46992
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4790228571428571e-06,
+      "loss": 1.1543,
+      "step": 46993
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4787542857142857e-06,
+      "loss": 1.2085,
+      "step": 46994
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4784857142857143e-06,
+      "loss": 0.9314,
+      "step": 46995
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.478217142857143e-06,
+      "loss": 1.0918,
+      "step": 46996
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4779485714285715e-06,
+      "loss": 1.189,
+      "step": 46997
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4776800000000001e-06,
+      "loss": 1.2158,
+      "step": 46998
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4774114285714285e-06,
+      "loss": 1.3004,
+      "step": 46999
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4771428571428573e-06,
+      "loss": 1.4372,
+      "step": 47000
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4768742857142857e-06,
+      "loss": 1.5127,
+      "step": 47001
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4766057142857143e-06,
+      "loss": 1.1647,
+      "step": 47002
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.476337142857143e-06,
+      "loss": 0.7644,
+      "step": 47003
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4760685714285715e-06,
+      "loss": 1.4453,
+      "step": 47004
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4758000000000001e-06,
+      "loss": 0.9078,
+      "step": 47005
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4755314285714285e-06,
+      "loss": 1.1599,
+      "step": 47006
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4752628571428573e-06,
+      "loss": 1.8376,
+      "step": 47007
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4749942857142857e-06,
+      "loss": 1.0852,
+      "step": 47008
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4747257142857143e-06,
+      "loss": 1.5865,
+      "step": 47009
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.474457142857143e-06,
+      "loss": 1.3245,
+      "step": 47010
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4741885714285715e-06,
+      "loss": 0.8241,
+      "step": 47011
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4739200000000001e-06,
+      "loss": 1.123,
+      "step": 47012
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4736514285714285e-06,
+      "loss": 0.9875,
+      "step": 47013
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4733828571428573e-06,
+      "loss": 1.0714,
+      "step": 47014
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4731142857142857e-06,
+      "loss": 1.0089,
+      "step": 47015
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4728457142857143e-06,
+      "loss": 0.9181,
+      "step": 47016
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.472577142857143e-06,
+      "loss": 1.2522,
+      "step": 47017
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4723085714285715e-06,
+      "loss": 0.8188,
+      "step": 47018
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.47204e-06,
+      "loss": 0.6799,
+      "step": 47019
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4717714285714287e-06,
+      "loss": 0.6483,
+      "step": 47020
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4715028571428573e-06,
+      "loss": 0.6462,
+      "step": 47021
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4712342857142857e-06,
+      "loss": 1.2004,
+      "step": 47022
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4709657142857143e-06,
+      "loss": 1.2024,
+      "step": 47023
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4706971428571429e-06,
+      "loss": 1.2474,
+      "step": 47024
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4704285714285715e-06,
+      "loss": 1.4014,
+      "step": 47025
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4701599999999999e-06,
+      "loss": 1.368,
+      "step": 47026
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4698914285714287e-06,
+      "loss": 1.2109,
+      "step": 47027
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4696228571428573e-06,
+      "loss": 1.0214,
+      "step": 47028
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4693542857142857e-06,
+      "loss": 1.2602,
+      "step": 47029
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4690857142857145e-06,
+      "loss": 1.534,
+      "step": 47030
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4688171428571429e-06,
+      "loss": 0.9865,
+      "step": 47031
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4685485714285715e-06,
+      "loss": 1.1929,
+      "step": 47032
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.46828e-06,
+      "loss": 1.1204,
+      "step": 47033
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4680114285714287e-06,
+      "loss": 0.753,
+      "step": 47034
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4677428571428573e-06,
+      "loss": 1.4811,
+      "step": 47035
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4674742857142857e-06,
+      "loss": 1.3148,
+      "step": 47036
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4672057142857145e-06,
+      "loss": 1.1822,
+      "step": 47037
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4669371428571429e-06,
+      "loss": 1.1717,
+      "step": 47038
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4666685714285715e-06,
+      "loss": 1.3846,
+      "step": 47039
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4664e-06,
+      "loss": 1.077,
+      "step": 47040
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4661314285714287e-06,
+      "loss": 1.1022,
+      "step": 47041
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.465862857142857e-06,
+      "loss": 1.4362,
+      "step": 47042
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4655942857142859e-06,
+      "loss": 1.2844,
+      "step": 47043
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4653257142857145e-06,
+      "loss": 0.9059,
+      "step": 47044
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4650571428571428e-06,
+      "loss": 1.3913,
+      "step": 47045
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4647885714285714e-06,
+      "loss": 1.0214,
+      "step": 47046
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.46452e-06,
+      "loss": 1.0273,
+      "step": 47047
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4642514285714286e-06,
+      "loss": 0.9404,
+      "step": 47048
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.463982857142857e-06,
+      "loss": 1.23,
+      "step": 47049
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4637142857142858e-06,
+      "loss": 0.7778,
+      "step": 47050
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4634457142857144e-06,
+      "loss": 1.0601,
+      "step": 47051
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4631771428571428e-06,
+      "loss": 1.0401,
+      "step": 47052
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4629085714285716e-06,
+      "loss": 1.1596,
+      "step": 47053
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.46264e-06,
+      "loss": 1.4417,
+      "step": 47054
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4623714285714286e-06,
+      "loss": 0.8961,
+      "step": 47055
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4621028571428572e-06,
+      "loss": 1.2615,
+      "step": 47056
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4618342857142858e-06,
+      "loss": 1.415,
+      "step": 47057
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4615657142857142e-06,
+      "loss": 1.0196,
+      "step": 47058
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4612971428571428e-06,
+      "loss": 0.9953,
+      "step": 47059
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4610285714285716e-06,
+      "loss": 1.2244,
+      "step": 47060
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.46076e-06,
+      "loss": 1.2943,
+      "step": 47061
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4604914285714286e-06,
+      "loss": 1.0069,
+      "step": 47062
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4602228571428572e-06,
+      "loss": 1.0059,
+      "step": 47063
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4599542857142858e-06,
+      "loss": 1.0525,
+      "step": 47064
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4596857142857142e-06,
+      "loss": 0.6286,
+      "step": 47065
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.459417142857143e-06,
+      "loss": 1.475,
+      "step": 47066
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4591485714285716e-06,
+      "loss": 1.4564,
+      "step": 47067
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.45888e-06,
+      "loss": 0.9494,
+      "step": 47068
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4586114285714286e-06,
+      "loss": 0.6895,
+      "step": 47069
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4583428571428572e-06,
+      "loss": 0.9258,
+      "step": 47070
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4580742857142858e-06,
+      "loss": 1.2519,
+      "step": 47071
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4578057142857142e-06,
+      "loss": 1.2613,
+      "step": 47072
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.457537142857143e-06,
+      "loss": 0.6153,
+      "step": 47073
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4572685714285714e-06,
+      "loss": 1.2908,
+      "step": 47074
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.457e-06,
+      "loss": 1.3623,
+      "step": 47075
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4567314285714288e-06,
+      "loss": 1.4147,
+      "step": 47076
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4564628571428572e-06,
+      "loss": 0.7174,
+      "step": 47077
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4561942857142858e-06,
+      "loss": 1.2483,
+      "step": 47078
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4559257142857144e-06,
+      "loss": 0.9286,
+      "step": 47079
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.455657142857143e-06,
+      "loss": 1.1633,
+      "step": 47080
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4553885714285714e-06,
+      "loss": 1.1884,
+      "step": 47081
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.45512e-06,
+      "loss": 1.164,
+      "step": 47082
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4548514285714288e-06,
+      "loss": 1.2443,
+      "step": 47083
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4545828571428572e-06,
+      "loss": 1.5122,
+      "step": 47084
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4543142857142858e-06,
+      "loss": 1.1211,
+      "step": 47085
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4540457142857144e-06,
+      "loss": 1.3139,
+      "step": 47086
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.453777142857143e-06,
+      "loss": 1.1788,
+      "step": 47087
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4535085714285714e-06,
+      "loss": 1.4333,
+      "step": 47088
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4532400000000002e-06,
+      "loss": 1.169,
+      "step": 47089
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4529714285714288e-06,
+      "loss": 1.4094,
+      "step": 47090
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4527028571428572e-06,
+      "loss": 0.8449,
+      "step": 47091
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4524342857142858e-06,
+      "loss": 1.1104,
+      "step": 47092
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4521657142857144e-06,
+      "loss": 0.979,
+      "step": 47093
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.451897142857143e-06,
+      "loss": 0.8082,
+      "step": 47094
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4516285714285713e-06,
+      "loss": 1.0558,
+      "step": 47095
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4513600000000002e-06,
+      "loss": 1.0227,
+      "step": 47096
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4510914285714285e-06,
+      "loss": 0.7373,
+      "step": 47097
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4508228571428571e-06,
+      "loss": 1.1329,
+      "step": 47098
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.450554285714286e-06,
+      "loss": 0.9875,
+      "step": 47099
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4502857142857143e-06,
+      "loss": 1.2978,
+      "step": 47100
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.450017142857143e-06,
+      "loss": 1.2005,
+      "step": 47101
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4497485714285713e-06,
+      "loss": 1.2876,
+      "step": 47102
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4494800000000001e-06,
+      "loss": 1.3933,
+      "step": 47103
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4492114285714285e-06,
+      "loss": 1.1183,
+      "step": 47104
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4489428571428571e-06,
+      "loss": 0.9737,
+      "step": 47105
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.448674285714286e-06,
+      "loss": 1.4486,
+      "step": 47106
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4484057142857143e-06,
+      "loss": 1.2136,
+      "step": 47107
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.448137142857143e-06,
+      "loss": 1.1077,
+      "step": 47108
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4478685714285715e-06,
+      "loss": 0.7393,
+      "step": 47109
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4476000000000001e-06,
+      "loss": 1.1706,
+      "step": 47110
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4473314285714285e-06,
+      "loss": 1.2227,
+      "step": 47111
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4470628571428573e-06,
+      "loss": 1.1247,
+      "step": 47112
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4467942857142857e-06,
+      "loss": 1.0523,
+      "step": 47113
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4465257142857143e-06,
+      "loss": 1.3726,
+      "step": 47114
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.446257142857143e-06,
+      "loss": 1.3449,
+      "step": 47115
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4459885714285715e-06,
+      "loss": 0.9835,
+      "step": 47116
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4457200000000001e-06,
+      "loss": 1.2217,
+      "step": 47117
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4454514285714285e-06,
+      "loss": 0.8159,
+      "step": 47118
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4451828571428573e-06,
+      "loss": 0.8953,
+      "step": 47119
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4449142857142857e-06,
+      "loss": 1.1085,
+      "step": 47120
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4446457142857143e-06,
+      "loss": 1.3902,
+      "step": 47121
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4443771428571431e-06,
+      "loss": 1.2937,
+      "step": 47122
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4441085714285715e-06,
+      "loss": 0.9827,
+      "step": 47123
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.44384e-06,
+      "loss": 1.3037,
+      "step": 47124
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4435714285714285e-06,
+      "loss": 1.2034,
+      "step": 47125
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4433028571428573e-06,
+      "loss": 1.6632,
+      "step": 47126
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4430342857142857e-06,
+      "loss": 1.0027,
+      "step": 47127
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4427657142857143e-06,
+      "loss": 1.393,
+      "step": 47128
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4424971428571429e-06,
+      "loss": 1.0235,
+      "step": 47129
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4422285714285715e-06,
+      "loss": 1.4277,
+      "step": 47130
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.44196e-06,
+      "loss": 1.3668,
+      "step": 47131
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4416914285714287e-06,
+      "loss": 1.1896,
+      "step": 47132
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4414228571428573e-06,
+      "loss": 1.0642,
+      "step": 47133
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4411542857142857e-06,
+      "loss": 1.2951,
+      "step": 47134
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4408857142857143e-06,
+      "loss": 0.8864,
+      "step": 47135
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4406171428571429e-06,
+      "loss": 1.2505,
+      "step": 47136
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4403485714285715e-06,
+      "loss": 1.1161,
+      "step": 47137
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.44008e-06,
+      "loss": 0.8554,
+      "step": 47138
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4398114285714287e-06,
+      "loss": 1.416,
+      "step": 47139
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4395428571428573e-06,
+      "loss": 1.1187,
+      "step": 47140
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4392742857142857e-06,
+      "loss": 0.9641,
+      "step": 47141
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4390057142857145e-06,
+      "loss": 1.2262,
+      "step": 47142
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4387371428571429e-06,
+      "loss": 0.8263,
+      "step": 47143
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4384685714285715e-06,
+      "loss": 1.2841,
+      "step": 47144
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4382e-06,
+      "loss": 0.8899,
+      "step": 47145
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4379314285714287e-06,
+      "loss": 0.8545,
+      "step": 47146
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4376628571428573e-06,
+      "loss": 0.8005,
+      "step": 47147
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4373942857142856e-06,
+      "loss": 1.1455,
+      "step": 47148
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4371257142857145e-06,
+      "loss": 1.1725,
+      "step": 47149
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4368571428571428e-06,
+      "loss": 1.3995,
+      "step": 47150
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4365885714285714e-06,
+      "loss": 1.1584,
+      "step": 47151
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.43632e-06,
+      "loss": 1.0838,
+      "step": 47152
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4360514285714286e-06,
+      "loss": 1.2278,
+      "step": 47153
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4357828571428572e-06,
+      "loss": 1.0212,
+      "step": 47154
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4355142857142858e-06,
+      "loss": 1.3128,
+      "step": 47155
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4352457142857144e-06,
+      "loss": 0.8916,
+      "step": 47156
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4349771428571428e-06,
+      "loss": 1.2989,
+      "step": 47157
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4347085714285714e-06,
+      "loss": 1.0232,
+      "step": 47158
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.43444e-06,
+      "loss": 0.9658,
+      "step": 47159
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4341714285714286e-06,
+      "loss": 1.3761,
+      "step": 47160
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.433902857142857e-06,
+      "loss": 1.1301,
+      "step": 47161
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.4336342857142858e-06,
+      "loss": 1.451,
+      "step": 47162
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4333657142857144e-06,
+      "loss": 0.7898,
+      "step": 47163
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4330971428571428e-06,
+      "loss": 1.5098,
+      "step": 47164
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4328285714285716e-06,
+      "loss": 0.9193,
+      "step": 47165
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.43256e-06,
+      "loss": 0.8256,
+      "step": 47166
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4322914285714286e-06,
+      "loss": 1.3001,
+      "step": 47167
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4320228571428572e-06,
+      "loss": 1.1633,
+      "step": 47168
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4317542857142858e-06,
+      "loss": 1.1531,
+      "step": 47169
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4314857142857144e-06,
+      "loss": 1.3412,
+      "step": 47170
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4312171428571428e-06,
+      "loss": 1.4532,
+      "step": 47171
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4309485714285716e-06,
+      "loss": 0.8335,
+      "step": 47172
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.43068e-06,
+      "loss": 1.1271,
+      "step": 47173
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4304114285714286e-06,
+      "loss": 1.0786,
+      "step": 47174
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4301428571428572e-06,
+      "loss": 1.0912,
+      "step": 47175
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4298742857142858e-06,
+      "loss": 1.1705,
+      "step": 47176
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4296057142857142e-06,
+      "loss": 1.2637,
+      "step": 47177
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.429337142857143e-06,
+      "loss": 1.0856,
+      "step": 47178
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4290685714285716e-06,
+      "loss": 1.2453,
+      "step": 47179
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4288e-06,
+      "loss": 0.8025,
+      "step": 47180
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4285314285714286e-06,
+      "loss": 1.0739,
+      "step": 47181
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4282628571428572e-06,
+      "loss": 1.3241,
+      "step": 47182
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4279942857142858e-06,
+      "loss": 1.2708,
+      "step": 47183
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4277257142857142e-06,
+      "loss": 1.3221,
+      "step": 47184
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.427457142857143e-06,
+      "loss": 1.1034,
+      "step": 47185
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4271885714285716e-06,
+      "loss": 0.9505,
+      "step": 47186
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.42692e-06,
+      "loss": 1.6631,
+      "step": 47187
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4266514285714288e-06,
+      "loss": 0.8457,
+      "step": 47188
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4263828571428572e-06,
+      "loss": 1.4816,
+      "step": 47189
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4261142857142858e-06,
+      "loss": 1.0907,
+      "step": 47190
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4258457142857144e-06,
+      "loss": 0.8497,
+      "step": 47191
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.425577142857143e-06,
+      "loss": 1.1194,
+      "step": 47192
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4253085714285716e-06,
+      "loss": 1.3449,
+      "step": 47193
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.42504e-06,
+      "loss": 1.2558,
+      "step": 47194
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4247714285714288e-06,
+      "loss": 0.8738,
+      "step": 47195
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4245028571428572e-06,
+      "loss": 1.2708,
+      "step": 47196
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4242342857142858e-06,
+      "loss": 1.514,
+      "step": 47197
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4239657142857144e-06,
+      "loss": 1.2782,
+      "step": 47198
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.423697142857143e-06,
+      "loss": 1.2286,
+      "step": 47199
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4234285714285713e-06,
+      "loss": 1.138,
+      "step": 47200
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4231600000000002e-06,
+      "loss": 1.0811,
+      "step": 47201
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4228914285714288e-06,
+      "loss": 1.2395,
+      "step": 47202
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4226228571428571e-06,
+      "loss": 1.1551,
+      "step": 47203
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4223542857142857e-06,
+      "loss": 1.0444,
+      "step": 47204
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4220857142857143e-06,
+      "loss": 1.105,
+      "step": 47205
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.421817142857143e-06,
+      "loss": 1.2877,
+      "step": 47206
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4215485714285713e-06,
+      "loss": 1.5556,
+      "step": 47207
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4212800000000001e-06,
+      "loss": 1.5093,
+      "step": 47208
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4210114285714287e-06,
+      "loss": 1.0245,
+      "step": 47209
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4207428571428571e-06,
+      "loss": 1.1334,
+      "step": 47210
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.420474285714286e-06,
+      "loss": 0.9829,
+      "step": 47211
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4202057142857143e-06,
+      "loss": 1.1751,
+      "step": 47212
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.419937142857143e-06,
+      "loss": 1.0115,
+      "step": 47213
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4196685714285713e-06,
+      "loss": 0.9959,
+      "step": 47214
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4194000000000001e-06,
+      "loss": 1.3007,
+      "step": 47215
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4191314285714285e-06,
+      "loss": 1.635,
+      "step": 47216
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4188628571428571e-06,
+      "loss": 1.155,
+      "step": 47217
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.418594285714286e-06,
+      "loss": 0.9069,
+      "step": 47218
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4183257142857143e-06,
+      "loss": 1.4425,
+      "step": 47219
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.418057142857143e-06,
+      "loss": 1.5466,
+      "step": 47220
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4177885714285715e-06,
+      "loss": 0.8969,
+      "step": 47221
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4175200000000001e-06,
+      "loss": 0.8257,
+      "step": 47222
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4172514285714285e-06,
+      "loss": 1.0769,
+      "step": 47223
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4169828571428573e-06,
+      "loss": 1.4216,
+      "step": 47224
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.416714285714286e-06,
+      "loss": 0.8195,
+      "step": 47225
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4164457142857143e-06,
+      "loss": 0.9453,
+      "step": 47226
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.416177142857143e-06,
+      "loss": 1.4761,
+      "step": 47227
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4159085714285715e-06,
+      "loss": 1.1162,
+      "step": 47228
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.41564e-06,
+      "loss": 1.1167,
+      "step": 47229
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4153714285714285e-06,
+      "loss": 0.9965,
+      "step": 47230
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4151028571428573e-06,
+      "loss": 1.1912,
+      "step": 47231
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4148342857142857e-06,
+      "loss": 1.1673,
+      "step": 47232
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4145657142857143e-06,
+      "loss": 0.9769,
+      "step": 47233
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.414297142857143e-06,
+      "loss": 1.2973,
+      "step": 47234
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4140285714285715e-06,
+      "loss": 1.0534,
+      "step": 47235
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.41376e-06,
+      "loss": 1.4767,
+      "step": 47236
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4134914285714285e-06,
+      "loss": 1.4785,
+      "step": 47237
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4132228571428573e-06,
+      "loss": 0.8698,
+      "step": 47238
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4129542857142857e-06,
+      "loss": 1.1398,
+      "step": 47239
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4126857142857143e-06,
+      "loss": 1.1478,
+      "step": 47240
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.412417142857143e-06,
+      "loss": 0.96,
+      "step": 47241
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4121485714285715e-06,
+      "loss": 1.0136,
+      "step": 47242
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.41188e-06,
+      "loss": 0.8752,
+      "step": 47243
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4116114285714287e-06,
+      "loss": 1.1321,
+      "step": 47244
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4113428571428573e-06,
+      "loss": 1.1527,
+      "step": 47245
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4110742857142857e-06,
+      "loss": 1.2976,
+      "step": 47246
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4108057142857143e-06,
+      "loss": 1.2701,
+      "step": 47247
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4105371428571429e-06,
+      "loss": 1.1188,
+      "step": 47248
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4102685714285715e-06,
+      "loss": 0.996,
+      "step": 47249
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.41e-06,
+      "loss": 1.2622,
+      "step": 47250
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4097314285714287e-06,
+      "loss": 0.9927,
+      "step": 47251
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4094628571428573e-06,
+      "loss": 1.0455,
+      "step": 47252
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4091942857142856e-06,
+      "loss": 0.9613,
+      "step": 47253
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4089257142857145e-06,
+      "loss": 0.9827,
+      "step": 47254
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4086571428571428e-06,
+      "loss": 0.8423,
+      "step": 47255
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4083885714285714e-06,
+      "loss": 1.0839,
+      "step": 47256
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.40812e-06,
+      "loss": 1.1661,
+      "step": 47257
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4078514285714286e-06,
+      "loss": 1.0678,
+      "step": 47258
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4075828571428572e-06,
+      "loss": 0.8992,
+      "step": 47259
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4073142857142856e-06,
+      "loss": 1.1951,
+      "step": 47260
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4070457142857144e-06,
+      "loss": 1.3283,
+      "step": 47261
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4067771428571428e-06,
+      "loss": 1.0764,
+      "step": 47262
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4065085714285714e-06,
+      "loss": 1.5219,
+      "step": 47263
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.40624e-06,
+      "loss": 1.5774,
+      "step": 47264
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4059714285714286e-06,
+      "loss": 1.499,
+      "step": 47265
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4057028571428572e-06,
+      "loss": 1.3637,
+      "step": 47266
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4054342857142858e-06,
+      "loss": 1.0774,
+      "step": 47267
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4051657142857144e-06,
+      "loss": 0.9692,
+      "step": 47268
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4048971428571428e-06,
+      "loss": 1.4203,
+      "step": 47269
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4046285714285714e-06,
+      "loss": 0.7123,
+      "step": 47270
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.40436e-06,
+      "loss": 1.5302,
+      "step": 47271
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4040914285714286e-06,
+      "loss": 0.9057,
+      "step": 47272
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4038228571428572e-06,
+      "loss": 1.1729,
+      "step": 47273
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4035542857142858e-06,
+      "loss": 1.1884,
+      "step": 47274
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4032857142857144e-06,
+      "loss": 1.3615,
+      "step": 47275
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4030171428571428e-06,
+      "loss": 0.7585,
+      "step": 47276
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4027485714285716e-06,
+      "loss": 0.9747,
+      "step": 47277
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.40248e-06,
+      "loss": 1.205,
+      "step": 47278
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4022114285714286e-06,
+      "loss": 1.6377,
+      "step": 47279
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4019428571428572e-06,
+      "loss": 1.3106,
+      "step": 47280
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4016742857142858e-06,
+      "loss": 1.2034,
+      "step": 47281
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4014057142857144e-06,
+      "loss": 1.1913,
+      "step": 47282
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4011371428571428e-06,
+      "loss": 0.904,
+      "step": 47283
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4008685714285716e-06,
+      "loss": 1.3823,
+      "step": 47284
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4006e-06,
+      "loss": 1.1102,
+      "step": 47285
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4003314285714286e-06,
+      "loss": 1.0458,
+      "step": 47286
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4000628571428572e-06,
+      "loss": 1.2109,
+      "step": 47287
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3997942857142858e-06,
+      "loss": 0.9616,
+      "step": 47288
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3995257142857144e-06,
+      "loss": 1.0253,
+      "step": 47289
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.399257142857143e-06,
+      "loss": 0.8404,
+      "step": 47290
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3989885714285716e-06,
+      "loss": 1.3391,
+      "step": 47291
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.39872e-06,
+      "loss": 1.3231,
+      "step": 47292
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3984514285714286e-06,
+      "loss": 1.0455,
+      "step": 47293
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3981828571428572e-06,
+      "loss": 1.1926,
+      "step": 47294
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3979142857142858e-06,
+      "loss": 1.2228,
+      "step": 47295
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3976457142857144e-06,
+      "loss": 1.2832,
+      "step": 47296
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.397377142857143e-06,
+      "loss": 1.4721,
+      "step": 47297
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3971085714285716e-06,
+      "loss": 0.9955,
+      "step": 47298
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.39684e-06,
+      "loss": 1.4148,
+      "step": 47299
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3965714285714288e-06,
+      "loss": 1.2113,
+      "step": 47300
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3963028571428572e-06,
+      "loss": 1.4056,
+      "step": 47301
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3960342857142858e-06,
+      "loss": 0.9922,
+      "step": 47302
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3957657142857144e-06,
+      "loss": 1.0661,
+      "step": 47303
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.395497142857143e-06,
+      "loss": 1.3153,
+      "step": 47304
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3952285714285716e-06,
+      "loss": 0.9395,
+      "step": 47305
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.39496e-06,
+      "loss": 0.6784,
+      "step": 47306
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3946914285714288e-06,
+      "loss": 1.0378,
+      "step": 47307
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3944228571428571e-06,
+      "loss": 1.1093,
+      "step": 47308
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3941542857142857e-06,
+      "loss": 0.9583,
+      "step": 47309
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3938857142857143e-06,
+      "loss": 1.3262,
+      "step": 47310
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.393617142857143e-06,
+      "loss": 0.8119,
+      "step": 47311
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3933485714285715e-06,
+      "loss": 1.1478,
+      "step": 47312
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3930800000000001e-06,
+      "loss": 1.411,
+      "step": 47313
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3928114285714287e-06,
+      "loss": 1.0674,
+      "step": 47314
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3925428571428571e-06,
+      "loss": 0.9623,
+      "step": 47315
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3922742857142857e-06,
+      "loss": 1.4817,
+      "step": 47316
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3920057142857143e-06,
+      "loss": 0.9159,
+      "step": 47317
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.391737142857143e-06,
+      "loss": 0.8776,
+      "step": 47318
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3914685714285713e-06,
+      "loss": 0.8791,
+      "step": 47319
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3912000000000001e-06,
+      "loss": 0.9323,
+      "step": 47320
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3909314285714287e-06,
+      "loss": 1.285,
+      "step": 47321
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3906628571428571e-06,
+      "loss": 1.1843,
+      "step": 47322
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.390394285714286e-06,
+      "loss": 1.5018,
+      "step": 47323
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3901257142857143e-06,
+      "loss": 1.2017,
+      "step": 47324
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.389857142857143e-06,
+      "loss": 1.6648,
+      "step": 47325
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3895885714285713e-06,
+      "loss": 0.9412,
+      "step": 47326
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3893200000000001e-06,
+      "loss": 0.9919,
+      "step": 47327
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3890514285714287e-06,
+      "loss": 1.0511,
+      "step": 47328
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.388782857142857e-06,
+      "loss": 0.987,
+      "step": 47329
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.388514285714286e-06,
+      "loss": 1.0735,
+      "step": 47330
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3882457142857143e-06,
+      "loss": 0.8045,
+      "step": 47331
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.387977142857143e-06,
+      "loss": 1.184,
+      "step": 47332
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3877085714285715e-06,
+      "loss": 1.4364,
+      "step": 47333
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.38744e-06,
+      "loss": 1.1711,
+      "step": 47334
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3871714285714285e-06,
+      "loss": 1.3578,
+      "step": 47335
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3869028571428573e-06,
+      "loss": 0.8783,
+      "step": 47336
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.3866342857142859e-06,
+      "loss": 0.7813,
+      "step": 47337
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3863657142857143e-06,
+      "loss": 1.1347,
+      "step": 47338
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3860971428571429e-06,
+      "loss": 0.9967,
+      "step": 47339
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3858285714285715e-06,
+      "loss": 1.182,
+      "step": 47340
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.38556e-06,
+      "loss": 1.4778,
+      "step": 47341
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3852914285714285e-06,
+      "loss": 1.0119,
+      "step": 47342
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3850228571428573e-06,
+      "loss": 1.3518,
+      "step": 47343
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3847542857142859e-06,
+      "loss": 1.0885,
+      "step": 47344
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3844857142857143e-06,
+      "loss": 0.9364,
+      "step": 47345
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.384217142857143e-06,
+      "loss": 1.2535,
+      "step": 47346
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3839485714285715e-06,
+      "loss": 1.2842,
+      "step": 47347
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.38368e-06,
+      "loss": 1.2837,
+      "step": 47348
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3834114285714285e-06,
+      "loss": 1.1944,
+      "step": 47349
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3831428571428573e-06,
+      "loss": 1.606,
+      "step": 47350
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3828742857142857e-06,
+      "loss": 1.3141,
+      "step": 47351
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3826057142857143e-06,
+      "loss": 1.3142,
+      "step": 47352
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.382337142857143e-06,
+      "loss": 0.8633,
+      "step": 47353
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3820685714285715e-06,
+      "loss": 1.3343,
+      "step": 47354
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3818e-06,
+      "loss": 1.3676,
+      "step": 47355
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3815314285714287e-06,
+      "loss": 1.184,
+      "step": 47356
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3812628571428573e-06,
+      "loss": 1.2384,
+      "step": 47357
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3809942857142856e-06,
+      "loss": 0.9218,
+      "step": 47358
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3807257142857142e-06,
+      "loss": 0.9313,
+      "step": 47359
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.380457142857143e-06,
+      "loss": 1.2954,
+      "step": 47360
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3801885714285714e-06,
+      "loss": 1.0454,
+      "step": 47361
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.37992e-06,
+      "loss": 1.0224,
+      "step": 47362
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3796514285714286e-06,
+      "loss": 1.0959,
+      "step": 47363
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3793828571428572e-06,
+      "loss": 1.0812,
+      "step": 47364
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3791142857142856e-06,
+      "loss": 1.2856,
+      "step": 47365
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3788457142857144e-06,
+      "loss": 1.287,
+      "step": 47366
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3785771428571428e-06,
+      "loss": 1.1065,
+      "step": 47367
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3783085714285714e-06,
+      "loss": 1.0114,
+      "step": 47368
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.37804e-06,
+      "loss": 1.322,
+      "step": 47369
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3777714285714286e-06,
+      "loss": 1.0813,
+      "step": 47370
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3775028571428572e-06,
+      "loss": 0.9884,
+      "step": 47371
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3772342857142856e-06,
+      "loss": 1.297,
+      "step": 47372
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3769657142857144e-06,
+      "loss": 1.0528,
+      "step": 47373
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3766971428571428e-06,
+      "loss": 0.897,
+      "step": 47374
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3764285714285714e-06,
+      "loss": 1.1838,
+      "step": 47375
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3761600000000002e-06,
+      "loss": 1.0254,
+      "step": 47376
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3758914285714286e-06,
+      "loss": 1.2737,
+      "step": 47377
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3756228571428572e-06,
+      "loss": 1.4805,
+      "step": 47378
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3753542857142858e-06,
+      "loss": 1.3216,
+      "step": 47379
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3750857142857144e-06,
+      "loss": 0.8547,
+      "step": 47380
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3748171428571428e-06,
+      "loss": 0.967,
+      "step": 47381
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3745485714285714e-06,
+      "loss": 1.1108,
+      "step": 47382
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.37428e-06,
+      "loss": 1.2136,
+      "step": 47383
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3740114285714286e-06,
+      "loss": 0.9771,
+      "step": 47384
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3737428571428572e-06,
+      "loss": 0.7334,
+      "step": 47385
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3734742857142858e-06,
+      "loss": 1.4678,
+      "step": 47386
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3732057142857144e-06,
+      "loss": 1.0027,
+      "step": 47387
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3729371428571428e-06,
+      "loss": 0.4791,
+      "step": 47388
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3726685714285716e-06,
+      "loss": 1.1101,
+      "step": 47389
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3724e-06,
+      "loss": 1.2699,
+      "step": 47390
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3721314285714286e-06,
+      "loss": 1.121,
+      "step": 47391
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3718628571428572e-06,
+      "loss": 1.6276,
+      "step": 47392
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3715942857142858e-06,
+      "loss": 1.7214,
+      "step": 47393
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3713257142857144e-06,
+      "loss": 1.0324,
+      "step": 47394
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3710571428571428e-06,
+      "loss": 0.9504,
+      "step": 47395
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3707885714285716e-06,
+      "loss": 1.352,
+      "step": 47396
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.37052e-06,
+      "loss": 0.9743,
+      "step": 47397
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3702514285714286e-06,
+      "loss": 1.0444,
+      "step": 47398
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3699828571428574e-06,
+      "loss": 1.0065,
+      "step": 47399
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3697142857142858e-06,
+      "loss": 0.9677,
+      "step": 47400
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3694457142857144e-06,
+      "loss": 1.0889,
+      "step": 47401
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.369177142857143e-06,
+      "loss": 1.0129,
+      "step": 47402
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3689085714285716e-06,
+      "loss": 0.9583,
+      "step": 47403
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.36864e-06,
+      "loss": 1.4355,
+      "step": 47404
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3683714285714286e-06,
+      "loss": 1.2044,
+      "step": 47405
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3681028571428572e-06,
+      "loss": 1.0915,
+      "step": 47406
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3678342857142858e-06,
+      "loss": 1.0066,
+      "step": 47407
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3675657142857144e-06,
+      "loss": 1.4227,
+      "step": 47408
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.367297142857143e-06,
+      "loss": 0.7392,
+      "step": 47409
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3670285714285716e-06,
+      "loss": 0.6607,
+      "step": 47410
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.36676e-06,
+      "loss": 1.0248,
+      "step": 47411
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3664914285714287e-06,
+      "loss": 1.0272,
+      "step": 47412
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3662228571428571e-06,
+      "loss": 1.3687,
+      "step": 47413
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3659542857142857e-06,
+      "loss": 0.8631,
+      "step": 47414
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3656857142857143e-06,
+      "loss": 1.0313,
+      "step": 47415
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.365417142857143e-06,
+      "loss": 1.1893,
+      "step": 47416
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3651485714285715e-06,
+      "loss": 1.0999,
+      "step": 47417
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.36488e-06,
+      "loss": 0.803,
+      "step": 47418
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3646114285714287e-06,
+      "loss": 0.6512,
+      "step": 47419
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3643428571428571e-06,
+      "loss": 1.2098,
+      "step": 47420
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3640742857142857e-06,
+      "loss": 1.0926,
+      "step": 47421
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3638057142857143e-06,
+      "loss": 0.8821,
+      "step": 47422
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.363537142857143e-06,
+      "loss": 1.5562,
+      "step": 47423
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3632685714285715e-06,
+      "loss": 0.8917,
+      "step": 47424
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3630000000000001e-06,
+      "loss": 0.8511,
+      "step": 47425
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3627314285714287e-06,
+      "loss": 1.213,
+      "step": 47426
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3624628571428571e-06,
+      "loss": 1.1079,
+      "step": 47427
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3621942857142857e-06,
+      "loss": 1.1676,
+      "step": 47428
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3619257142857143e-06,
+      "loss": 1.4226,
+      "step": 47429
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.361657142857143e-06,
+      "loss": 0.7261,
+      "step": 47430
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3613885714285715e-06,
+      "loss": 0.5641,
+      "step": 47431
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3611200000000001e-06,
+      "loss": 1.1177,
+      "step": 47432
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3608514285714287e-06,
+      "loss": 1.2417,
+      "step": 47433
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.360582857142857e-06,
+      "loss": 1.3538,
+      "step": 47434
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.360314285714286e-06,
+      "loss": 0.9913,
+      "step": 47435
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3600457142857143e-06,
+      "loss": 1.0211,
+      "step": 47436
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3597771428571429e-06,
+      "loss": 1.274,
+      "step": 47437
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3595085714285715e-06,
+      "loss": 1.1859,
+      "step": 47438
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.35924e-06,
+      "loss": 1.2048,
+      "step": 47439
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3589714285714287e-06,
+      "loss": 1.0832,
+      "step": 47440
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.358702857142857e-06,
+      "loss": 1.2798,
+      "step": 47441
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3584342857142859e-06,
+      "loss": 1.273,
+      "step": 47442
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3581657142857143e-06,
+      "loss": 0.8971,
+      "step": 47443
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3578971428571429e-06,
+      "loss": 1.2725,
+      "step": 47444
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3576285714285715e-06,
+      "loss": 1.2817,
+      "step": 47445
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.35736e-06,
+      "loss": 1.1165,
+      "step": 47446
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3570914285714287e-06,
+      "loss": 0.9625,
+      "step": 47447
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3568228571428573e-06,
+      "loss": 0.888,
+      "step": 47448
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3565542857142859e-06,
+      "loss": 1.2158,
+      "step": 47449
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3562857142857143e-06,
+      "loss": 0.8961,
+      "step": 47450
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3560171428571429e-06,
+      "loss": 1.3993,
+      "step": 47451
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3557485714285715e-06,
+      "loss": 1.1997,
+      "step": 47452
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.35548e-06,
+      "loss": 1.0777,
+      "step": 47453
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3552114285714285e-06,
+      "loss": 0.8852,
+      "step": 47454
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3549428571428573e-06,
+      "loss": 1.3389,
+      "step": 47455
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3546742857142859e-06,
+      "loss": 1.1457,
+      "step": 47456
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3544057142857143e-06,
+      "loss": 1.0106,
+      "step": 47457
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.354137142857143e-06,
+      "loss": 0.842,
+      "step": 47458
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3538685714285715e-06,
+      "loss": 1.1759,
+      "step": 47459
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3536e-06,
+      "loss": 1.0186,
+      "step": 47460
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3533314285714284e-06,
+      "loss": 0.7654,
+      "step": 47461
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3530628571428572e-06,
+      "loss": 0.7736,
+      "step": 47462
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3527942857142858e-06,
+      "loss": 1.2613,
+      "step": 47463
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3525257142857142e-06,
+      "loss": 1.3893,
+      "step": 47464
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.352257142857143e-06,
+      "loss": 0.9032,
+      "step": 47465
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3519885714285714e-06,
+      "loss": 1.1739,
+      "step": 47466
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.35172e-06,
+      "loss": 1.1816,
+      "step": 47467
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3514514285714286e-06,
+      "loss": 0.9816,
+      "step": 47468
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3511828571428572e-06,
+      "loss": 1.3682,
+      "step": 47469
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3509142857142856e-06,
+      "loss": 1.1513,
+      "step": 47470
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3506457142857142e-06,
+      "loss": 1.4391,
+      "step": 47471
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.350377142857143e-06,
+      "loss": 1.217,
+      "step": 47472
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3501085714285714e-06,
+      "loss": 0.9625,
+      "step": 47473
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.34984e-06,
+      "loss": 1.2365,
+      "step": 47474
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3495714285714286e-06,
+      "loss": 0.7313,
+      "step": 47475
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3493028571428572e-06,
+      "loss": 1.0909,
+      "step": 47476
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3490342857142856e-06,
+      "loss": 1.0804,
+      "step": 47477
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3487657142857144e-06,
+      "loss": 1.491,
+      "step": 47478
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.348497142857143e-06,
+      "loss": 0.6797,
+      "step": 47479
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3482285714285714e-06,
+      "loss": 1.1445,
+      "step": 47480
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.34796e-06,
+      "loss": 1.132,
+      "step": 47481
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3476914285714286e-06,
+      "loss": 1.3616,
+      "step": 47482
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3474228571428572e-06,
+      "loss": 0.7903,
+      "step": 47483
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3471542857142856e-06,
+      "loss": 1.2892,
+      "step": 47484
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3468857142857144e-06,
+      "loss": 0.7634,
+      "step": 47485
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3466171428571428e-06,
+      "loss": 0.7838,
+      "step": 47486
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3463485714285714e-06,
+      "loss": 0.7618,
+      "step": 47487
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3460800000000002e-06,
+      "loss": 0.8519,
+      "step": 47488
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3458114285714286e-06,
+      "loss": 0.9707,
+      "step": 47489
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3455428571428572e-06,
+      "loss": 0.8983,
+      "step": 47490
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3452742857142858e-06,
+      "loss": 1.8545,
+      "step": 47491
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3450057142857144e-06,
+      "loss": 1.4447,
+      "step": 47492
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3447371428571428e-06,
+      "loss": 0.8318,
+      "step": 47493
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3444685714285714e-06,
+      "loss": 0.7021,
+      "step": 47494
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3442000000000002e-06,
+      "loss": 0.8587,
+      "step": 47495
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3439314285714286e-06,
+      "loss": 0.6406,
+      "step": 47496
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3436628571428572e-06,
+      "loss": 1.3769,
+      "step": 47497
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3433942857142858e-06,
+      "loss": 1.2597,
+      "step": 47498
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3431257142857144e-06,
+      "loss": 1.4048,
+      "step": 47499
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3428571428571428e-06,
+      "loss": 1.2388,
+      "step": 47500
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3425885714285716e-06,
+      "loss": 1.1658,
+      "step": 47501
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3423200000000002e-06,
+      "loss": 1.1375,
+      "step": 47502
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3420514285714286e-06,
+      "loss": 1.0846,
+      "step": 47503
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3417828571428572e-06,
+      "loss": 1.4641,
+      "step": 47504
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3415142857142858e-06,
+      "loss": 0.9391,
+      "step": 47505
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3412457142857144e-06,
+      "loss": 0.7637,
+      "step": 47506
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3409771428571428e-06,
+      "loss": 1.1908,
+      "step": 47507
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3407085714285716e-06,
+      "loss": 1.2524,
+      "step": 47508
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.34044e-06,
+      "loss": 1.0916,
+      "step": 47509
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3401714285714286e-06,
+      "loss": 1.2337,
+      "step": 47510
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3399028571428574e-06,
+      "loss": 1.3286,
+      "step": 47511
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3396342857142858e-06,
+      "loss": 1.0782,
+      "step": 47512
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3393657142857143e-06,
+      "loss": 1.6777,
+      "step": 47513
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.339097142857143e-06,
+      "loss": 1.2356,
+      "step": 47514
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3388285714285715e-06,
+      "loss": 1.2836,
+      "step": 47515
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.33856e-06,
+      "loss": 1.3588,
+      "step": 47516
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3382914285714285e-06,
+      "loss": 1.3919,
+      "step": 47517
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3380228571428573e-06,
+      "loss": 1.3184,
+      "step": 47518
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3377542857142857e-06,
+      "loss": 1.5001,
+      "step": 47519
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3374857142857143e-06,
+      "loss": 0.8959,
+      "step": 47520
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.337217142857143e-06,
+      "loss": 1.0899,
+      "step": 47521
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3369485714285715e-06,
+      "loss": 1.4453,
+      "step": 47522
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.33668e-06,
+      "loss": 1.5977,
+      "step": 47523
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3364114285714287e-06,
+      "loss": 1.5325,
+      "step": 47524
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3361428571428571e-06,
+      "loss": 1.35,
+      "step": 47525
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3358742857142857e-06,
+      "loss": 1.3487,
+      "step": 47526
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3356057142857143e-06,
+      "loss": 1.0599,
+      "step": 47527
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.335337142857143e-06,
+      "loss": 0.9638,
+      "step": 47528
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3350685714285715e-06,
+      "loss": 1.1459,
+      "step": 47529
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3348e-06,
+      "loss": 1.1229,
+      "step": 47530
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3345314285714287e-06,
+      "loss": 1.7637,
+      "step": 47531
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3342628571428571e-06,
+      "loss": 1.066,
+      "step": 47532
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3339942857142857e-06,
+      "loss": 1.5125,
+      "step": 47533
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3337257142857145e-06,
+      "loss": 0.9675,
+      "step": 47534
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.333457142857143e-06,
+      "loss": 1.3043,
+      "step": 47535
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3331885714285715e-06,
+      "loss": 0.9289,
+      "step": 47536
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3329200000000001e-06,
+      "loss": 1.2171,
+      "step": 47537
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3326514285714287e-06,
+      "loss": 1.2456,
+      "step": 47538
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.332382857142857e-06,
+      "loss": 1.3231,
+      "step": 47539
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3321142857142857e-06,
+      "loss": 0.9559,
+      "step": 47540
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3318457142857143e-06,
+      "loss": 1.058,
+      "step": 47541
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3315771428571429e-06,
+      "loss": 0.8513,
+      "step": 47542
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3313085714285715e-06,
+      "loss": 1.1328,
+      "step": 47543
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.33104e-06,
+      "loss": 0.8768,
+      "step": 47544
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3307714285714287e-06,
+      "loss": 0.6771,
+      "step": 47545
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.330502857142857e-06,
+      "loss": 1.0984,
+      "step": 47546
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3302342857142859e-06,
+      "loss": 0.8443,
+      "step": 47547
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3299657142857143e-06,
+      "loss": 0.9582,
+      "step": 47548
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3296971428571429e-06,
+      "loss": 0.8109,
+      "step": 47549
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3294285714285715e-06,
+      "loss": 0.7447,
+      "step": 47550
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.32916e-06,
+      "loss": 1.2564,
+      "step": 47551
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3288914285714287e-06,
+      "loss": 1.1905,
+      "step": 47552
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.328622857142857e-06,
+      "loss": 1.0871,
+      "step": 47553
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3283542857142859e-06,
+      "loss": 1.4211,
+      "step": 47554
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3280857142857143e-06,
+      "loss": 0.9415,
+      "step": 47555
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3278171428571429e-06,
+      "loss": 1.5677,
+      "step": 47556
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3275485714285715e-06,
+      "loss": 0.9232,
+      "step": 47557
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.32728e-06,
+      "loss": 0.9294,
+      "step": 47558
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3270114285714287e-06,
+      "loss": 0.9941,
+      "step": 47559
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3267428571428573e-06,
+      "loss": 1.1949,
+      "step": 47560
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3264742857142859e-06,
+      "loss": 1.0689,
+      "step": 47561
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3262057142857143e-06,
+      "loss": 0.7198,
+      "step": 47562
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3259371428571429e-06,
+      "loss": 0.9779,
+      "step": 47563
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3256685714285714e-06,
+      "loss": 1.3176,
+      "step": 47564
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3254e-06,
+      "loss": 1.1464,
+      "step": 47565
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3251314285714286e-06,
+      "loss": 1.0017,
+      "step": 47566
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3248628571428572e-06,
+      "loss": 1.2161,
+      "step": 47567
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3245942857142858e-06,
+      "loss": 0.7821,
+      "step": 47568
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3243257142857142e-06,
+      "loss": 0.8567,
+      "step": 47569
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.324057142857143e-06,
+      "loss": 1.6372,
+      "step": 47570
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3237885714285714e-06,
+      "loss": 0.7118,
+      "step": 47571
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.32352e-06,
+      "loss": 1.5161,
+      "step": 47572
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3232514285714284e-06,
+      "loss": 1.0872,
+      "step": 47573
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3229828571428572e-06,
+      "loss": 1.2743,
+      "step": 47574
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3227142857142858e-06,
+      "loss": 1.2543,
+      "step": 47575
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3224457142857142e-06,
+      "loss": 0.6667,
+      "step": 47576
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.322177142857143e-06,
+      "loss": 1.0419,
+      "step": 47577
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3219085714285714e-06,
+      "loss": 1.2472,
+      "step": 47578
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.32164e-06,
+      "loss": 1.0366,
+      "step": 47579
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3213714285714286e-06,
+      "loss": 1.1162,
+      "step": 47580
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3211028571428572e-06,
+      "loss": 1.5093,
+      "step": 47581
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3208342857142858e-06,
+      "loss": 1.4337,
+      "step": 47582
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3205657142857142e-06,
+      "loss": 1.0032,
+      "step": 47583
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.320297142857143e-06,
+      "loss": 0.9887,
+      "step": 47584
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3200285714285714e-06,
+      "loss": 1.2355,
+      "step": 47585
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.31976e-06,
+      "loss": 1.3243,
+      "step": 47586
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3194914285714286e-06,
+      "loss": 0.9901,
+      "step": 47587
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3192228571428572e-06,
+      "loss": 1.1118,
+      "step": 47588
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3189542857142856e-06,
+      "loss": 0.9734,
+      "step": 47589
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3186857142857144e-06,
+      "loss": 1.0117,
+      "step": 47590
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.318417142857143e-06,
+      "loss": 1.219,
+      "step": 47591
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3181485714285714e-06,
+      "loss": 0.9232,
+      "step": 47592
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.31788e-06,
+      "loss": 1.2924,
+      "step": 47593
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3176114285714286e-06,
+      "loss": 1.0383,
+      "step": 47594
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3173428571428572e-06,
+      "loss": 1.2242,
+      "step": 47595
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3170742857142856e-06,
+      "loss": 1.2392,
+      "step": 47596
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3168057142857144e-06,
+      "loss": 0.8463,
+      "step": 47597
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.316537142857143e-06,
+      "loss": 1.1089,
+      "step": 47598
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3162685714285714e-06,
+      "loss": 0.9038,
+      "step": 47599
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3160000000000002e-06,
+      "loss": 1.2399,
+      "step": 47600
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3157314285714286e-06,
+      "loss": 1.0014,
+      "step": 47601
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3154628571428572e-06,
+      "loss": 1.1726,
+      "step": 47602
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3151942857142858e-06,
+      "loss": 1.0441,
+      "step": 47603
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3149257142857144e-06,
+      "loss": 1.2217,
+      "step": 47604
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.314657142857143e-06,
+      "loss": 0.7874,
+      "step": 47605
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3143885714285714e-06,
+      "loss": 0.5215,
+      "step": 47606
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3141200000000002e-06,
+      "loss": 1.1082,
+      "step": 47607
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3138514285714286e-06,
+      "loss": 1.1901,
+      "step": 47608
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3135828571428572e-06,
+      "loss": 1.5596,
+      "step": 47609
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3133142857142858e-06,
+      "loss": 1.1074,
+      "step": 47610
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3130457142857144e-06,
+      "loss": 1.4375,
+      "step": 47611
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3127771428571428e-06,
+      "loss": 1.117,
+      "step": 47612
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3125085714285716e-06,
+      "loss": 1.2309,
+      "step": 47613
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3122400000000002e-06,
+      "loss": 1.0096,
+      "step": 47614
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3119714285714285e-06,
+      "loss": 1.3647,
+      "step": 47615
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3117028571428571e-06,
+      "loss": 1.1876,
+      "step": 47616
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3114342857142857e-06,
+      "loss": 0.9096,
+      "step": 47617
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3111657142857143e-06,
+      "loss": 1.1112,
+      "step": 47618
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3108971428571427e-06,
+      "loss": 1.1869,
+      "step": 47619
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3106285714285715e-06,
+      "loss": 0.9836,
+      "step": 47620
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3103600000000001e-06,
+      "loss": 1.3148,
+      "step": 47621
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3100914285714285e-06,
+      "loss": 1.1254,
+      "step": 47622
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3098228571428573e-06,
+      "loss": 1.1875,
+      "step": 47623
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3095542857142857e-06,
+      "loss": 1.1516,
+      "step": 47624
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3092857142857143e-06,
+      "loss": 1.2026,
+      "step": 47625
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.309017142857143e-06,
+      "loss": 0.8845,
+      "step": 47626
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3087485714285715e-06,
+      "loss": 0.8977,
+      "step": 47627
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.30848e-06,
+      "loss": 1.2337,
+      "step": 47628
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3082114285714285e-06,
+      "loss": 1.1036,
+      "step": 47629
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3079428571428573e-06,
+      "loss": 1.0768,
+      "step": 47630
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3076742857142857e-06,
+      "loss": 1.4998,
+      "step": 47631
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3074057142857143e-06,
+      "loss": 1.0999,
+      "step": 47632
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.307137142857143e-06,
+      "loss": 1.1515,
+      "step": 47633
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3068685714285715e-06,
+      "loss": 1.1285,
+      "step": 47634
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3066e-06,
+      "loss": 1.368,
+      "step": 47635
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3063314285714287e-06,
+      "loss": 1.145,
+      "step": 47636
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3060628571428573e-06,
+      "loss": 1.74,
+      "step": 47637
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3057942857142857e-06,
+      "loss": 1.3018,
+      "step": 47638
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3055257142857143e-06,
+      "loss": 0.9649,
+      "step": 47639
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.305257142857143e-06,
+      "loss": 0.9909,
+      "step": 47640
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3049885714285715e-06,
+      "loss": 0.7997,
+      "step": 47641
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3047199999999999e-06,
+      "loss": 1.2552,
+      "step": 47642
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3044514285714287e-06,
+      "loss": 1.1966,
+      "step": 47643
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.304182857142857e-06,
+      "loss": 1.2965,
+      "step": 47644
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3039142857142857e-06,
+      "loss": 1.0433,
+      "step": 47645
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3036457142857145e-06,
+      "loss": 1.1789,
+      "step": 47646
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3033771428571429e-06,
+      "loss": 1.2866,
+      "step": 47647
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3031085714285715e-06,
+      "loss": 0.8148,
+      "step": 47648
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.30284e-06,
+      "loss": 1.1266,
+      "step": 47649
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3025714285714287e-06,
+      "loss": 1.2238,
+      "step": 47650
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.302302857142857e-06,
+      "loss": 1.1438,
+      "step": 47651
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3020342857142857e-06,
+      "loss": 0.8744,
+      "step": 47652
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3017657142857145e-06,
+      "loss": 1.4252,
+      "step": 47653
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3014971428571429e-06,
+      "loss": 1.3825,
+      "step": 47654
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3012285714285715e-06,
+      "loss": 0.7987,
+      "step": 47655
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.30096e-06,
+      "loss": 0.8562,
+      "step": 47656
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3006914285714287e-06,
+      "loss": 0.813,
+      "step": 47657
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.300422857142857e-06,
+      "loss": 1.2207,
+      "step": 47658
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.3001542857142859e-06,
+      "loss": 1.0619,
+      "step": 47659
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2998857142857143e-06,
+      "loss": 1.1384,
+      "step": 47660
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2996171428571429e-06,
+      "loss": 1.2287,
+      "step": 47661
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2993485714285715e-06,
+      "loss": 1.0646,
+      "step": 47662
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.29908e-06,
+      "loss": 1.5114,
+      "step": 47663
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2988114285714287e-06,
+      "loss": 0.8994,
+      "step": 47664
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.298542857142857e-06,
+      "loss": 1.4384,
+      "step": 47665
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2982742857142859e-06,
+      "loss": 1.0733,
+      "step": 47666
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2980057142857142e-06,
+      "loss": 0.8251,
+      "step": 47667
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2977371428571428e-06,
+      "loss": 1.0502,
+      "step": 47668
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2974685714285717e-06,
+      "loss": 1.2345,
+      "step": 47669
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2972e-06,
+      "loss": 1.0133,
+      "step": 47670
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2969314285714286e-06,
+      "loss": 1.1797,
+      "step": 47671
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2966628571428572e-06,
+      "loss": 1.3689,
+      "step": 47672
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2963942857142858e-06,
+      "loss": 1.2343,
+      "step": 47673
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2961257142857142e-06,
+      "loss": 1.0776,
+      "step": 47674
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2958571428571428e-06,
+      "loss": 0.9263,
+      "step": 47675
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2955885714285714e-06,
+      "loss": 1.2111,
+      "step": 47676
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.29532e-06,
+      "loss": 1.3729,
+      "step": 47677
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2950514285714286e-06,
+      "loss": 1.259,
+      "step": 47678
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2947828571428572e-06,
+      "loss": 1.4964,
+      "step": 47679
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2945142857142858e-06,
+      "loss": 1.4351,
+      "step": 47680
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2942457142857142e-06,
+      "loss": 1.2746,
+      "step": 47681
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.293977142857143e-06,
+      "loss": 1.2782,
+      "step": 47682
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2937085714285714e-06,
+      "loss": 1.2495,
+      "step": 47683
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.29344e-06,
+      "loss": 0.988,
+      "step": 47684
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2931714285714286e-06,
+      "loss": 1.1521,
+      "step": 47685
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2929028571428572e-06,
+      "loss": 1.1859,
+      "step": 47686
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.2926342857142858e-06,
+      "loss": 0.9987,
+      "step": 47687
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2923657142857142e-06,
+      "loss": 1.0815,
+      "step": 47688
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.292097142857143e-06,
+      "loss": 0.6571,
+      "step": 47689
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2918285714285714e-06,
+      "loss": 1.1176,
+      "step": 47690
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.29156e-06,
+      "loss": 1.3329,
+      "step": 47691
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2912914285714288e-06,
+      "loss": 0.9442,
+      "step": 47692
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2910228571428572e-06,
+      "loss": 1.1051,
+      "step": 47693
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2907542857142858e-06,
+      "loss": 1.2945,
+      "step": 47694
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2904857142857142e-06,
+      "loss": 1.2976,
+      "step": 47695
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.290217142857143e-06,
+      "loss": 1.2248,
+      "step": 47696
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2899485714285714e-06,
+      "loss": 0.9593,
+      "step": 47697
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.28968e-06,
+      "loss": 1.4808,
+      "step": 47698
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2894114285714286e-06,
+      "loss": 1.8115,
+      "step": 47699
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2891428571428572e-06,
+      "loss": 1.028,
+      "step": 47700
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2888742857142858e-06,
+      "loss": 0.9737,
+      "step": 47701
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2886057142857144e-06,
+      "loss": 0.9294,
+      "step": 47702
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.288337142857143e-06,
+      "loss": 0.8159,
+      "step": 47703
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2880685714285714e-06,
+      "loss": 1.2474,
+      "step": 47704
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2878e-06,
+      "loss": 0.7513,
+      "step": 47705
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2875314285714286e-06,
+      "loss": 1.4638,
+      "step": 47706
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2872628571428572e-06,
+      "loss": 0.8147,
+      "step": 47707
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2869942857142858e-06,
+      "loss": 0.8965,
+      "step": 47708
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2867257142857144e-06,
+      "loss": 0.8511,
+      "step": 47709
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.286457142857143e-06,
+      "loss": 0.8728,
+      "step": 47710
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2861885714285714e-06,
+      "loss": 0.7787,
+      "step": 47711
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2859200000000002e-06,
+      "loss": 0.9595,
+      "step": 47712
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2856514285714286e-06,
+      "loss": 1.8268,
+      "step": 47713
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2853828571428572e-06,
+      "loss": 1.0118,
+      "step": 47714
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2851142857142858e-06,
+      "loss": 0.7608,
+      "step": 47715
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2848457142857144e-06,
+      "loss": 1.4559,
+      "step": 47716
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.284577142857143e-06,
+      "loss": 0.9321,
+      "step": 47717
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2843085714285713e-06,
+      "loss": 1.268,
+      "step": 47718
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2840400000000002e-06,
+      "loss": 1.6202,
+      "step": 47719
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2837714285714285e-06,
+      "loss": 1.1508,
+      "step": 47720
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2835028571428571e-06,
+      "loss": 1.094,
+      "step": 47721
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2832342857142857e-06,
+      "loss": 1.0299,
+      "step": 47722
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2829657142857143e-06,
+      "loss": 1.3286,
+      "step": 47723
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.282697142857143e-06,
+      "loss": 1.1178,
+      "step": 47724
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2824285714285715e-06,
+      "loss": 0.8019,
+      "step": 47725
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2821600000000001e-06,
+      "loss": 1.2644,
+      "step": 47726
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2818914285714285e-06,
+      "loss": 1.1695,
+      "step": 47727
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2816228571428571e-06,
+      "loss": 0.8691,
+      "step": 47728
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2813542857142857e-06,
+      "loss": 1.0214,
+      "step": 47729
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2810857142857143e-06,
+      "loss": 1.3266,
+      "step": 47730
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2808171428571427e-06,
+      "loss": 0.9666,
+      "step": 47731
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2805485714285715e-06,
+      "loss": 0.9138,
+      "step": 47732
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2802800000000001e-06,
+      "loss": 1.3156,
+      "step": 47733
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2800114285714285e-06,
+      "loss": 1.3675,
+      "step": 47734
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2797428571428573e-06,
+      "loss": 1.3106,
+      "step": 47735
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2794742857142857e-06,
+      "loss": 0.964,
+      "step": 47736
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2792057142857143e-06,
+      "loss": 0.6893,
+      "step": 47737
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.278937142857143e-06,
+      "loss": 1.0604,
+      "step": 47738
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2786685714285715e-06,
+      "loss": 0.964,
+      "step": 47739
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2784000000000001e-06,
+      "loss": 0.7575,
+      "step": 47740
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2781314285714285e-06,
+      "loss": 1.1252,
+      "step": 47741
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2778628571428573e-06,
+      "loss": 1.2528,
+      "step": 47742
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2775942857142857e-06,
+      "loss": 1.1641,
+      "step": 47743
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2773257142857143e-06,
+      "loss": 1.6901,
+      "step": 47744
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.277057142857143e-06,
+      "loss": 1.0057,
+      "step": 47745
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2767885714285715e-06,
+      "loss": 1.0567,
+      "step": 47746
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2765199999999999e-06,
+      "loss": 1.2409,
+      "step": 47747
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2762514285714287e-06,
+      "loss": 0.7751,
+      "step": 47748
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2759828571428573e-06,
+      "loss": 1.2733,
+      "step": 47749
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2757142857142857e-06,
+      "loss": 1.0231,
+      "step": 47750
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2754457142857143e-06,
+      "loss": 1.0221,
+      "step": 47751
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2751771428571429e-06,
+      "loss": 1.2267,
+      "step": 47752
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2749085714285715e-06,
+      "loss": 0.8672,
+      "step": 47753
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2746399999999999e-06,
+      "loss": 0.9685,
+      "step": 47754
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2743714285714287e-06,
+      "loss": 1.444,
+      "step": 47755
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2741028571428573e-06,
+      "loss": 1.1128,
+      "step": 47756
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2738342857142857e-06,
+      "loss": 1.1203,
+      "step": 47757
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2735657142857145e-06,
+      "loss": 1.1047,
+      "step": 47758
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2732971428571429e-06,
+      "loss": 0.9197,
+      "step": 47759
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2730285714285715e-06,
+      "loss": 0.9116,
+      "step": 47760
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.27276e-06,
+      "loss": 0.8657,
+      "step": 47761
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2724914285714287e-06,
+      "loss": 1.2157,
+      "step": 47762
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.272222857142857e-06,
+      "loss": 1.0502,
+      "step": 47763
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2719542857142857e-06,
+      "loss": 1.3038,
+      "step": 47764
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2716857142857145e-06,
+      "loss": 0.8227,
+      "step": 47765
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2714171428571429e-06,
+      "loss": 1.4382,
+      "step": 47766
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2711485714285715e-06,
+      "loss": 1.1913,
+      "step": 47767
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.27088e-06,
+      "loss": 1.6941,
+      "step": 47768
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2706114285714287e-06,
+      "loss": 1.1337,
+      "step": 47769
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.270342857142857e-06,
+      "loss": 1.0776,
+      "step": 47770
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2700742857142859e-06,
+      "loss": 1.3044,
+      "step": 47771
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2698057142857145e-06,
+      "loss": 0.759,
+      "step": 47772
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2695371428571428e-06,
+      "loss": 1.3521,
+      "step": 47773
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2692685714285714e-06,
+      "loss": 0.8305,
+      "step": 47774
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.269e-06,
+      "loss": 1.022,
+      "step": 47775
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2687314285714286e-06,
+      "loss": 1.2053,
+      "step": 47776
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.268462857142857e-06,
+      "loss": 1.0198,
+      "step": 47777
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2681942857142858e-06,
+      "loss": 1.1184,
+      "step": 47778
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2679257142857142e-06,
+      "loss": 1.281,
+      "step": 47779
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2676571428571428e-06,
+      "loss": 1.1834,
+      "step": 47780
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2673885714285716e-06,
+      "loss": 1.1173,
+      "step": 47781
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.26712e-06,
+      "loss": 0.8056,
+      "step": 47782
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2668514285714286e-06,
+      "loss": 1.3373,
+      "step": 47783
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2665828571428572e-06,
+      "loss": 1.1838,
+      "step": 47784
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2663142857142858e-06,
+      "loss": 0.8468,
+      "step": 47785
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2660457142857142e-06,
+      "loss": 1.083,
+      "step": 47786
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2657771428571428e-06,
+      "loss": 1.5349,
+      "step": 47787
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2655085714285716e-06,
+      "loss": 1.3524,
+      "step": 47788
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.26524e-06,
+      "loss": 0.8865,
+      "step": 47789
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2649714285714286e-06,
+      "loss": 1.2793,
+      "step": 47790
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2647028571428572e-06,
+      "loss": 1.134,
+      "step": 47791
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2644342857142858e-06,
+      "loss": 1.3209,
+      "step": 47792
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2641657142857142e-06,
+      "loss": 0.6385,
+      "step": 47793
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.263897142857143e-06,
+      "loss": 1.1971,
+      "step": 47794
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2636285714285716e-06,
+      "loss": 0.9886,
+      "step": 47795
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.26336e-06,
+      "loss": 1.0814,
+      "step": 47796
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2630914285714286e-06,
+      "loss": 1.0823,
+      "step": 47797
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2628228571428572e-06,
+      "loss": 0.9777,
+      "step": 47798
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2625542857142858e-06,
+      "loss": 1.0424,
+      "step": 47799
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2622857142857142e-06,
+      "loss": 1.2145,
+      "step": 47800
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.262017142857143e-06,
+      "loss": 1.1,
+      "step": 47801
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2617485714285714e-06,
+      "loss": 1.2166,
+      "step": 47802
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.26148e-06,
+      "loss": 1.176,
+      "step": 47803
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2612114285714288e-06,
+      "loss": 0.6626,
+      "step": 47804
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2609428571428572e-06,
+      "loss": 1.2584,
+      "step": 47805
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2606742857142858e-06,
+      "loss": 0.9319,
+      "step": 47806
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2604057142857142e-06,
+      "loss": 1.0871,
+      "step": 47807
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.260137142857143e-06,
+      "loss": 1.0917,
+      "step": 47808
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2598685714285714e-06,
+      "loss": 1.0814,
+      "step": 47809
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2596e-06,
+      "loss": 1.2134,
+      "step": 47810
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2593314285714288e-06,
+      "loss": 0.8962,
+      "step": 47811
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2590628571428572e-06,
+      "loss": 1.213,
+      "step": 47812
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2587942857142858e-06,
+      "loss": 1.3253,
+      "step": 47813
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2585257142857144e-06,
+      "loss": 1.0588,
+      "step": 47814
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.258257142857143e-06,
+      "loss": 1.0272,
+      "step": 47815
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2579885714285714e-06,
+      "loss": 0.8248,
+      "step": 47816
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2577200000000002e-06,
+      "loss": 0.6135,
+      "step": 47817
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2574514285714286e-06,
+      "loss": 0.9602,
+      "step": 47818
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2571828571428572e-06,
+      "loss": 0.8989,
+      "step": 47819
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2569142857142858e-06,
+      "loss": 0.7108,
+      "step": 47820
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2566457142857144e-06,
+      "loss": 0.9847,
+      "step": 47821
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.256377142857143e-06,
+      "loss": 1.0018,
+      "step": 47822
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2561085714285713e-06,
+      "loss": 0.9144,
+      "step": 47823
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2558400000000002e-06,
+      "loss": 0.8138,
+      "step": 47824
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2555714285714285e-06,
+      "loss": 1.5199,
+      "step": 47825
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2553028571428571e-06,
+      "loss": 1.0615,
+      "step": 47826
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.255034285714286e-06,
+      "loss": 1.4068,
+      "step": 47827
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2547657142857143e-06,
+      "loss": 1.4047,
+      "step": 47828
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.254497142857143e-06,
+      "loss": 1.6263,
+      "step": 47829
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2542285714285713e-06,
+      "loss": 1.0412,
+      "step": 47830
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2539600000000001e-06,
+      "loss": 1.2368,
+      "step": 47831
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2536914285714285e-06,
+      "loss": 0.9386,
+      "step": 47832
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2534228571428571e-06,
+      "loss": 1.4012,
+      "step": 47833
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2531542857142857e-06,
+      "loss": 1.2544,
+      "step": 47834
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2528857142857143e-06,
+      "loss": 0.8601,
+      "step": 47835
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.252617142857143e-06,
+      "loss": 1.4673,
+      "step": 47836
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2523485714285715e-06,
+      "loss": 1.3927,
+      "step": 47837
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2520800000000001e-06,
+      "loss": 0.9553,
+      "step": 47838
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2518114285714285e-06,
+      "loss": 0.8139,
+      "step": 47839
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2515428571428571e-06,
+      "loss": 1.1914,
+      "step": 47840
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2512742857142857e-06,
+      "loss": 1.2453,
+      "step": 47841
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2510057142857143e-06,
+      "loss": 1.1229,
+      "step": 47842
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.250737142857143e-06,
+      "loss": 0.9368,
+      "step": 47843
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2504685714285715e-06,
+      "loss": 1.1995,
+      "step": 47844
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2502000000000001e-06,
+      "loss": 1.6353,
+      "step": 47845
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2499314285714285e-06,
+      "loss": 0.9476,
+      "step": 47846
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2496628571428573e-06,
+      "loss": 1.3561,
+      "step": 47847
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2493942857142857e-06,
+      "loss": 0.9047,
+      "step": 47848
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2491257142857143e-06,
+      "loss": 1.0722,
+      "step": 47849
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.248857142857143e-06,
+      "loss": 1.0294,
+      "step": 47850
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2485885714285715e-06,
+      "loss": 1.1547,
+      "step": 47851
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.24832e-06,
+      "loss": 1.2013,
+      "step": 47852
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2480514285714285e-06,
+      "loss": 1.3383,
+      "step": 47853
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2477828571428573e-06,
+      "loss": 1.1663,
+      "step": 47854
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2475142857142857e-06,
+      "loss": 1.4328,
+      "step": 47855
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2472457142857143e-06,
+      "loss": 1.0257,
+      "step": 47856
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2469771428571429e-06,
+      "loss": 0.9119,
+      "step": 47857
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2467085714285715e-06,
+      "loss": 0.6108,
+      "step": 47858
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.24644e-06,
+      "loss": 1.1895,
+      "step": 47859
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2461714285714287e-06,
+      "loss": 1.1763,
+      "step": 47860
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2459028571428573e-06,
+      "loss": 1.0108,
+      "step": 47861
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.2456342857142857e-06,
+      "loss": 0.972,
+      "step": 47862
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2453657142857143e-06,
+      "loss": 0.9629,
+      "step": 47863
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2450971428571429e-06,
+      "loss": 0.9008,
+      "step": 47864
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2448285714285715e-06,
+      "loss": 1.1333,
+      "step": 47865
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2445599999999999e-06,
+      "loss": 1.2904,
+      "step": 47866
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2442914285714287e-06,
+      "loss": 1.3148,
+      "step": 47867
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2440228571428573e-06,
+      "loss": 1.3113,
+      "step": 47868
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2437542857142857e-06,
+      "loss": 1.6745,
+      "step": 47869
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2434857142857145e-06,
+      "loss": 0.8202,
+      "step": 47870
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2432171428571429e-06,
+      "loss": 1.5225,
+      "step": 47871
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2429485714285715e-06,
+      "loss": 0.875,
+      "step": 47872
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.24268e-06,
+      "loss": 1.0109,
+      "step": 47873
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2424114285714287e-06,
+      "loss": 0.8737,
+      "step": 47874
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2421428571428573e-06,
+      "loss": 1.2177,
+      "step": 47875
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2418742857142856e-06,
+      "loss": 1.0473,
+      "step": 47876
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2416057142857145e-06,
+      "loss": 1.0952,
+      "step": 47877
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2413371428571428e-06,
+      "loss": 1.0011,
+      "step": 47878
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2410685714285714e-06,
+      "loss": 0.5908,
+      "step": 47879
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2408e-06,
+      "loss": 0.8407,
+      "step": 47880
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2405314285714286e-06,
+      "loss": 1.6935,
+      "step": 47881
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.240262857142857e-06,
+      "loss": 1.2453,
+      "step": 47882
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2399942857142858e-06,
+      "loss": 0.7719,
+      "step": 47883
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2397257142857144e-06,
+      "loss": 0.9509,
+      "step": 47884
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2394571428571428e-06,
+      "loss": 1.1509,
+      "step": 47885
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2391885714285714e-06,
+      "loss": 1.1124,
+      "step": 47886
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.23892e-06,
+      "loss": 1.1563,
+      "step": 47887
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2386514285714286e-06,
+      "loss": 1.3546,
+      "step": 47888
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.238382857142857e-06,
+      "loss": 1.2776,
+      "step": 47889
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2381142857142858e-06,
+      "loss": 0.9566,
+      "step": 47890
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2378457142857144e-06,
+      "loss": 1.2019,
+      "step": 47891
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2375771428571428e-06,
+      "loss": 0.9301,
+      "step": 47892
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2373085714285716e-06,
+      "loss": 1.508,
+      "step": 47893
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.23704e-06,
+      "loss": 1.0981,
+      "step": 47894
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2367714285714286e-06,
+      "loss": 1.0043,
+      "step": 47895
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2365028571428572e-06,
+      "loss": 1.4693,
+      "step": 47896
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2362342857142858e-06,
+      "loss": 1.557,
+      "step": 47897
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2359657142857144e-06,
+      "loss": 1.1196,
+      "step": 47898
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2356971428571428e-06,
+      "loss": 0.8516,
+      "step": 47899
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2354285714285716e-06,
+      "loss": 1.2837,
+      "step": 47900
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.23516e-06,
+      "loss": 1.2776,
+      "step": 47901
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2348914285714286e-06,
+      "loss": 1.1694,
+      "step": 47902
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2346228571428572e-06,
+      "loss": 1.042,
+      "step": 47903
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2343542857142858e-06,
+      "loss": 0.9305,
+      "step": 47904
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2340857142857142e-06,
+      "loss": 0.9744,
+      "step": 47905
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.233817142857143e-06,
+      "loss": 1.0688,
+      "step": 47906
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2335485714285716e-06,
+      "loss": 1.3258,
+      "step": 47907
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.23328e-06,
+      "loss": 1.0795,
+      "step": 47908
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2330114285714286e-06,
+      "loss": 1.1332,
+      "step": 47909
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2327428571428572e-06,
+      "loss": 0.8232,
+      "step": 47910
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2324742857142858e-06,
+      "loss": 1.2434,
+      "step": 47911
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2322057142857142e-06,
+      "loss": 1.2974,
+      "step": 47912
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.231937142857143e-06,
+      "loss": 1.2411,
+      "step": 47913
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2316685714285716e-06,
+      "loss": 1.199,
+      "step": 47914
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2314e-06,
+      "loss": 1.1776,
+      "step": 47915
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2311314285714288e-06,
+      "loss": 1.0513,
+      "step": 47916
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2308628571428572e-06,
+      "loss": 1.4507,
+      "step": 47917
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2305942857142858e-06,
+      "loss": 0.6856,
+      "step": 47918
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2303257142857142e-06,
+      "loss": 1.0319,
+      "step": 47919
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.230057142857143e-06,
+      "loss": 0.9228,
+      "step": 47920
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2297885714285714e-06,
+      "loss": 1.4093,
+      "step": 47921
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.22952e-06,
+      "loss": 1.4752,
+      "step": 47922
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2292514285714288e-06,
+      "loss": 1.0953,
+      "step": 47923
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2289828571428572e-06,
+      "loss": 0.7689,
+      "step": 47924
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2287142857142858e-06,
+      "loss": 1.0805,
+      "step": 47925
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2284457142857144e-06,
+      "loss": 1.1667,
+      "step": 47926
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.228177142857143e-06,
+      "loss": 1.281,
+      "step": 47927
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2279085714285713e-06,
+      "loss": 1.2148,
+      "step": 47928
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2276400000000002e-06,
+      "loss": 0.8984,
+      "step": 47929
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2273714285714288e-06,
+      "loss": 1.3956,
+      "step": 47930
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2271028571428571e-06,
+      "loss": 0.9685,
+      "step": 47931
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2268342857142857e-06,
+      "loss": 0.8532,
+      "step": 47932
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2265657142857143e-06,
+      "loss": 1.356,
+      "step": 47933
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.226297142857143e-06,
+      "loss": 1.2665,
+      "step": 47934
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2260285714285713e-06,
+      "loss": 1.3067,
+      "step": 47935
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2257600000000001e-06,
+      "loss": 1.0906,
+      "step": 47936
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2254914285714285e-06,
+      "loss": 0.799,
+      "step": 47937
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2252228571428571e-06,
+      "loss": 1.2887,
+      "step": 47938
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.224954285714286e-06,
+      "loss": 0.9358,
+      "step": 47939
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2246857142857143e-06,
+      "loss": 1.0248,
+      "step": 47940
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.224417142857143e-06,
+      "loss": 0.9565,
+      "step": 47941
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2241485714285713e-06,
+      "loss": 0.5846,
+      "step": 47942
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2238800000000001e-06,
+      "loss": 1.1339,
+      "step": 47943
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2236114285714285e-06,
+      "loss": 0.9456,
+      "step": 47944
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2233428571428571e-06,
+      "loss": 1.2327,
+      "step": 47945
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.223074285714286e-06,
+      "loss": 0.731,
+      "step": 47946
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2228057142857143e-06,
+      "loss": 1.2224,
+      "step": 47947
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.222537142857143e-06,
+      "loss": 0.852,
+      "step": 47948
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2222685714285715e-06,
+      "loss": 1.4635,
+      "step": 47949
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2220000000000001e-06,
+      "loss": 1.1118,
+      "step": 47950
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2217314285714285e-06,
+      "loss": 0.8295,
+      "step": 47951
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.221462857142857e-06,
+      "loss": 1.1105,
+      "step": 47952
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2211942857142857e-06,
+      "loss": 1.0276,
+      "step": 47953
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2209257142857143e-06,
+      "loss": 1.2604,
+      "step": 47954
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.220657142857143e-06,
+      "loss": 1.3462,
+      "step": 47955
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2203885714285715e-06,
+      "loss": 1.1324,
+      "step": 47956
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.22012e-06,
+      "loss": 1.1743,
+      "step": 47957
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2198514285714285e-06,
+      "loss": 0.5749,
+      "step": 47958
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2195828571428573e-06,
+      "loss": 1.1077,
+      "step": 47959
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2193142857142857e-06,
+      "loss": 1.1051,
+      "step": 47960
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2190457142857143e-06,
+      "loss": 1.1077,
+      "step": 47961
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2187771428571429e-06,
+      "loss": 1.5137,
+      "step": 47962
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2185085714285715e-06,
+      "loss": 0.7646,
+      "step": 47963
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.21824e-06,
+      "loss": 1.0676,
+      "step": 47964
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2179714285714285e-06,
+      "loss": 1.0329,
+      "step": 47965
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2177028571428573e-06,
+      "loss": 0.8268,
+      "step": 47966
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2174342857142857e-06,
+      "loss": 1.107,
+      "step": 47967
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2171657142857143e-06,
+      "loss": 0.9189,
+      "step": 47968
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2168971428571429e-06,
+      "loss": 1.2265,
+      "step": 47969
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2166285714285715e-06,
+      "loss": 1.0378,
+      "step": 47970
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.21636e-06,
+      "loss": 1.0496,
+      "step": 47971
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2160914285714287e-06,
+      "loss": 1.2338,
+      "step": 47972
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2158228571428573e-06,
+      "loss": 1.1329,
+      "step": 47973
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2155542857142857e-06,
+      "loss": 1.0011,
+      "step": 47974
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2152857142857143e-06,
+      "loss": 1.4283,
+      "step": 47975
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2150171428571429e-06,
+      "loss": 0.9656,
+      "step": 47976
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2147485714285715e-06,
+      "loss": 1.138,
+      "step": 47977
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.21448e-06,
+      "loss": 0.7824,
+      "step": 47978
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2142114285714287e-06,
+      "loss": 1.5026,
+      "step": 47979
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2139428571428573e-06,
+      "loss": 1.4673,
+      "step": 47980
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2136742857142856e-06,
+      "loss": 0.8636,
+      "step": 47981
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2134057142857145e-06,
+      "loss": 1.0862,
+      "step": 47982
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2131371428571428e-06,
+      "loss": 1.1478,
+      "step": 47983
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2128685714285714e-06,
+      "loss": 1.3197,
+      "step": 47984
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2126e-06,
+      "loss": 1.1621,
+      "step": 47985
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2123314285714286e-06,
+      "loss": 0.9293,
+      "step": 47986
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2120628571428572e-06,
+      "loss": 1.0577,
+      "step": 47987
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2117942857142856e-06,
+      "loss": 1.0907,
+      "step": 47988
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2115257142857144e-06,
+      "loss": 1.4821,
+      "step": 47989
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2112571428571428e-06,
+      "loss": 1.3599,
+      "step": 47990
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2109885714285714e-06,
+      "loss": 1.1031,
+      "step": 47991
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.21072e-06,
+      "loss": 1.1756,
+      "step": 47992
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2104514285714286e-06,
+      "loss": 0.9535,
+      "step": 47993
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2101828571428572e-06,
+      "loss": 0.8622,
+      "step": 47994
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2099142857142858e-06,
+      "loss": 1.457,
+      "step": 47995
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2096457142857144e-06,
+      "loss": 1.3222,
+      "step": 47996
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2093771428571428e-06,
+      "loss": 0.8656,
+      "step": 47997
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2091085714285714e-06,
+      "loss": 1.2479,
+      "step": 47998
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.20884e-06,
+      "loss": 0.9864,
+      "step": 47999
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2085714285714286e-06,
+      "loss": 1.4256,
+      "step": 48000
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2083028571428572e-06,
+      "loss": 1.3459,
+      "step": 48001
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2080342857142858e-06,
+      "loss": 0.8914,
+      "step": 48002
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2077657142857144e-06,
+      "loss": 0.901,
+      "step": 48003
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2074971428571428e-06,
+      "loss": 1.0059,
+      "step": 48004
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2072285714285716e-06,
+      "loss": 1.1308,
+      "step": 48005
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.20696e-06,
+      "loss": 1.3447,
+      "step": 48006
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2066914285714286e-06,
+      "loss": 0.8548,
+      "step": 48007
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2064228571428572e-06,
+      "loss": 1.1082,
+      "step": 48008
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2061542857142858e-06,
+      "loss": 1.1666,
+      "step": 48009
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2058857142857144e-06,
+      "loss": 1.3795,
+      "step": 48010
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2056171428571428e-06,
+      "loss": 1.3198,
+      "step": 48011
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2053485714285716e-06,
+      "loss": 0.7479,
+      "step": 48012
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.20508e-06,
+      "loss": 1.2834,
+      "step": 48013
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2048114285714286e-06,
+      "loss": 0.9044,
+      "step": 48014
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2045428571428572e-06,
+      "loss": 0.9931,
+      "step": 48015
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2042742857142858e-06,
+      "loss": 1.3462,
+      "step": 48016
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2040057142857144e-06,
+      "loss": 0.7627,
+      "step": 48017
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.203737142857143e-06,
+      "loss": 1.0289,
+      "step": 48018
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2034685714285716e-06,
+      "loss": 1.2491,
+      "step": 48019
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2032e-06,
+      "loss": 1.3887,
+      "step": 48020
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2029314285714286e-06,
+      "loss": 1.2182,
+      "step": 48021
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2026628571428572e-06,
+      "loss": 1.8157,
+      "step": 48022
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2023942857142858e-06,
+      "loss": 0.9775,
+      "step": 48023
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2021257142857142e-06,
+      "loss": 1.2125,
+      "step": 48024
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.201857142857143e-06,
+      "loss": 1.6475,
+      "step": 48025
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2015885714285716e-06,
+      "loss": 1.1152,
+      "step": 48026
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.20132e-06,
+      "loss": 1.4022,
+      "step": 48027
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2010514285714288e-06,
+      "loss": 0.8435,
+      "step": 48028
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2007828571428572e-06,
+      "loss": 1.0306,
+      "step": 48029
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2005142857142858e-06,
+      "loss": 1.2629,
+      "step": 48030
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.2002457142857141e-06,
+      "loss": 1.3356,
+      "step": 48031
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.199977142857143e-06,
+      "loss": 1.2067,
+      "step": 48032
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.1997085714285716e-06,
+      "loss": 0.8855,
+      "step": 48033
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.19944e-06,
+      "loss": 1.5101,
+      "step": 48034
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.1991714285714288e-06,
+      "loss": 1.2816,
+      "step": 48035
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.1989028571428571e-06,
+      "loss": 0.8798,
+      "step": 48036
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.1986342857142857e-06,
+      "loss": 1.2665,
+      "step": 48037
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1983657142857143e-06,
+      "loss": 1.5454,
+      "step": 48038
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.198097142857143e-06,
+      "loss": 0.7926,
+      "step": 48039
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1978285714285713e-06,
+      "loss": 1.119,
+      "step": 48040
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1975600000000001e-06,
+      "loss": 1.1098,
+      "step": 48041
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1972914285714287e-06,
+      "loss": 1.0387,
+      "step": 48042
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1970228571428571e-06,
+      "loss": 1.2583,
+      "step": 48043
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1967542857142857e-06,
+      "loss": 1.4021,
+      "step": 48044
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1964857142857143e-06,
+      "loss": 1.334,
+      "step": 48045
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.196217142857143e-06,
+      "loss": 0.6237,
+      "step": 48046
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1959485714285713e-06,
+      "loss": 0.888,
+      "step": 48047
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1956800000000001e-06,
+      "loss": 0.8413,
+      "step": 48048
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1954114285714287e-06,
+      "loss": 0.8753,
+      "step": 48049
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1951428571428571e-06,
+      "loss": 0.9664,
+      "step": 48050
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.194874285714286e-06,
+      "loss": 0.8282,
+      "step": 48051
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1946057142857143e-06,
+      "loss": 1.3736,
+      "step": 48052
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.194337142857143e-06,
+      "loss": 1.2584,
+      "step": 48053
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1940685714285713e-06,
+      "loss": 1.2416,
+      "step": 48054
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1938000000000001e-06,
+      "loss": 1.271,
+      "step": 48055
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1935314285714285e-06,
+      "loss": 1.4205,
+      "step": 48056
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.193262857142857e-06,
+      "loss": 1.2133,
+      "step": 48057
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.192994285714286e-06,
+      "loss": 1.0508,
+      "step": 48058
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1927257142857143e-06,
+      "loss": 1.1775,
+      "step": 48059
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.192457142857143e-06,
+      "loss": 1.5885,
+      "step": 48060
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1921885714285715e-06,
+      "loss": 1.1788,
+      "step": 48061
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.19192e-06,
+      "loss": 1.2966,
+      "step": 48062
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1916514285714285e-06,
+      "loss": 0.9584,
+      "step": 48063
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.191382857142857e-06,
+      "loss": 1.0321,
+      "step": 48064
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.191114285714286e-06,
+      "loss": 1.5106,
+      "step": 48065
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1908457142857143e-06,
+      "loss": 1.2912,
+      "step": 48066
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1905771428571429e-06,
+      "loss": 1.3662,
+      "step": 48067
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1903085714285715e-06,
+      "loss": 0.9208,
+      "step": 48068
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.19004e-06,
+      "loss": 1.0786,
+      "step": 48069
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1897714285714285e-06,
+      "loss": 1.3095,
+      "step": 48070
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1895028571428573e-06,
+      "loss": 1.3317,
+      "step": 48071
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1892342857142857e-06,
+      "loss": 1.1371,
+      "step": 48072
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1889657142857143e-06,
+      "loss": 1.2224,
+      "step": 48073
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1886971428571429e-06,
+      "loss": 1.1843,
+      "step": 48074
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1884285714285715e-06,
+      "loss": 1.1717,
+      "step": 48075
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.18816e-06,
+      "loss": 1.3085,
+      "step": 48076
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1878914285714285e-06,
+      "loss": 0.8501,
+      "step": 48077
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1876228571428573e-06,
+      "loss": 1.0072,
+      "step": 48078
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1873542857142857e-06,
+      "loss": 1.4492,
+      "step": 48079
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1870857142857143e-06,
+      "loss": 0.811,
+      "step": 48080
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.186817142857143e-06,
+      "loss": 1.088,
+      "step": 48081
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1865485714285715e-06,
+      "loss": 1.0258,
+      "step": 48082
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.18628e-06,
+      "loss": 1.0984,
+      "step": 48083
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1860114285714287e-06,
+      "loss": 1.0556,
+      "step": 48084
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1857428571428573e-06,
+      "loss": 1.1372,
+      "step": 48085
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1854742857142856e-06,
+      "loss": 1.0809,
+      "step": 48086
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1852057142857142e-06,
+      "loss": 1.4073,
+      "step": 48087
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1849371428571428e-06,
+      "loss": 1.2306,
+      "step": 48088
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1846685714285714e-06,
+      "loss": 0.8117,
+      "step": 48089
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1844e-06,
+      "loss": 1.2894,
+      "step": 48090
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1841314285714286e-06,
+      "loss": 1.1216,
+      "step": 48091
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1838628571428572e-06,
+      "loss": 1.1667,
+      "step": 48092
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1835942857142856e-06,
+      "loss": 1.0875,
+      "step": 48093
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1833257142857144e-06,
+      "loss": 1.3814,
+      "step": 48094
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1830571428571428e-06,
+      "loss": 1.1421,
+      "step": 48095
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1827885714285714e-06,
+      "loss": 1.04,
+      "step": 48096
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.18252e-06,
+      "loss": 0.842,
+      "step": 48097
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1822514285714286e-06,
+      "loss": 1.4421,
+      "step": 48098
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1819828571428572e-06,
+      "loss": 0.7747,
+      "step": 48099
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1817142857142856e-06,
+      "loss": 1.2629,
+      "step": 48100
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1814457142857144e-06,
+      "loss": 1.275,
+      "step": 48101
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1811771428571428e-06,
+      "loss": 1.035,
+      "step": 48102
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1809085714285714e-06,
+      "loss": 0.9636,
+      "step": 48103
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1806400000000002e-06,
+      "loss": 1.2576,
+      "step": 48104
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1803714285714286e-06,
+      "loss": 1.0174,
+      "step": 48105
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1801028571428572e-06,
+      "loss": 0.9321,
+      "step": 48106
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1798342857142858e-06,
+      "loss": 1.0443,
+      "step": 48107
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1795657142857144e-06,
+      "loss": 0.9794,
+      "step": 48108
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1792971428571428e-06,
+      "loss": 1.032,
+      "step": 48109
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1790285714285714e-06,
+      "loss": 1.1672,
+      "step": 48110
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.17876e-06,
+      "loss": 0.7347,
+      "step": 48111
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1784914285714286e-06,
+      "loss": 1.0428,
+      "step": 48112
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1782228571428572e-06,
+      "loss": 1.3659,
+      "step": 48113
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1779542857142858e-06,
+      "loss": 1.092,
+      "step": 48114
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1776857142857144e-06,
+      "loss": 0.9689,
+      "step": 48115
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1774171428571428e-06,
+      "loss": 1.3849,
+      "step": 48116
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1771485714285716e-06,
+      "loss": 1.4029,
+      "step": 48117
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.17688e-06,
+      "loss": 0.9874,
+      "step": 48118
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1766114285714286e-06,
+      "loss": 1.1942,
+      "step": 48119
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1763428571428572e-06,
+      "loss": 0.9547,
+      "step": 48120
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1760742857142858e-06,
+      "loss": 1.096,
+      "step": 48121
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1758057142857144e-06,
+      "loss": 1.4027,
+      "step": 48122
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1755371428571428e-06,
+      "loss": 1.1951,
+      "step": 48123
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1752685714285716e-06,
+      "loss": 0.8933,
+      "step": 48124
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.175e-06,
+      "loss": 1.2596,
+      "step": 48125
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1747314285714286e-06,
+      "loss": 1.1142,
+      "step": 48126
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1744628571428572e-06,
+      "loss": 1.0135,
+      "step": 48127
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1741942857142858e-06,
+      "loss": 0.9411,
+      "step": 48128
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1739257142857144e-06,
+      "loss": 1.1397,
+      "step": 48129
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.173657142857143e-06,
+      "loss": 1.035,
+      "step": 48130
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1733885714285716e-06,
+      "loss": 1.54,
+      "step": 48131
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.17312e-06,
+      "loss": 1.1336,
+      "step": 48132
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1728514285714286e-06,
+      "loss": 0.856,
+      "step": 48133
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1725828571428572e-06,
+      "loss": 1.2741,
+      "step": 48134
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1723142857142858e-06,
+      "loss": 1.3396,
+      "step": 48135
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1720457142857144e-06,
+      "loss": 0.8715,
+      "step": 48136
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.171777142857143e-06,
+      "loss": 0.7436,
+      "step": 48137
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1715085714285716e-06,
+      "loss": 1.1232,
+      "step": 48138
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.17124e-06,
+      "loss": 1.3528,
+      "step": 48139
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1709714285714288e-06,
+      "loss": 0.8737,
+      "step": 48140
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1707028571428571e-06,
+      "loss": 1.1905,
+      "step": 48141
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1704342857142857e-06,
+      "loss": 1.1379,
+      "step": 48142
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1701657142857141e-06,
+      "loss": 0.763,
+      "step": 48143
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.169897142857143e-06,
+      "loss": 1.2444,
+      "step": 48144
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1696285714285715e-06,
+      "loss": 1.6029,
+      "step": 48145
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.16936e-06,
+      "loss": 0.7502,
+      "step": 48146
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1690914285714287e-06,
+      "loss": 1.0042,
+      "step": 48147
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1688228571428571e-06,
+      "loss": 0.8912,
+      "step": 48148
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1685542857142857e-06,
+      "loss": 0.9509,
+      "step": 48149
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1682857142857143e-06,
+      "loss": 0.9167,
+      "step": 48150
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.168017142857143e-06,
+      "loss": 0.9215,
+      "step": 48151
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1677485714285715e-06,
+      "loss": 1.3037,
+      "step": 48152
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1674800000000001e-06,
+      "loss": 1.0488,
+      "step": 48153
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1672114285714287e-06,
+      "loss": 1.1731,
+      "step": 48154
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1669428571428571e-06,
+      "loss": 0.8615,
+      "step": 48155
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1666742857142857e-06,
+      "loss": 1.4402,
+      "step": 48156
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1664057142857143e-06,
+      "loss": 1.2019,
+      "step": 48157
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.166137142857143e-06,
+      "loss": 1.3084,
+      "step": 48158
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1658685714285713e-06,
+      "loss": 1.1884,
+      "step": 48159
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1656000000000001e-06,
+      "loss": 0.9107,
+      "step": 48160
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1653314285714287e-06,
+      "loss": 1.4865,
+      "step": 48161
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.165062857142857e-06,
+      "loss": 1.1704,
+      "step": 48162
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.164794285714286e-06,
+      "loss": 0.8755,
+      "step": 48163
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1645257142857143e-06,
+      "loss": 1.1472,
+      "step": 48164
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.164257142857143e-06,
+      "loss": 0.9977,
+      "step": 48165
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1639885714285713e-06,
+      "loss": 1.1013,
+      "step": 48166
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.16372e-06,
+      "loss": 1.1955,
+      "step": 48167
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1634514285714287e-06,
+      "loss": 0.9846,
+      "step": 48168
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.163182857142857e-06,
+      "loss": 0.7027,
+      "step": 48169
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.162914285714286e-06,
+      "loss": 1.6296,
+      "step": 48170
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1626457142857143e-06,
+      "loss": 0.953,
+      "step": 48171
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1623771428571429e-06,
+      "loss": 1.6962,
+      "step": 48172
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1621085714285715e-06,
+      "loss": 0.7738,
+      "step": 48173
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.16184e-06,
+      "loss": 1.0116,
+      "step": 48174
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1615714285714285e-06,
+      "loss": 0.8772,
+      "step": 48175
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.161302857142857e-06,
+      "loss": 0.7518,
+      "step": 48176
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1610342857142859e-06,
+      "loss": 1.2487,
+      "step": 48177
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1607657142857143e-06,
+      "loss": 0.8587,
+      "step": 48178
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1604971428571429e-06,
+      "loss": 0.9915,
+      "step": 48179
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1602285714285715e-06,
+      "loss": 0.9272,
+      "step": 48180
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.15996e-06,
+      "loss": 1.0001,
+      "step": 48181
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1596914285714285e-06,
+      "loss": 1.23,
+      "step": 48182
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1594228571428573e-06,
+      "loss": 0.9321,
+      "step": 48183
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1591542857142859e-06,
+      "loss": 0.7445,
+      "step": 48184
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1588857142857143e-06,
+      "loss": 0.9696,
+      "step": 48185
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1586171428571429e-06,
+      "loss": 1.0942,
+      "step": 48186
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1583485714285715e-06,
+      "loss": 1.2597,
+      "step": 48187
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.15808e-06,
+      "loss": 1.2534,
+      "step": 48188
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1578114285714284e-06,
+      "loss": 1.4116,
+      "step": 48189
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1575428571428573e-06,
+      "loss": 0.7477,
+      "step": 48190
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1572742857142856e-06,
+      "loss": 0.7104,
+      "step": 48191
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1570057142857142e-06,
+      "loss": 1.3219,
+      "step": 48192
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.156737142857143e-06,
+      "loss": 1.5296,
+      "step": 48193
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1564685714285714e-06,
+      "loss": 1.4704,
+      "step": 48194
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1562e-06,
+      "loss": 1.4263,
+      "step": 48195
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1559314285714286e-06,
+      "loss": 1.2325,
+      "step": 48196
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1556628571428572e-06,
+      "loss": 0.9863,
+      "step": 48197
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1553942857142856e-06,
+      "loss": 1.2692,
+      "step": 48198
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1551257142857142e-06,
+      "loss": 1.0782,
+      "step": 48199
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.154857142857143e-06,
+      "loss": 0.973,
+      "step": 48200
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1545885714285714e-06,
+      "loss": 1.0049,
+      "step": 48201
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.15432e-06,
+      "loss": 0.9165,
+      "step": 48202
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1540514285714286e-06,
+      "loss": 0.9497,
+      "step": 48203
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1537828571428572e-06,
+      "loss": 1.1637,
+      "step": 48204
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1535142857142856e-06,
+      "loss": 0.8619,
+      "step": 48205
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1532457142857144e-06,
+      "loss": 0.962,
+      "step": 48206
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.152977142857143e-06,
+      "loss": 1.2467,
+      "step": 48207
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1527085714285714e-06,
+      "loss": 1.3992,
+      "step": 48208
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.15244e-06,
+      "loss": 1.1143,
+      "step": 48209
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1521714285714286e-06,
+      "loss": 1.1477,
+      "step": 48210
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1519028571428572e-06,
+      "loss": 1.2781,
+      "step": 48211
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.1516342857142856e-06,
+      "loss": 1.5956,
+      "step": 48212
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1513657142857144e-06,
+      "loss": 1.6075,
+      "step": 48213
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1510971428571428e-06,
+      "loss": 0.8127,
+      "step": 48214
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1508285714285714e-06,
+      "loss": 1.7374,
+      "step": 48215
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1505600000000002e-06,
+      "loss": 1.5016,
+      "step": 48216
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1502914285714286e-06,
+      "loss": 1.563,
+      "step": 48217
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1500228571428572e-06,
+      "loss": 1.0476,
+      "step": 48218
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1497542857142858e-06,
+      "loss": 1.1717,
+      "step": 48219
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1494857142857144e-06,
+      "loss": 0.9654,
+      "step": 48220
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1492171428571428e-06,
+      "loss": 1.2968,
+      "step": 48221
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1489485714285714e-06,
+      "loss": 1.2928,
+      "step": 48222
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1486800000000002e-06,
+      "loss": 1.1618,
+      "step": 48223
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1484114285714286e-06,
+      "loss": 0.8838,
+      "step": 48224
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1481428571428572e-06,
+      "loss": 1.4356,
+      "step": 48225
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1478742857142858e-06,
+      "loss": 0.9146,
+      "step": 48226
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1476057142857144e-06,
+      "loss": 0.9232,
+      "step": 48227
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1473371428571428e-06,
+      "loss": 1.0598,
+      "step": 48228
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1470685714285716e-06,
+      "loss": 1.2887,
+      "step": 48229
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1468e-06,
+      "loss": 1.4972,
+      "step": 48230
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1465314285714286e-06,
+      "loss": 1.2373,
+      "step": 48231
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1462628571428572e-06,
+      "loss": 1.0544,
+      "step": 48232
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1459942857142858e-06,
+      "loss": 1.2391,
+      "step": 48233
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1457257142857144e-06,
+      "loss": 1.0473,
+      "step": 48234
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1454571428571428e-06,
+      "loss": 1.5808,
+      "step": 48235
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1451885714285716e-06,
+      "loss": 0.9223,
+      "step": 48236
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.14492e-06,
+      "loss": 1.1134,
+      "step": 48237
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1446514285714286e-06,
+      "loss": 1.0815,
+      "step": 48238
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1443828571428574e-06,
+      "loss": 1.0265,
+      "step": 48239
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1441142857142858e-06,
+      "loss": 1.3952,
+      "step": 48240
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1438457142857144e-06,
+      "loss": 1.1913,
+      "step": 48241
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.143577142857143e-06,
+      "loss": 1.4433,
+      "step": 48242
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1433085714285716e-06,
+      "loss": 1.4276,
+      "step": 48243
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.14304e-06,
+      "loss": 1.2126,
+      "step": 48244
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1427714285714285e-06,
+      "loss": 1.1614,
+      "step": 48245
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1425028571428571e-06,
+      "loss": 1.1403,
+      "step": 48246
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1422342857142857e-06,
+      "loss": 1.5559,
+      "step": 48247
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1419657142857143e-06,
+      "loss": 1.0104,
+      "step": 48248
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.141697142857143e-06,
+      "loss": 1.0803,
+      "step": 48249
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1414285714285715e-06,
+      "loss": 1.1927,
+      "step": 48250
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.14116e-06,
+      "loss": 1.1512,
+      "step": 48251
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1408914285714287e-06,
+      "loss": 1.1714,
+      "step": 48252
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1406228571428571e-06,
+      "loss": 1.2332,
+      "step": 48253
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1403542857142857e-06,
+      "loss": 0.8432,
+      "step": 48254
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1400857142857143e-06,
+      "loss": 1.3741,
+      "step": 48255
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.139817142857143e-06,
+      "loss": 1.0963,
+      "step": 48256
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1395485714285715e-06,
+      "loss": 0.8846,
+      "step": 48257
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.13928e-06,
+      "loss": 0.8322,
+      "step": 48258
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1390114285714287e-06,
+      "loss": 1.0298,
+      "step": 48259
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1387428571428571e-06,
+      "loss": 1.282,
+      "step": 48260
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1384742857142857e-06,
+      "loss": 1.1414,
+      "step": 48261
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1382057142857143e-06,
+      "loss": 0.6976,
+      "step": 48262
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.137937142857143e-06,
+      "loss": 0.8927,
+      "step": 48263
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1376685714285715e-06,
+      "loss": 0.8322,
+      "step": 48264
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1374000000000001e-06,
+      "loss": 1.0652,
+      "step": 48265
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1371314285714287e-06,
+      "loss": 1.1306,
+      "step": 48266
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.136862857142857e-06,
+      "loss": 1.3118,
+      "step": 48267
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1365942857142857e-06,
+      "loss": 1.2115,
+      "step": 48268
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1363257142857143e-06,
+      "loss": 1.1758,
+      "step": 48269
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.136057142857143e-06,
+      "loss": 1.352,
+      "step": 48270
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1357885714285715e-06,
+      "loss": 0.9165,
+      "step": 48271
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.13552e-06,
+      "loss": 0.6691,
+      "step": 48272
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1352514285714287e-06,
+      "loss": 1.445,
+      "step": 48273
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.134982857142857e-06,
+      "loss": 0.9736,
+      "step": 48274
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.134714285714286e-06,
+      "loss": 1.1825,
+      "step": 48275
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1344457142857143e-06,
+      "loss": 1.0488,
+      "step": 48276
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1341771428571429e-06,
+      "loss": 1.0523,
+      "step": 48277
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1339085714285713e-06,
+      "loss": 0.8181,
+      "step": 48278
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.13364e-06,
+      "loss": 1.1525,
+      "step": 48279
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1333714285714287e-06,
+      "loss": 1.6299,
+      "step": 48280
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.133102857142857e-06,
+      "loss": 1.5234,
+      "step": 48281
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1328342857142859e-06,
+      "loss": 1.1473,
+      "step": 48282
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1325657142857143e-06,
+      "loss": 0.7781,
+      "step": 48283
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1322971428571429e-06,
+      "loss": 0.8336,
+      "step": 48284
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1320285714285715e-06,
+      "loss": 1.4552,
+      "step": 48285
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.13176e-06,
+      "loss": 1.2569,
+      "step": 48286
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1314914285714287e-06,
+      "loss": 1.4798,
+      "step": 48287
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.131222857142857e-06,
+      "loss": 1.4646,
+      "step": 48288
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1309542857142859e-06,
+      "loss": 1.0501,
+      "step": 48289
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1306857142857143e-06,
+      "loss": 1.1769,
+      "step": 48290
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1304171428571429e-06,
+      "loss": 1.4073,
+      "step": 48291
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1301485714285715e-06,
+      "loss": 1.2829,
+      "step": 48292
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.12988e-06,
+      "loss": 1.2147,
+      "step": 48293
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1296114285714287e-06,
+      "loss": 0.9776,
+      "step": 48294
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1293428571428573e-06,
+      "loss": 0.8552,
+      "step": 48295
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1290742857142859e-06,
+      "loss": 1.003,
+      "step": 48296
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1288057142857142e-06,
+      "loss": 0.922,
+      "step": 48297
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1285371428571428e-06,
+      "loss": 0.859,
+      "step": 48298
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1282685714285714e-06,
+      "loss": 1.2744,
+      "step": 48299
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.128e-06,
+      "loss": 1.2393,
+      "step": 48300
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1277314285714284e-06,
+      "loss": 1.1155,
+      "step": 48301
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1274628571428572e-06,
+      "loss": 1.1709,
+      "step": 48302
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1271942857142858e-06,
+      "loss": 1.0165,
+      "step": 48303
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1269257142857142e-06,
+      "loss": 0.7473,
+      "step": 48304
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.126657142857143e-06,
+      "loss": 1.1593,
+      "step": 48305
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1263885714285714e-06,
+      "loss": 1.3282,
+      "step": 48306
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.12612e-06,
+      "loss": 1.1832,
+      "step": 48307
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1258514285714286e-06,
+      "loss": 1.272,
+      "step": 48308
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1255828571428572e-06,
+      "loss": 1.0131,
+      "step": 48309
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1253142857142858e-06,
+      "loss": 0.8469,
+      "step": 48310
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1250457142857142e-06,
+      "loss": 0.7813,
+      "step": 48311
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.124777142857143e-06,
+      "loss": 1.1197,
+      "step": 48312
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1245085714285714e-06,
+      "loss": 1.1221,
+      "step": 48313
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.12424e-06,
+      "loss": 1.4789,
+      "step": 48314
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1239714285714286e-06,
+      "loss": 1.4993,
+      "step": 48315
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1237028571428572e-06,
+      "loss": 1.2074,
+      "step": 48316
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1234342857142856e-06,
+      "loss": 1.4829,
+      "step": 48317
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1231657142857144e-06,
+      "loss": 1.0423,
+      "step": 48318
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.122897142857143e-06,
+      "loss": 1.0628,
+      "step": 48319
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1226285714285714e-06,
+      "loss": 1.0776,
+      "step": 48320
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.12236e-06,
+      "loss": 1.058,
+      "step": 48321
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1220914285714286e-06,
+      "loss": 0.7525,
+      "step": 48322
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1218228571428572e-06,
+      "loss": 1.2654,
+      "step": 48323
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1215542857142856e-06,
+      "loss": 1.0676,
+      "step": 48324
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1212857142857144e-06,
+      "loss": 1.2852,
+      "step": 48325
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.121017142857143e-06,
+      "loss": 0.9062,
+      "step": 48326
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1207485714285714e-06,
+      "loss": 1.2808,
+      "step": 48327
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1204800000000002e-06,
+      "loss": 1.1579,
+      "step": 48328
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1202114285714286e-06,
+      "loss": 1.6443,
+      "step": 48329
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1199428571428572e-06,
+      "loss": 1.1369,
+      "step": 48330
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1196742857142858e-06,
+      "loss": 1.2931,
+      "step": 48331
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1194057142857144e-06,
+      "loss": 0.9887,
+      "step": 48332
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1191371428571428e-06,
+      "loss": 1.2458,
+      "step": 48333
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1188685714285714e-06,
+      "loss": 0.4619,
+      "step": 48334
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1186000000000002e-06,
+      "loss": 1.7116,
+      "step": 48335
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1183314285714286e-06,
+      "loss": 1.435,
+      "step": 48336
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1180628571428572e-06,
+      "loss": 1.4899,
+      "step": 48337
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1177942857142858e-06,
+      "loss": 0.7766,
+      "step": 48338
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1175257142857144e-06,
+      "loss": 0.7176,
+      "step": 48339
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1172571428571428e-06,
+      "loss": 1.1047,
+      "step": 48340
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1169885714285716e-06,
+      "loss": 0.5499,
+      "step": 48341
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1167200000000002e-06,
+      "loss": 1.1299,
+      "step": 48342
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1164514285714286e-06,
+      "loss": 0.8563,
+      "step": 48343
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1161828571428572e-06,
+      "loss": 1.4023,
+      "step": 48344
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1159142857142858e-06,
+      "loss": 1.2822,
+      "step": 48345
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1156457142857144e-06,
+      "loss": 1.5193,
+      "step": 48346
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1153771428571427e-06,
+      "loss": 0.9931,
+      "step": 48347
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1151085714285716e-06,
+      "loss": 0.8357,
+      "step": 48348
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.11484e-06,
+      "loss": 1.0185,
+      "step": 48349
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1145714285714285e-06,
+      "loss": 1.183,
+      "step": 48350
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1143028571428574e-06,
+      "loss": 0.5968,
+      "step": 48351
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1140342857142857e-06,
+      "loss": 1.2764,
+      "step": 48352
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1137657142857143e-06,
+      "loss": 1.1465,
+      "step": 48353
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.113497142857143e-06,
+      "loss": 1.0923,
+      "step": 48354
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1132285714285715e-06,
+      "loss": 1.2784,
+      "step": 48355
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.11296e-06,
+      "loss": 0.8347,
+      "step": 48356
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1126914285714285e-06,
+      "loss": 1.0633,
+      "step": 48357
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1124228571428573e-06,
+      "loss": 1.3141,
+      "step": 48358
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1121542857142857e-06,
+      "loss": 1.3673,
+      "step": 48359
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1118857142857143e-06,
+      "loss": 1.0063,
+      "step": 48360
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.111617142857143e-06,
+      "loss": 0.6814,
+      "step": 48361
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1113485714285715e-06,
+      "loss": 0.83,
+      "step": 48362
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.11108e-06,
+      "loss": 0.7442,
+      "step": 48363
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1108114285714287e-06,
+      "loss": 1.6465,
+      "step": 48364
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1105428571428571e-06,
+      "loss": 1.7602,
+      "step": 48365
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1102742857142857e-06,
+      "loss": 0.7552,
+      "step": 48366
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1100057142857143e-06,
+      "loss": 1.4252,
+      "step": 48367
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.109737142857143e-06,
+      "loss": 1.4601,
+      "step": 48368
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1094685714285715e-06,
+      "loss": 0.8042,
+      "step": 48369
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1092e-06,
+      "loss": 0.9976,
+      "step": 48370
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1089314285714287e-06,
+      "loss": 0.8608,
+      "step": 48371
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.108662857142857e-06,
+      "loss": 1.2963,
+      "step": 48372
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1083942857142857e-06,
+      "loss": 0.9164,
+      "step": 48373
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1081257142857145e-06,
+      "loss": 1.2001,
+      "step": 48374
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.107857142857143e-06,
+      "loss": 1.0042,
+      "step": 48375
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1075885714285715e-06,
+      "loss": 1.3006,
+      "step": 48376
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.10732e-06,
+      "loss": 1.1766,
+      "step": 48377
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1070514285714287e-06,
+      "loss": 1.2112,
+      "step": 48378
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.106782857142857e-06,
+      "loss": 1.3801,
+      "step": 48379
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1065142857142857e-06,
+      "loss": 0.8254,
+      "step": 48380
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1062457142857143e-06,
+      "loss": 1.384,
+      "step": 48381
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1059771428571429e-06,
+      "loss": 1.0562,
+      "step": 48382
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1057085714285715e-06,
+      "loss": 0.9161,
+      "step": 48383
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.10544e-06,
+      "loss": 1.0252,
+      "step": 48384
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1051714285714287e-06,
+      "loss": 0.9592,
+      "step": 48385
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.104902857142857e-06,
+      "loss": 1.1064,
+      "step": 48386
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.1046342857142859e-06,
+      "loss": 1.3582,
+      "step": 48387
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1043657142857143e-06,
+      "loss": 0.4318,
+      "step": 48388
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1040971428571429e-06,
+      "loss": 1.0416,
+      "step": 48389
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1038285714285715e-06,
+      "loss": 1.2162,
+      "step": 48390
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.10356e-06,
+      "loss": 1.105,
+      "step": 48391
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1032914285714287e-06,
+      "loss": 0.8474,
+      "step": 48392
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.103022857142857e-06,
+      "loss": 1.0386,
+      "step": 48393
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1027542857142859e-06,
+      "loss": 1.1902,
+      "step": 48394
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1024857142857143e-06,
+      "loss": 0.9696,
+      "step": 48395
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1022171428571429e-06,
+      "loss": 1.0127,
+      "step": 48396
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1019485714285717e-06,
+      "loss": 1.443,
+      "step": 48397
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.10168e-06,
+      "loss": 0.8919,
+      "step": 48398
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1014114285714287e-06,
+      "loss": 1.1042,
+      "step": 48399
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.101142857142857e-06,
+      "loss": 1.2843,
+      "step": 48400
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1008742857142859e-06,
+      "loss": 0.9147,
+      "step": 48401
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1006057142857142e-06,
+      "loss": 1.2147,
+      "step": 48402
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1003371428571428e-06,
+      "loss": 1.0108,
+      "step": 48403
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.1000685714285714e-06,
+      "loss": 1.3822,
+      "step": 48404
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0998e-06,
+      "loss": 1.3107,
+      "step": 48405
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0995314285714286e-06,
+      "loss": 0.8648,
+      "step": 48406
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0992628571428572e-06,
+      "loss": 1.3361,
+      "step": 48407
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0989942857142858e-06,
+      "loss": 1.1434,
+      "step": 48408
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0987257142857142e-06,
+      "loss": 1.1661,
+      "step": 48409
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0984571428571428e-06,
+      "loss": 0.9388,
+      "step": 48410
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0981885714285714e-06,
+      "loss": 0.9338,
+      "step": 48411
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.09792e-06,
+      "loss": 1.4792,
+      "step": 48412
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0976514285714286e-06,
+      "loss": 1.2122,
+      "step": 48413
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0973828571428572e-06,
+      "loss": 0.9079,
+      "step": 48414
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0971142857142858e-06,
+      "loss": 1.2488,
+      "step": 48415
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0968457142857142e-06,
+      "loss": 1.0824,
+      "step": 48416
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.096577142857143e-06,
+      "loss": 0.8269,
+      "step": 48417
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0963085714285714e-06,
+      "loss": 1.1049,
+      "step": 48418
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.09604e-06,
+      "loss": 1.4502,
+      "step": 48419
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0957714285714286e-06,
+      "loss": 1.1681,
+      "step": 48420
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0955028571428572e-06,
+      "loss": 0.852,
+      "step": 48421
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0952342857142858e-06,
+      "loss": 1.2111,
+      "step": 48422
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0949657142857142e-06,
+      "loss": 1.1039,
+      "step": 48423
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.094697142857143e-06,
+      "loss": 1.1792,
+      "step": 48424
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0944285714285714e-06,
+      "loss": 0.8407,
+      "step": 48425
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.09416e-06,
+      "loss": 1.0234,
+      "step": 48426
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0938914285714286e-06,
+      "loss": 1.205,
+      "step": 48427
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0936228571428572e-06,
+      "loss": 1.0878,
+      "step": 48428
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0933542857142858e-06,
+      "loss": 1.2292,
+      "step": 48429
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0930857142857144e-06,
+      "loss": 1.0544,
+      "step": 48430
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.092817142857143e-06,
+      "loss": 1.3421,
+      "step": 48431
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0925485714285714e-06,
+      "loss": 1.3539,
+      "step": 48432
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.09228e-06,
+      "loss": 1.208,
+      "step": 48433
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0920114285714286e-06,
+      "loss": 1.4046,
+      "step": 48434
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0917428571428572e-06,
+      "loss": 0.6917,
+      "step": 48435
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0914742857142856e-06,
+      "loss": 0.8407,
+      "step": 48436
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0912057142857144e-06,
+      "loss": 1.2527,
+      "step": 48437
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.090937142857143e-06,
+      "loss": 1.5119,
+      "step": 48438
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0906685714285714e-06,
+      "loss": 1.3845,
+      "step": 48439
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0904000000000002e-06,
+      "loss": 0.8047,
+      "step": 48440
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0901314285714286e-06,
+      "loss": 1.1236,
+      "step": 48441
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0898628571428572e-06,
+      "loss": 1.1984,
+      "step": 48442
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0895942857142858e-06,
+      "loss": 1.22,
+      "step": 48443
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0893257142857144e-06,
+      "loss": 1.0922,
+      "step": 48444
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.089057142857143e-06,
+      "loss": 0.9993,
+      "step": 48445
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0887885714285714e-06,
+      "loss": 0.863,
+      "step": 48446
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0885200000000002e-06,
+      "loss": 1.213,
+      "step": 48447
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0882514285714286e-06,
+      "loss": 1.0467,
+      "step": 48448
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0879828571428572e-06,
+      "loss": 1.3347,
+      "step": 48449
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0877142857142858e-06,
+      "loss": 1.4114,
+      "step": 48450
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0874457142857144e-06,
+      "loss": 1.1099,
+      "step": 48451
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0871771428571427e-06,
+      "loss": 1.1138,
+      "step": 48452
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0869085714285716e-06,
+      "loss": 1.196,
+      "step": 48453
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0866400000000002e-06,
+      "loss": 1.1211,
+      "step": 48454
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0863714285714285e-06,
+      "loss": 0.9808,
+      "step": 48455
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0861028571428571e-06,
+      "loss": 0.8938,
+      "step": 48456
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0858342857142857e-06,
+      "loss": 1.1147,
+      "step": 48457
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0855657142857143e-06,
+      "loss": 1.1211,
+      "step": 48458
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0852971428571427e-06,
+      "loss": 1.3156,
+      "step": 48459
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0850285714285715e-06,
+      "loss": 1.1334,
+      "step": 48460
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0847600000000001e-06,
+      "loss": 1.0117,
+      "step": 48461
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0844914285714285e-06,
+      "loss": 1.0874,
+      "step": 48462
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0842228571428573e-06,
+      "loss": 1.1775,
+      "step": 48463
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0839542857142857e-06,
+      "loss": 1.2337,
+      "step": 48464
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0836857142857143e-06,
+      "loss": 1.1686,
+      "step": 48465
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.083417142857143e-06,
+      "loss": 0.9631,
+      "step": 48466
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0831485714285715e-06,
+      "loss": 1.1294,
+      "step": 48467
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.08288e-06,
+      "loss": 0.8612,
+      "step": 48468
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0826114285714285e-06,
+      "loss": 1.1606,
+      "step": 48469
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0823428571428573e-06,
+      "loss": 1.3579,
+      "step": 48470
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0820742857142857e-06,
+      "loss": 1.2669,
+      "step": 48471
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0818057142857143e-06,
+      "loss": 0.8341,
+      "step": 48472
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.081537142857143e-06,
+      "loss": 1.5507,
+      "step": 48473
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0812685714285715e-06,
+      "loss": 0.9161,
+      "step": 48474
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.081e-06,
+      "loss": 1.1359,
+      "step": 48475
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0807314285714287e-06,
+      "loss": 1.4452,
+      "step": 48476
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0804628571428573e-06,
+      "loss": 1.3716,
+      "step": 48477
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0801942857142857e-06,
+      "loss": 1.1967,
+      "step": 48478
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0799257142857143e-06,
+      "loss": 1.1318,
+      "step": 48479
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.079657142857143e-06,
+      "loss": 0.6248,
+      "step": 48480
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0793885714285715e-06,
+      "loss": 1.2186,
+      "step": 48481
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0791199999999999e-06,
+      "loss": 1.1993,
+      "step": 48482
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0788514285714287e-06,
+      "loss": 0.8253,
+      "step": 48483
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.078582857142857e-06,
+      "loss": 1.5914,
+      "step": 48484
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0783142857142857e-06,
+      "loss": 1.231,
+      "step": 48485
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0780457142857145e-06,
+      "loss": 1.2702,
+      "step": 48486
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0777771428571429e-06,
+      "loss": 0.8178,
+      "step": 48487
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0775085714285715e-06,
+      "loss": 1.1419,
+      "step": 48488
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.07724e-06,
+      "loss": 0.9441,
+      "step": 48489
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0769714285714287e-06,
+      "loss": 0.9334,
+      "step": 48490
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.076702857142857e-06,
+      "loss": 0.6657,
+      "step": 48491
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0764342857142857e-06,
+      "loss": 0.978,
+      "step": 48492
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0761657142857145e-06,
+      "loss": 1.4053,
+      "step": 48493
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0758971428571429e-06,
+      "loss": 0.9628,
+      "step": 48494
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0756285714285715e-06,
+      "loss": 1.0659,
+      "step": 48495
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.07536e-06,
+      "loss": 1.1327,
+      "step": 48496
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0750914285714287e-06,
+      "loss": 1.0866,
+      "step": 48497
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.074822857142857e-06,
+      "loss": 1.0885,
+      "step": 48498
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0745542857142859e-06,
+      "loss": 1.0905,
+      "step": 48499
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0742857142857145e-06,
+      "loss": 1.3492,
+      "step": 48500
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0740171428571429e-06,
+      "loss": 1.096,
+      "step": 48501
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0737485714285715e-06,
+      "loss": 1.2365,
+      "step": 48502
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.07348e-06,
+      "loss": 1.0846,
+      "step": 48503
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0732114285714287e-06,
+      "loss": 1.1282,
+      "step": 48504
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.072942857142857e-06,
+      "loss": 1.2444,
+      "step": 48505
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0726742857142859e-06,
+      "loss": 0.9819,
+      "step": 48506
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0724057142857142e-06,
+      "loss": 0.9764,
+      "step": 48507
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0721371428571428e-06,
+      "loss": 1.2355,
+      "step": 48508
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0718685714285717e-06,
+      "loss": 0.977,
+      "step": 48509
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0716e-06,
+      "loss": 1.4484,
+      "step": 48510
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0713314285714286e-06,
+      "loss": 1.3279,
+      "step": 48511
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.071062857142857e-06,
+      "loss": 0.9697,
+      "step": 48512
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0707942857142858e-06,
+      "loss": 1.315,
+      "step": 48513
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0705257142857142e-06,
+      "loss": 1.2029,
+      "step": 48514
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0702571428571428e-06,
+      "loss": 0.9039,
+      "step": 48515
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0699885714285716e-06,
+      "loss": 0.7986,
+      "step": 48516
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.06972e-06,
+      "loss": 0.9484,
+      "step": 48517
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0694514285714286e-06,
+      "loss": 1.1939,
+      "step": 48518
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0691828571428572e-06,
+      "loss": 1.1226,
+      "step": 48519
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0689142857142858e-06,
+      "loss": 0.9738,
+      "step": 48520
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0686457142857142e-06,
+      "loss": 1.2728,
+      "step": 48521
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0683771428571428e-06,
+      "loss": 0.9564,
+      "step": 48522
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0681085714285714e-06,
+      "loss": 1.0495,
+      "step": 48523
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.06784e-06,
+      "loss": 0.9876,
+      "step": 48524
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0675714285714286e-06,
+      "loss": 0.795,
+      "step": 48525
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0673028571428572e-06,
+      "loss": 0.9744,
+      "step": 48526
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0670342857142858e-06,
+      "loss": 1.4139,
+      "step": 48527
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0667657142857142e-06,
+      "loss": 1.3367,
+      "step": 48528
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.066497142857143e-06,
+      "loss": 1.0387,
+      "step": 48529
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0662285714285714e-06,
+      "loss": 1.5401,
+      "step": 48530
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.06596e-06,
+      "loss": 1.0173,
+      "step": 48531
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0656914285714288e-06,
+      "loss": 1.0171,
+      "step": 48532
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0654228571428572e-06,
+      "loss": 1.335,
+      "step": 48533
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0651542857142858e-06,
+      "loss": 1.6934,
+      "step": 48534
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0648857142857142e-06,
+      "loss": 1.018,
+      "step": 48535
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.064617142857143e-06,
+      "loss": 1.5516,
+      "step": 48536
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0643485714285714e-06,
+      "loss": 1.0076,
+      "step": 48537
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.06408e-06,
+      "loss": 0.9522,
+      "step": 48538
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0638114285714286e-06,
+      "loss": 0.8144,
+      "step": 48539
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0635428571428572e-06,
+      "loss": 1.2697,
+      "step": 48540
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0632742857142858e-06,
+      "loss": 0.8544,
+      "step": 48541
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0630057142857144e-06,
+      "loss": 1.0817,
+      "step": 48542
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.062737142857143e-06,
+      "loss": 1.3233,
+      "step": 48543
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0624685714285714e-06,
+      "loss": 0.8477,
+      "step": 48544
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0622e-06,
+      "loss": 1.1422,
+      "step": 48545
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0619314285714286e-06,
+      "loss": 1.5243,
+      "step": 48546
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0616628571428572e-06,
+      "loss": 0.7633,
+      "step": 48547
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0613942857142858e-06,
+      "loss": 0.934,
+      "step": 48548
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0611257142857144e-06,
+      "loss": 1.2519,
+      "step": 48549
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.060857142857143e-06,
+      "loss": 1.2716,
+      "step": 48550
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0605885714285714e-06,
+      "loss": 1.4859,
+      "step": 48551
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0603200000000002e-06,
+      "loss": 1.5308,
+      "step": 48552
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0600514285714286e-06,
+      "loss": 1.6031,
+      "step": 48553
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0597828571428572e-06,
+      "loss": 0.92,
+      "step": 48554
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0595142857142858e-06,
+      "loss": 1.1685,
+      "step": 48555
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0592457142857144e-06,
+      "loss": 1.0009,
+      "step": 48556
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.058977142857143e-06,
+      "loss": 1.2465,
+      "step": 48557
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0587085714285713e-06,
+      "loss": 0.9788,
+      "step": 48558
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0584400000000002e-06,
+      "loss": 1.0552,
+      "step": 48559
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0581714285714285e-06,
+      "loss": 0.893,
+      "step": 48560
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0579028571428571e-06,
+      "loss": 1.2484,
+      "step": 48561
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.0576342857142857e-06,
+      "loss": 1.0505,
+      "step": 48562
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0573657142857143e-06,
+      "loss": 1.035,
+      "step": 48563
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.057097142857143e-06,
+      "loss": 0.9457,
+      "step": 48564
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0568285714285715e-06,
+      "loss": 1.3379,
+      "step": 48565
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0565600000000001e-06,
+      "loss": 0.9622,
+      "step": 48566
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0562914285714285e-06,
+      "loss": 1.5905,
+      "step": 48567
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0560228571428571e-06,
+      "loss": 1.1269,
+      "step": 48568
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0557542857142857e-06,
+      "loss": 1.2082,
+      "step": 48569
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0554857142857143e-06,
+      "loss": 0.9313,
+      "step": 48570
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0552171428571427e-06,
+      "loss": 1.1548,
+      "step": 48571
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0549485714285715e-06,
+      "loss": 0.9767,
+      "step": 48572
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0546800000000001e-06,
+      "loss": 1.458,
+      "step": 48573
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0544114285714285e-06,
+      "loss": 1.5606,
+      "step": 48574
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0541428571428573e-06,
+      "loss": 1.5808,
+      "step": 48575
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0538742857142857e-06,
+      "loss": 1.0831,
+      "step": 48576
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0536057142857143e-06,
+      "loss": 1.329,
+      "step": 48577
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.053337142857143e-06,
+      "loss": 1.2864,
+      "step": 48578
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0530685714285715e-06,
+      "loss": 1.1231,
+      "step": 48579
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0528000000000001e-06,
+      "loss": 1.0553,
+      "step": 48580
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0525314285714285e-06,
+      "loss": 0.9747,
+      "step": 48581
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0522628571428573e-06,
+      "loss": 1.1648,
+      "step": 48582
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0519942857142857e-06,
+      "loss": 1.3822,
+      "step": 48583
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0517257142857143e-06,
+      "loss": 0.9023,
+      "step": 48584
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.051457142857143e-06,
+      "loss": 0.848,
+      "step": 48585
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0511885714285715e-06,
+      "loss": 1.0558,
+      "step": 48586
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0509199999999999e-06,
+      "loss": 1.3724,
+      "step": 48587
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0506514285714287e-06,
+      "loss": 1.1312,
+      "step": 48588
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0503828571428573e-06,
+      "loss": 1.2746,
+      "step": 48589
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0501142857142857e-06,
+      "loss": 1.128,
+      "step": 48590
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0498457142857143e-06,
+      "loss": 1.4834,
+      "step": 48591
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0495771428571429e-06,
+      "loss": 1.2922,
+      "step": 48592
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0493085714285715e-06,
+      "loss": 1.3385,
+      "step": 48593
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0490399999999999e-06,
+      "loss": 1.2225,
+      "step": 48594
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0487714285714287e-06,
+      "loss": 0.8482,
+      "step": 48595
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0485028571428573e-06,
+      "loss": 1.1787,
+      "step": 48596
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0482342857142857e-06,
+      "loss": 1.137,
+      "step": 48597
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0479657142857145e-06,
+      "loss": 1.0865,
+      "step": 48598
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0476971428571429e-06,
+      "loss": 1.054,
+      "step": 48599
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0474285714285715e-06,
+      "loss": 0.9783,
+      "step": 48600
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.04716e-06,
+      "loss": 0.7874,
+      "step": 48601
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0468914285714287e-06,
+      "loss": 0.8416,
+      "step": 48602
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0466228571428573e-06,
+      "loss": 1.0174,
+      "step": 48603
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0463542857142857e-06,
+      "loss": 1.2315,
+      "step": 48604
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0460857142857145e-06,
+      "loss": 0.9046,
+      "step": 48605
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0458171428571429e-06,
+      "loss": 1.0814,
+      "step": 48606
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0455485714285715e-06,
+      "loss": 0.9171,
+      "step": 48607
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.04528e-06,
+      "loss": 1.0626,
+      "step": 48608
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0450114285714287e-06,
+      "loss": 1.0032,
+      "step": 48609
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.044742857142857e-06,
+      "loss": 0.8647,
+      "step": 48610
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0444742857142859e-06,
+      "loss": 1.0187,
+      "step": 48611
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0442057142857145e-06,
+      "loss": 1.2666,
+      "step": 48612
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0439371428571428e-06,
+      "loss": 1.4827,
+      "step": 48613
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0436685714285714e-06,
+      "loss": 1.2341,
+      "step": 48614
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0434e-06,
+      "loss": 1.556,
+      "step": 48615
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0431314285714286e-06,
+      "loss": 1.0665,
+      "step": 48616
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.042862857142857e-06,
+      "loss": 0.6395,
+      "step": 48617
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0425942857142858e-06,
+      "loss": 1.3547,
+      "step": 48618
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0423257142857144e-06,
+      "loss": 1.0082,
+      "step": 48619
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0420571428571428e-06,
+      "loss": 0.9921,
+      "step": 48620
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0417885714285716e-06,
+      "loss": 1.433,
+      "step": 48621
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.04152e-06,
+      "loss": 1.0037,
+      "step": 48622
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0412514285714286e-06,
+      "loss": 0.9527,
+      "step": 48623
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.040982857142857e-06,
+      "loss": 1.4546,
+      "step": 48624
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0407142857142858e-06,
+      "loss": 1.4204,
+      "step": 48625
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0404457142857142e-06,
+      "loss": 1.1982,
+      "step": 48626
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0401771428571428e-06,
+      "loss": 1.7876,
+      "step": 48627
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0399085714285716e-06,
+      "loss": 0.6986,
+      "step": 48628
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.03964e-06,
+      "loss": 0.8682,
+      "step": 48629
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0393714285714286e-06,
+      "loss": 1.1543,
+      "step": 48630
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0391028571428572e-06,
+      "loss": 1.0421,
+      "step": 48631
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0388342857142858e-06,
+      "loss": 1.298,
+      "step": 48632
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0385657142857142e-06,
+      "loss": 1.3663,
+      "step": 48633
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.038297142857143e-06,
+      "loss": 1.2283,
+      "step": 48634
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0380285714285716e-06,
+      "loss": 1.1774,
+      "step": 48635
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.03776e-06,
+      "loss": 1.022,
+      "step": 48636
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0374914285714286e-06,
+      "loss": 1.2743,
+      "step": 48637
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0372228571428572e-06,
+      "loss": 0.9839,
+      "step": 48638
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0369542857142858e-06,
+      "loss": 0.5683,
+      "step": 48639
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0366857142857142e-06,
+      "loss": 1.0135,
+      "step": 48640
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.036417142857143e-06,
+      "loss": 1.2514,
+      "step": 48641
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0361485714285714e-06,
+      "loss": 1.0772,
+      "step": 48642
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.03588e-06,
+      "loss": 1.1737,
+      "step": 48643
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0356114285714288e-06,
+      "loss": 1.0925,
+      "step": 48644
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0353428571428572e-06,
+      "loss": 1.1679,
+      "step": 48645
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0350742857142858e-06,
+      "loss": 1.0832,
+      "step": 48646
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0348057142857142e-06,
+      "loss": 0.848,
+      "step": 48647
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.034537142857143e-06,
+      "loss": 0.9867,
+      "step": 48648
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0342685714285714e-06,
+      "loss": 1.2675,
+      "step": 48649
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.034e-06,
+      "loss": 1.584,
+      "step": 48650
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0337314285714288e-06,
+      "loss": 1.2104,
+      "step": 48651
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0334628571428572e-06,
+      "loss": 0.9088,
+      "step": 48652
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0331942857142858e-06,
+      "loss": 0.9758,
+      "step": 48653
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0329257142857144e-06,
+      "loss": 1.6196,
+      "step": 48654
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.032657142857143e-06,
+      "loss": 0.9321,
+      "step": 48655
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0323885714285714e-06,
+      "loss": 1.1943,
+      "step": 48656
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.03212e-06,
+      "loss": 1.2706,
+      "step": 48657
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0318514285714286e-06,
+      "loss": 1.1308,
+      "step": 48658
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0315828571428572e-06,
+      "loss": 1.3402,
+      "step": 48659
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0313142857142858e-06,
+      "loss": 1.2506,
+      "step": 48660
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0310457142857144e-06,
+      "loss": 1.1623,
+      "step": 48661
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.030777142857143e-06,
+      "loss": 0.796,
+      "step": 48662
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0305085714285713e-06,
+      "loss": 1.1177,
+      "step": 48663
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0302400000000002e-06,
+      "loss": 0.8957,
+      "step": 48664
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0299714285714285e-06,
+      "loss": 1.4133,
+      "step": 48665
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0297028571428571e-06,
+      "loss": 1.1413,
+      "step": 48666
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0294342857142857e-06,
+      "loss": 0.8442,
+      "step": 48667
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0291657142857143e-06,
+      "loss": 0.8813,
+      "step": 48668
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.028897142857143e-06,
+      "loss": 0.9419,
+      "step": 48669
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0286285714285713e-06,
+      "loss": 1.0893,
+      "step": 48670
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0283600000000001e-06,
+      "loss": 1.3695,
+      "step": 48671
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0280914285714285e-06,
+      "loss": 0.9888,
+      "step": 48672
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0278228571428571e-06,
+      "loss": 1.023,
+      "step": 48673
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0275542857142857e-06,
+      "loss": 1.0178,
+      "step": 48674
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0272857142857143e-06,
+      "loss": 1.4692,
+      "step": 48675
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.027017142857143e-06,
+      "loss": 1.0808,
+      "step": 48676
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0267485714285715e-06,
+      "loss": 1.4116,
+      "step": 48677
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0264800000000001e-06,
+      "loss": 1.4881,
+      "step": 48678
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0262114285714285e-06,
+      "loss": 1.0096,
+      "step": 48679
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0259428571428571e-06,
+      "loss": 1.0894,
+      "step": 48680
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0256742857142857e-06,
+      "loss": 1.09,
+      "step": 48681
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0254057142857143e-06,
+      "loss": 1.2936,
+      "step": 48682
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.025137142857143e-06,
+      "loss": 1.2028,
+      "step": 48683
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0248685714285715e-06,
+      "loss": 1.5298,
+      "step": 48684
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0246000000000001e-06,
+      "loss": 1.5703,
+      "step": 48685
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0243314285714285e-06,
+      "loss": 0.6689,
+      "step": 48686
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0240628571428573e-06,
+      "loss": 1.3383,
+      "step": 48687
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0237942857142857e-06,
+      "loss": 1.1967,
+      "step": 48688
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0235257142857143e-06,
+      "loss": 0.769,
+      "step": 48689
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.023257142857143e-06,
+      "loss": 1.0626,
+      "step": 48690
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0229885714285715e-06,
+      "loss": 1.2052,
+      "step": 48691
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.02272e-06,
+      "loss": 1.3963,
+      "step": 48692
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0224514285714285e-06,
+      "loss": 1.0261,
+      "step": 48693
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0221828571428573e-06,
+      "loss": 1.0714,
+      "step": 48694
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0219142857142857e-06,
+      "loss": 0.9106,
+      "step": 48695
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0216457142857143e-06,
+      "loss": 0.8022,
+      "step": 48696
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0213771428571429e-06,
+      "loss": 0.7957,
+      "step": 48697
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0211085714285715e-06,
+      "loss": 1.196,
+      "step": 48698
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.02084e-06,
+      "loss": 0.9268,
+      "step": 48699
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0205714285714287e-06,
+      "loss": 1.1475,
+      "step": 48700
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0203028571428573e-06,
+      "loss": 1.0548,
+      "step": 48701
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0200342857142857e-06,
+      "loss": 1.2213,
+      "step": 48702
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0197657142857143e-06,
+      "loss": 1.2559,
+      "step": 48703
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0194971428571429e-06,
+      "loss": 1.2699,
+      "step": 48704
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0192285714285715e-06,
+      "loss": 1.4358,
+      "step": 48705
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.01896e-06,
+      "loss": 1.0683,
+      "step": 48706
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0186914285714287e-06,
+      "loss": 1.1167,
+      "step": 48707
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0184228571428573e-06,
+      "loss": 1.4662,
+      "step": 48708
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0181542857142857e-06,
+      "loss": 1.1486,
+      "step": 48709
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0178857142857145e-06,
+      "loss": 1.1134,
+      "step": 48710
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0176171428571429e-06,
+      "loss": 0.7655,
+      "step": 48711
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0173485714285715e-06,
+      "loss": 1.1145,
+      "step": 48712
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.01708e-06,
+      "loss": 0.6492,
+      "step": 48713
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0168114285714287e-06,
+      "loss": 1.186,
+      "step": 48714
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0165428571428573e-06,
+      "loss": 0.8699,
+      "step": 48715
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0162742857142856e-06,
+      "loss": 1.1971,
+      "step": 48716
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0160057142857145e-06,
+      "loss": 1.104,
+      "step": 48717
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0157371428571428e-06,
+      "loss": 1.0076,
+      "step": 48718
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0154685714285714e-06,
+      "loss": 1.0254,
+      "step": 48719
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0152e-06,
+      "loss": 1.1501,
+      "step": 48720
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0149314285714286e-06,
+      "loss": 1.4742,
+      "step": 48721
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0146628571428572e-06,
+      "loss": 1.0784,
+      "step": 48722
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0143942857142858e-06,
+      "loss": 1.118,
+      "step": 48723
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0141257142857144e-06,
+      "loss": 1.3108,
+      "step": 48724
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0138571428571428e-06,
+      "loss": 0.8289,
+      "step": 48725
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0135885714285714e-06,
+      "loss": 0.8011,
+      "step": 48726
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.01332e-06,
+      "loss": 1.187,
+      "step": 48727
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0130514285714286e-06,
+      "loss": 1.2827,
+      "step": 48728
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.012782857142857e-06,
+      "loss": 0.8732,
+      "step": 48729
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0125142857142858e-06,
+      "loss": 1.0538,
+      "step": 48730
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0122457142857144e-06,
+      "loss": 1.2807,
+      "step": 48731
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0119771428571428e-06,
+      "loss": 1.2228,
+      "step": 48732
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0117085714285716e-06,
+      "loss": 1.0645,
+      "step": 48733
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.01144e-06,
+      "loss": 0.9703,
+      "step": 48734
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0111714285714286e-06,
+      "loss": 1.2522,
+      "step": 48735
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.010902857142857e-06,
+      "loss": 1.039,
+      "step": 48736
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.0106342857142858e-06,
+      "loss": 1.1038,
+      "step": 48737
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0103657142857144e-06,
+      "loss": 1.2288,
+      "step": 48738
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0100971428571428e-06,
+      "loss": 1.2529,
+      "step": 48739
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0098285714285716e-06,
+      "loss": 1.3738,
+      "step": 48740
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.00956e-06,
+      "loss": 1.1443,
+      "step": 48741
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0092914285714286e-06,
+      "loss": 1.0376,
+      "step": 48742
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0090228571428572e-06,
+      "loss": 1.1885,
+      "step": 48743
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0087542857142858e-06,
+      "loss": 1.1164,
+      "step": 48744
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0084857142857142e-06,
+      "loss": 1.2882,
+      "step": 48745
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.008217142857143e-06,
+      "loss": 1.1565,
+      "step": 48746
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0079485714285716e-06,
+      "loss": 1.1619,
+      "step": 48747
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.00768e-06,
+      "loss": 1.2222,
+      "step": 48748
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0074114285714286e-06,
+      "loss": 1.0545,
+      "step": 48749
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0071428571428572e-06,
+      "loss": 0.9453,
+      "step": 48750
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0068742857142858e-06,
+      "loss": 1.1267,
+      "step": 48751
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0066057142857142e-06,
+      "loss": 1.0849,
+      "step": 48752
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.006337142857143e-06,
+      "loss": 1.081,
+      "step": 48753
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0060685714285716e-06,
+      "loss": 0.8489,
+      "step": 48754
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0058e-06,
+      "loss": 0.8977,
+      "step": 48755
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0055314285714288e-06,
+      "loss": 1.2303,
+      "step": 48756
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0052628571428572e-06,
+      "loss": 1.1881,
+      "step": 48757
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0049942857142858e-06,
+      "loss": 0.9034,
+      "step": 48758
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0047257142857142e-06,
+      "loss": 0.7743,
+      "step": 48759
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.004457142857143e-06,
+      "loss": 1.3393,
+      "step": 48760
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0041885714285714e-06,
+      "loss": 1.2408,
+      "step": 48761
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.00392e-06,
+      "loss": 1.0392,
+      "step": 48762
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0036514285714288e-06,
+      "loss": 0.9939,
+      "step": 48763
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0033828571428572e-06,
+      "loss": 1.465,
+      "step": 48764
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0031142857142858e-06,
+      "loss": 1.2098,
+      "step": 48765
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0028457142857144e-06,
+      "loss": 0.9139,
+      "step": 48766
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.002577142857143e-06,
+      "loss": 1.0964,
+      "step": 48767
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0023085714285713e-06,
+      "loss": 1.156,
+      "step": 48768
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.00204e-06,
+      "loss": 1.3854,
+      "step": 48769
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0017714285714288e-06,
+      "loss": 1.5742,
+      "step": 48770
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0015028571428571e-06,
+      "loss": 1.0176,
+      "step": 48771
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0012342857142857e-06,
+      "loss": 1.0047,
+      "step": 48772
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0009657142857143e-06,
+      "loss": 1.0105,
+      "step": 48773
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.000697142857143e-06,
+      "loss": 1.0537,
+      "step": 48774
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0004285714285713e-06,
+      "loss": 0.6542,
+      "step": 48775
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.0001600000000001e-06,
+      "loss": 1.2597,
+      "step": 48776
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.998914285714285e-07,
+      "loss": 0.9389,
+      "step": 48777
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.996228571428571e-07,
+      "loss": 1.2802,
+      "step": 48778
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.993542857142857e-07,
+      "loss": 1.3906,
+      "step": 48779
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.990857142857143e-07,
+      "loss": 1.1053,
+      "step": 48780
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.98817142857143e-07,
+      "loss": 1.1165,
+      "step": 48781
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.985485714285713e-07,
+      "loss": 1.2234,
+      "step": 48782
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.982800000000001e-07,
+      "loss": 1.0486,
+      "step": 48783
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.980114285714285e-07,
+      "loss": 1.49,
+      "step": 48784
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.977428571428571e-07,
+      "loss": 1.1858,
+      "step": 48785
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.97474285714286e-07,
+      "loss": 1.1813,
+      "step": 48786
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.972057142857143e-07,
+      "loss": 1.1988,
+      "step": 48787
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.96937142857143e-07,
+      "loss": 0.9973,
+      "step": 48788
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.966685714285715e-07,
+      "loss": 1.3882,
+      "step": 48789
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.964000000000001e-07,
+      "loss": 0.782,
+      "step": 48790
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.961314285714285e-07,
+      "loss": 0.9456,
+      "step": 48791
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.95862857142857e-07,
+      "loss": 1.056,
+      "step": 48792
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.95594285714286e-07,
+      "loss": 1.1784,
+      "step": 48793
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.953257142857143e-07,
+      "loss": 1.0178,
+      "step": 48794
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.95057142857143e-07,
+      "loss": 0.8239,
+      "step": 48795
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.947885714285715e-07,
+      "loss": 1.3581,
+      "step": 48796
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.9452e-07,
+      "loss": 1.484,
+      "step": 48797
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.942514285714285e-07,
+      "loss": 1.1712,
+      "step": 48798
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.939828571428573e-07,
+      "loss": 1.0276,
+      "step": 48799
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.937142857142857e-07,
+      "loss": 0.957,
+      "step": 48800
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.934457142857143e-07,
+      "loss": 1.0078,
+      "step": 48801
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.931771428571429e-07,
+      "loss": 1.018,
+      "step": 48802
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.929085714285715e-07,
+      "loss": 1.447,
+      "step": 48803
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.9264e-07,
+      "loss": 1.2614,
+      "step": 48804
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.923714285714285e-07,
+      "loss": 1.0178,
+      "step": 48805
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.921028571428573e-07,
+      "loss": 0.8394,
+      "step": 48806
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.918342857142857e-07,
+      "loss": 0.9896,
+      "step": 48807
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.915657142857143e-07,
+      "loss": 1.1182,
+      "step": 48808
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.91297142857143e-07,
+      "loss": 1.1033,
+      "step": 48809
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.910285714285715e-07,
+      "loss": 1.3928,
+      "step": 48810
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.9076e-07,
+      "loss": 0.9738,
+      "step": 48811
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.904914285714287e-07,
+      "loss": 1.1567,
+      "step": 48812
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.902228571428573e-07,
+      "loss": 1.4488,
+      "step": 48813
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.899542857142857e-07,
+      "loss": 1.1843,
+      "step": 48814
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.896857142857143e-07,
+      "loss": 0.9399,
+      "step": 48815
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.894171428571429e-07,
+      "loss": 0.9018,
+      "step": 48816
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.891485714285715e-07,
+      "loss": 1.1988,
+      "step": 48817
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.8888e-07,
+      "loss": 0.8551,
+      "step": 48818
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.886114285714287e-07,
+      "loss": 1.4806,
+      "step": 48819
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.883428571428573e-07,
+      "loss": 0.9071,
+      "step": 48820
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.880742857142856e-07,
+      "loss": 0.978,
+      "step": 48821
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.878057142857144e-07,
+      "loss": 1.1241,
+      "step": 48822
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.875371428571428e-07,
+      "loss": 0.8185,
+      "step": 48823
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.872685714285714e-07,
+      "loss": 1.0462,
+      "step": 48824
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.87e-07,
+      "loss": 1.2306,
+      "step": 48825
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.867314285714286e-07,
+      "loss": 0.8031,
+      "step": 48826
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.864628571428572e-07,
+      "loss": 0.9978,
+      "step": 48827
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.861942857142856e-07,
+      "loss": 0.8711,
+      "step": 48828
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.859257142857144e-07,
+      "loss": 0.9197,
+      "step": 48829
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.856571428571428e-07,
+      "loss": 1.5128,
+      "step": 48830
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.853885714285714e-07,
+      "loss": 0.9877,
+      "step": 48831
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.8512e-07,
+      "loss": 1.3291,
+      "step": 48832
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.848514285714286e-07,
+      "loss": 1.0254,
+      "step": 48833
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.845828571428572e-07,
+      "loss": 1.0903,
+      "step": 48834
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.843142857142858e-07,
+      "loss": 0.9085,
+      "step": 48835
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.840457142857144e-07,
+      "loss": 1.0001,
+      "step": 48836
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.837771428571428e-07,
+      "loss": 0.9711,
+      "step": 48837
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.835085714285714e-07,
+      "loss": 0.743,
+      "step": 48838
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.8324e-07,
+      "loss": 1.3702,
+      "step": 48839
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.829714285714286e-07,
+      "loss": 1.0455,
+      "step": 48840
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.827028571428572e-07,
+      "loss": 1.513,
+      "step": 48841
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.824342857142858e-07,
+      "loss": 0.8947,
+      "step": 48842
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.821657142857144e-07,
+      "loss": 1.3154,
+      "step": 48843
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.818971428571428e-07,
+      "loss": 1.2404,
+      "step": 48844
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.816285714285716e-07,
+      "loss": 0.9358,
+      "step": 48845
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.8136e-07,
+      "loss": 0.8806,
+      "step": 48846
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.810914285714286e-07,
+      "loss": 1.508,
+      "step": 48847
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.80822857142857e-07,
+      "loss": 1.4196,
+      "step": 48848
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.805542857142858e-07,
+      "loss": 1.0396,
+      "step": 48849
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.802857142857144e-07,
+      "loss": 1.4247,
+      "step": 48850
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.800171428571428e-07,
+      "loss": 0.8912,
+      "step": 48851
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.797485714285716e-07,
+      "loss": 0.973,
+      "step": 48852
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.7948e-07,
+      "loss": 1.283,
+      "step": 48853
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.792114285714286e-07,
+      "loss": 1.2124,
+      "step": 48854
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.789428571428572e-07,
+      "loss": 1.2428,
+      "step": 48855
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.786742857142858e-07,
+      "loss": 1.011,
+      "step": 48856
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.784057142857144e-07,
+      "loss": 0.8227,
+      "step": 48857
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.78137142857143e-07,
+      "loss": 0.97,
+      "step": 48858
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.778685714285716e-07,
+      "loss": 1.4773,
+      "step": 48859
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.776e-07,
+      "loss": 1.2625,
+      "step": 48860
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.773314285714286e-07,
+      "loss": 1.0117,
+      "step": 48861
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.770628571428572e-07,
+      "loss": 0.7337,
+      "step": 48862
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.767942857142858e-07,
+      "loss": 1.2048,
+      "step": 48863
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.765257142857142e-07,
+      "loss": 1.3223,
+      "step": 48864
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.76257142857143e-07,
+      "loss": 1.0727,
+      "step": 48865
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.759885714285716e-07,
+      "loss": 1.1679,
+      "step": 48866
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.7572e-07,
+      "loss": 0.9723,
+      "step": 48867
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.754514285714288e-07,
+      "loss": 1.3329,
+      "step": 48868
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.751828571428572e-07,
+      "loss": 1.5354,
+      "step": 48869
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.749142857142858e-07,
+      "loss": 1.2973,
+      "step": 48870
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.746457142857141e-07,
+      "loss": 0.853,
+      "step": 48871
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.74377142857143e-07,
+      "loss": 0.989,
+      "step": 48872
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.741085714285715e-07,
+      "loss": 1.0605,
+      "step": 48873
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.7384e-07,
+      "loss": 0.9374,
+      "step": 48874
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.735714285714287e-07,
+      "loss": 1.0383,
+      "step": 48875
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.733028571428571e-07,
+      "loss": 0.9451,
+      "step": 48876
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.730342857142857e-07,
+      "loss": 1.3528,
+      "step": 48877
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.727657142857143e-07,
+      "loss": 1.4773,
+      "step": 48878
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.72497142857143e-07,
+      "loss": 1.4617,
+      "step": 48879
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.722285714285713e-07,
+      "loss": 1.0531,
+      "step": 48880
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.7196e-07,
+      "loss": 1.3362,
+      "step": 48881
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.716914285714287e-07,
+      "loss": 1.0176,
+      "step": 48882
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.714228571428571e-07,
+      "loss": 1.1628,
+      "step": 48883
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.711542857142857e-07,
+      "loss": 1.6268,
+      "step": 48884
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.708857142857143e-07,
+      "loss": 0.8168,
+      "step": 48885
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.70617142857143e-07,
+      "loss": 0.8868,
+      "step": 48886
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.703485714285713e-07,
+      "loss": 1.0106,
+      "step": 48887
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.700800000000001e-07,
+      "loss": 1.2182,
+      "step": 48888
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.698114285714287e-07,
+      "loss": 1.0063,
+      "step": 48889
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.695428571428571e-07,
+      "loss": 0.947,
+      "step": 48890
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.692742857142857e-07,
+      "loss": 0.8061,
+      "step": 48891
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.690057142857143e-07,
+      "loss": 1.204,
+      "step": 48892
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.68737142857143e-07,
+      "loss": 1.5609,
+      "step": 48893
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.684685714285713e-07,
+      "loss": 1.4391,
+      "step": 48894
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.682000000000001e-07,
+      "loss": 1.2455,
+      "step": 48895
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.679314285714287e-07,
+      "loss": 1.2746,
+      "step": 48896
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.67662857142857e-07,
+      "loss": 1.0285,
+      "step": 48897
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.67394285714286e-07,
+      "loss": 0.8199,
+      "step": 48898
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.671257142857143e-07,
+      "loss": 1.4123,
+      "step": 48899
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.668571428571429e-07,
+      "loss": 1.264,
+      "step": 48900
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.665885714285715e-07,
+      "loss": 1.469,
+      "step": 48901
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.6632e-07,
+      "loss": 0.8786,
+      "step": 48902
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.660514285714285e-07,
+      "loss": 0.9065,
+      "step": 48903
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.65782857142857e-07,
+      "loss": 1.3777,
+      "step": 48904
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.655142857142859e-07,
+      "loss": 1.0417,
+      "step": 48905
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.652457142857143e-07,
+      "loss": 1.4265,
+      "step": 48906
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.649771428571429e-07,
+      "loss": 1.1592,
+      "step": 48907
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.647085714285715e-07,
+      "loss": 1.2435,
+      "step": 48908
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.6444e-07,
+      "loss": 0.5901,
+      "step": 48909
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.641714285714285e-07,
+      "loss": 1.1295,
+      "step": 48910
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.639028571428573e-07,
+      "loss": 0.9535,
+      "step": 48911
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.636342857142859e-07,
+      "loss": 1.0569,
+      "step": 48912
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.633657142857143e-07,
+      "loss": 1.0712,
+      "step": 48913
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.630971428571429e-07,
+      "loss": 0.7756,
+      "step": 48914
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.628285714285715e-07,
+      "loss": 0.8801,
+      "step": 48915
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.6256e-07,
+      "loss": 1.1142,
+      "step": 48916
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.622914285714285e-07,
+      "loss": 1.2392,
+      "step": 48917
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.620228571428573e-07,
+      "loss": 0.776,
+      "step": 48918
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.617542857142857e-07,
+      "loss": 1.2838,
+      "step": 48919
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.614857142857143e-07,
+      "loss": 1.4467,
+      "step": 48920
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.61217142857143e-07,
+      "loss": 0.8257,
+      "step": 48921
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.609485714285715e-07,
+      "loss": 1.1832,
+      "step": 48922
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.6068e-07,
+      "loss": 1.1354,
+      "step": 48923
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.604114285714286e-07,
+      "loss": 1.2087,
+      "step": 48924
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.601428571428572e-07,
+      "loss": 1.037,
+      "step": 48925
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.598742857142856e-07,
+      "loss": 1.1736,
+      "step": 48926
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.596057142857142e-07,
+      "loss": 1.1259,
+      "step": 48927
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.59337142857143e-07,
+      "loss": 1.0561,
+      "step": 48928
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.590685714285714e-07,
+      "loss": 0.6811,
+      "step": 48929
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.588e-07,
+      "loss": 1.3901,
+      "step": 48930
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.585314285714286e-07,
+      "loss": 1.2205,
+      "step": 48931
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.582628571428572e-07,
+      "loss": 1.0103,
+      "step": 48932
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.579942857142856e-07,
+      "loss": 1.0313,
+      "step": 48933
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.577257142857144e-07,
+      "loss": 1.1816,
+      "step": 48934
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.574571428571428e-07,
+      "loss": 1.3374,
+      "step": 48935
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.571885714285714e-07,
+      "loss": 1.3376,
+      "step": 48936
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.5692e-07,
+      "loss": 0.7496,
+      "step": 48937
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.566514285714286e-07,
+      "loss": 0.9685,
+      "step": 48938
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.563828571428572e-07,
+      "loss": 1.0221,
+      "step": 48939
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.561142857142856e-07,
+      "loss": 1.1299,
+      "step": 48940
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.558457142857144e-07,
+      "loss": 1.4876,
+      "step": 48941
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.555771428571428e-07,
+      "loss": 1.0377,
+      "step": 48942
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.553085714285714e-07,
+      "loss": 1.4192,
+      "step": 48943
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.550400000000002e-07,
+      "loss": 1.2629,
+      "step": 48944
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.547714285714286e-07,
+      "loss": 1.4417,
+      "step": 48945
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.545028571428572e-07,
+      "loss": 0.8917,
+      "step": 48946
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.542342857142858e-07,
+      "loss": 0.7596,
+      "step": 48947
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.539657142857144e-07,
+      "loss": 0.8715,
+      "step": 48948
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.536971428571428e-07,
+      "loss": 1.0883,
+      "step": 48949
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.534285714285715e-07,
+      "loss": 1.2223,
+      "step": 48950
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.5316e-07,
+      "loss": 0.7648,
+      "step": 48951
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.528914285714286e-07,
+      "loss": 1.1914,
+      "step": 48952
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.526228571428573e-07,
+      "loss": 1.5757,
+      "step": 48953
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.523542857142857e-07,
+      "loss": 1.4123,
+      "step": 48954
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.520857142857144e-07,
+      "loss": 1.2527,
+      "step": 48955
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.518171428571429e-07,
+      "loss": 1.4293,
+      "step": 48956
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.515485714285715e-07,
+      "loss": 1.0038,
+      "step": 48957
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.5128e-07,
+      "loss": 1.2387,
+      "step": 48958
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.510114285714286e-07,
+      "loss": 0.8182,
+      "step": 48959
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.507428571428573e-07,
+      "loss": 0.8944,
+      "step": 48960
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.504742857142858e-07,
+      "loss": 0.7034,
+      "step": 48961
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.502057142857144e-07,
+      "loss": 1.3238,
+      "step": 48962
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.499371428571429e-07,
+      "loss": 1.1182,
+      "step": 48963
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.496685714285715e-07,
+      "loss": 1.455,
+      "step": 48964
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.494e-07,
+      "loss": 0.9408,
+      "step": 48965
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.491314285714287e-07,
+      "loss": 0.6947,
+      "step": 48966
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.488628571428571e-07,
+      "loss": 1.3383,
+      "step": 48967
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.485942857142858e-07,
+      "loss": 1.1626,
+      "step": 48968
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.483257142857144e-07,
+      "loss": 1.0111,
+      "step": 48969
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.480571428571429e-07,
+      "loss": 1.1736,
+      "step": 48970
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.477885714285716e-07,
+      "loss": 1.2507,
+      "step": 48971
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.4752e-07,
+      "loss": 1.0124,
+      "step": 48972
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.472514285714287e-07,
+      "loss": 1.3001,
+      "step": 48973
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.46982857142857e-07,
+      "loss": 0.7942,
+      "step": 48974
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.467142857142857e-07,
+      "loss": 1.5428,
+      "step": 48975
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.464457142857145e-07,
+      "loss": 1.1296,
+      "step": 48976
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.461771428571428e-07,
+      "loss": 1.355,
+      "step": 48977
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.459085714285715e-07,
+      "loss": 1.0074,
+      "step": 48978
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.456399999999999e-07,
+      "loss": 1.0071,
+      "step": 48979
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.453714285714286e-07,
+      "loss": 1.3562,
+      "step": 48980
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.451028571428571e-07,
+      "loss": 1.2075,
+      "step": 48981
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.448342857142857e-07,
+      "loss": 0.8202,
+      "step": 48982
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.445657142857142e-07,
+      "loss": 1.2366,
+      "step": 48983
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.442971428571428e-07,
+      "loss": 1.1969,
+      "step": 48984
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.440285714285715e-07,
+      "loss": 1.6574,
+      "step": 48985
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.4376e-07,
+      "loss": 1.0525,
+      "step": 48986
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.434914285714286e-07,
+      "loss": 1.1784,
+      "step": 48987
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.432228571428571e-07,
+      "loss": 0.7908,
+      "step": 48988
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.429542857142858e-07,
+      "loss": 1.1556,
+      "step": 48989
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.426857142857142e-07,
+      "loss": 0.8382,
+      "step": 48990
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.424171428571429e-07,
+      "loss": 1.1822,
+      "step": 48991
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.421485714285715e-07,
+      "loss": 1.0586,
+      "step": 48992
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.4188e-07,
+      "loss": 0.7517,
+      "step": 48993
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.416114285714287e-07,
+      "loss": 1.3166,
+      "step": 48994
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.413428571428571e-07,
+      "loss": 0.9749,
+      "step": 48995
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.410742857142858e-07,
+      "loss": 1.442,
+      "step": 48996
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.408057142857142e-07,
+      "loss": 0.8367,
+      "step": 48997
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.405371428571429e-07,
+      "loss": 1.086,
+      "step": 48998
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.402685714285716e-07,
+      "loss": 0.7325,
+      "step": 48999
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.4e-07,
+      "loss": 1.005,
+      "step": 49000
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.397314285714287e-07,
+      "loss": 0.9886,
+      "step": 49001
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.394628571428571e-07,
+      "loss": 1.2493,
+      "step": 49002
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.391942857142858e-07,
+      "loss": 1.0214,
+      "step": 49003
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.389257142857143e-07,
+      "loss": 0.8273,
+      "step": 49004
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.386571428571429e-07,
+      "loss": 1.0298,
+      "step": 49005
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.383885714285714e-07,
+      "loss": 0.9315,
+      "step": 49006
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.3812e-07,
+      "loss": 0.8497,
+      "step": 49007
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.378514285714287e-07,
+      "loss": 1.603,
+      "step": 49008
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.375828571428572e-07,
+      "loss": 1.4662,
+      "step": 49009
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.373142857142858e-07,
+      "loss": 0.8437,
+      "step": 49010
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.370457142857143e-07,
+      "loss": 1.2655,
+      "step": 49011
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.367771428571429e-07,
+      "loss": 1.433,
+      "step": 49012
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.365085714285714e-07,
+      "loss": 0.7614,
+      "step": 49013
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.362400000000001e-07,
+      "loss": 1.0176,
+      "step": 49014
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.359714285714287e-07,
+      "loss": 1.3233,
+      "step": 49015
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.357028571428572e-07,
+      "loss": 1.2812,
+      "step": 49016
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.354342857142858e-07,
+      "loss": 1.1357,
+      "step": 49017
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.351657142857143e-07,
+      "loss": 1.1216,
+      "step": 49018
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.34897142857143e-07,
+      "loss": 1.0633,
+      "step": 49019
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.346285714285714e-07,
+      "loss": 1.1626,
+      "step": 49020
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.343600000000001e-07,
+      "loss": 0.8882,
+      "step": 49021
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.340914285714286e-07,
+      "loss": 1.106,
+      "step": 49022
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.338228571428572e-07,
+      "loss": 1.1282,
+      "step": 49023
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.335542857142859e-07,
+      "loss": 0.9672,
+      "step": 49024
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.332857142857142e-07,
+      "loss": 0.89,
+      "step": 49025
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.33017142857143e-07,
+      "loss": 0.9025,
+      "step": 49026
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.327485714285714e-07,
+      "loss": 1.2171,
+      "step": 49027
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.3248e-07,
+      "loss": 1.195,
+      "step": 49028
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.322114285714285e-07,
+      "loss": 0.6792,
+      "step": 49029
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.319428571428571e-07,
+      "loss": 0.69,
+      "step": 49030
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.316742857142858e-07,
+      "loss": 1.0151,
+      "step": 49031
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.314057142857143e-07,
+      "loss": 1.3964,
+      "step": 49032
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.311371428571429e-07,
+      "loss": 1.0004,
+      "step": 49033
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.308685714285714e-07,
+      "loss": 0.9613,
+      "step": 49034
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.306e-07,
+      "loss": 1.1629,
+      "step": 49035
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.303314285714285e-07,
+      "loss": 1.6158,
+      "step": 49036
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.300628571428572e-07,
+      "loss": 0.9024,
+      "step": 49037
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.297942857142856e-07,
+      "loss": 1.2345,
+      "step": 49038
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.295257142857143e-07,
+      "loss": 1.2376,
+      "step": 49039
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.292571428571429e-07,
+      "loss": 1.0908,
+      "step": 49040
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.289885714285714e-07,
+      "loss": 1.0278,
+      "step": 49041
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.287200000000001e-07,
+      "loss": 1.2247,
+      "step": 49042
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.284514285714285e-07,
+      "loss": 0.9356,
+      "step": 49043
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.281828571428572e-07,
+      "loss": 1.3065,
+      "step": 49044
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.279142857142857e-07,
+      "loss": 1.0068,
+      "step": 49045
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.276457142857143e-07,
+      "loss": 1.1945,
+      "step": 49046
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.27377142857143e-07,
+      "loss": 0.7892,
+      "step": 49047
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.271085714285714e-07,
+      "loss": 1.4283,
+      "step": 49048
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.268400000000001e-07,
+      "loss": 1.0344,
+      "step": 49049
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.265714285714286e-07,
+      "loss": 1.303,
+      "step": 49050
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.263028571428572e-07,
+      "loss": 0.932,
+      "step": 49051
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.260342857142857e-07,
+      "loss": 0.9498,
+      "step": 49052
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.257657142857143e-07,
+      "loss": 0.8299,
+      "step": 49053
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.254971428571428e-07,
+      "loss": 1.4752,
+      "step": 49054
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.252285714285715e-07,
+      "loss": 1.3462,
+      "step": 49055
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.249600000000001e-07,
+      "loss": 1.4332,
+      "step": 49056
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.246914285714286e-07,
+      "loss": 0.9724,
+      "step": 49057
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.244228571428572e-07,
+      "loss": 1.3494,
+      "step": 49058
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.241542857142857e-07,
+      "loss": 1.4546,
+      "step": 49059
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.238857142857144e-07,
+      "loss": 1.1126,
+      "step": 49060
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.236171428571428e-07,
+      "loss": 0.8178,
+      "step": 49061
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.233485714285715e-07,
+      "loss": 0.951,
+      "step": 49062
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.230800000000001e-07,
+      "loss": 1.6427,
+      "step": 49063
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.228114285714286e-07,
+      "loss": 1.3823,
+      "step": 49064
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.225428571428573e-07,
+      "loss": 1.0397,
+      "step": 49065
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.222742857142857e-07,
+      "loss": 1.4857,
+      "step": 49066
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.220057142857144e-07,
+      "loss": 1.1769,
+      "step": 49067
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.217371428571429e-07,
+      "loss": 1.2502,
+      "step": 49068
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.214685714285715e-07,
+      "loss": 1.0103,
+      "step": 49069
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.212e-07,
+      "loss": 1.4689,
+      "step": 49070
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.209314285714286e-07,
+      "loss": 0.848,
+      "step": 49071
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.206628571428573e-07,
+      "loss": 1.3664,
+      "step": 49072
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.203942857142858e-07,
+      "loss": 0.7474,
+      "step": 49073
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.201257142857144e-07,
+      "loss": 1.4643,
+      "step": 49074
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.198571428571429e-07,
+      "loss": 0.903,
+      "step": 49075
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.195885714285715e-07,
+      "loss": 1.4742,
+      "step": 49076
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.1932e-07,
+      "loss": 1.526,
+      "step": 49077
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.190514285714287e-07,
+      "loss": 1.3959,
+      "step": 49078
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.187828571428573e-07,
+      "loss": 1.1985,
+      "step": 49079
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.185142857142857e-07,
+      "loss": 1.165,
+      "step": 49080
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.182457142857143e-07,
+      "loss": 0.9461,
+      "step": 49081
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.179771428571428e-07,
+      "loss": 0.7953,
+      "step": 49082
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.177085714285715e-07,
+      "loss": 0.8475,
+      "step": 49083
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.174399999999999e-07,
+      "loss": 0.9393,
+      "step": 49084
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.171714285714286e-07,
+      "loss": 0.788,
+      "step": 49085
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.16902857142857e-07,
+      "loss": 1.3888,
+      "step": 49086
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.166342857142857e-07,
+      "loss": 0.9813,
+      "step": 49087
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.163657142857144e-07,
+      "loss": 0.8895,
+      "step": 49088
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.160971428571428e-07,
+      "loss": 0.9607,
+      "step": 49089
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.158285714285715e-07,
+      "loss": 1.3495,
+      "step": 49090
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.155599999999999e-07,
+      "loss": 0.8936,
+      "step": 49091
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.152914285714286e-07,
+      "loss": 0.6592,
+      "step": 49092
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.150228571428571e-07,
+      "loss": 0.9481,
+      "step": 49093
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.147542857142857e-07,
+      "loss": 0.759,
+      "step": 49094
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.144857142857144e-07,
+      "loss": 0.8484,
+      "step": 49095
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.142171428571428e-07,
+      "loss": 1.2222,
+      "step": 49096
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.139485714285715e-07,
+      "loss": 1.2457,
+      "step": 49097
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.1368e-07,
+      "loss": 1.1195,
+      "step": 49098
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.134114285714286e-07,
+      "loss": 0.9011,
+      "step": 49099
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.131428571428571e-07,
+      "loss": 1.5451,
+      "step": 49100
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.128742857142858e-07,
+      "loss": 1.2327,
+      "step": 49101
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.126057142857144e-07,
+      "loss": 0.9524,
+      "step": 49102
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.123371428571429e-07,
+      "loss": 1.0173,
+      "step": 49103
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.120685714285715e-07,
+      "loss": 1.0228,
+      "step": 49104
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.118e-07,
+      "loss": 0.9164,
+      "step": 49105
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.115314285714287e-07,
+      "loss": 1.1603,
+      "step": 49106
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.112628571428571e-07,
+      "loss": 1.0568,
+      "step": 49107
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.109942857142858e-07,
+      "loss": 1.3305,
+      "step": 49108
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.107257142857142e-07,
+      "loss": 1.0079,
+      "step": 49109
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.104571428571429e-07,
+      "loss": 1.1885,
+      "step": 49110
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.101885714285716e-07,
+      "loss": 0.9788,
+      "step": 49111
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.0992e-07,
+      "loss": 1.271,
+      "step": 49112
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.096514285714287e-07,
+      "loss": 0.9243,
+      "step": 49113
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.093828571428571e-07,
+      "loss": 1.3263,
+      "step": 49114
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.091142857142858e-07,
+      "loss": 1.29,
+      "step": 49115
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.088457142857143e-07,
+      "loss": 0.7527,
+      "step": 49116
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.085771428571429e-07,
+      "loss": 0.9652,
+      "step": 49117
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.083085714285716e-07,
+      "loss": 1.0994,
+      "step": 49118
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.0804e-07,
+      "loss": 1.3109,
+      "step": 49119
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.077714285714287e-07,
+      "loss": 1.0855,
+      "step": 49120
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.075028571428572e-07,
+      "loss": 1.164,
+      "step": 49121
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.072342857142858e-07,
+      "loss": 1.6773,
+      "step": 49122
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.069657142857143e-07,
+      "loss": 1.0266,
+      "step": 49123
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.066971428571429e-07,
+      "loss": 1.5079,
+      "step": 49124
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.064285714285714e-07,
+      "loss": 1.1841,
+      "step": 49125
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.061600000000001e-07,
+      "loss": 0.744,
+      "step": 49126
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.058914285714287e-07,
+      "loss": 1.0215,
+      "step": 49127
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.056228571428572e-07,
+      "loss": 0.9765,
+      "step": 49128
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.053542857142858e-07,
+      "loss": 0.8396,
+      "step": 49129
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.050857142857142e-07,
+      "loss": 1.3519,
+      "step": 49130
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.04817142857143e-07,
+      "loss": 1.2387,
+      "step": 49131
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.045485714285713e-07,
+      "loss": 1.0455,
+      "step": 49132
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.0428e-07,
+      "loss": 1.0999,
+      "step": 49133
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.040114285714286e-07,
+      "loss": 1.1523,
+      "step": 49134
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.037428571428571e-07,
+      "loss": 1.0834,
+      "step": 49135
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.034742857142858e-07,
+      "loss": 1.6477,
+      "step": 49136
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.032057142857142e-07,
+      "loss": 1.0373,
+      "step": 49137
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.029371428571429e-07,
+      "loss": 0.985,
+      "step": 49138
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.026685714285714e-07,
+      "loss": 0.7489,
+      "step": 49139
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.024e-07,
+      "loss": 1.1343,
+      "step": 49140
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.021314285714285e-07,
+      "loss": 1.0229,
+      "step": 49141
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.018628571428571e-07,
+      "loss": 1.3171,
+      "step": 49142
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.015942857142858e-07,
+      "loss": 0.9584,
+      "step": 49143
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.013257142857143e-07,
+      "loss": 0.9244,
+      "step": 49144
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.010571428571429e-07,
+      "loss": 1.0114,
+      "step": 49145
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.007885714285714e-07,
+      "loss": 1.3438,
+      "step": 49146
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.0052e-07,
+      "loss": 1.2382,
+      "step": 49147
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.002514285714285e-07,
+      "loss": 1.0282,
+      "step": 49148
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.999828571428572e-07,
+      "loss": 0.9628,
+      "step": 49149
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.997142857142858e-07,
+      "loss": 1.7007,
+      "step": 49150
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.994457142857143e-07,
+      "loss": 1.0565,
+      "step": 49151
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.991771428571429e-07,
+      "loss": 0.9478,
+      "step": 49152
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.989085714285714e-07,
+      "loss": 1.1281,
+      "step": 49153
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.986400000000001e-07,
+      "loss": 1.0592,
+      "step": 49154
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.983714285714285e-07,
+      "loss": 0.9307,
+      "step": 49155
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.981028571428572e-07,
+      "loss": 0.8855,
+      "step": 49156
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.978342857142857e-07,
+      "loss": 1.1902,
+      "step": 49157
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.975657142857143e-07,
+      "loss": 0.8735,
+      "step": 49158
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.97297142857143e-07,
+      "loss": 1.1281,
+      "step": 49159
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.970285714285714e-07,
+      "loss": 1.3118,
+      "step": 49160
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.967600000000001e-07,
+      "loss": 1.0412,
+      "step": 49161
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.964914285714286e-07,
+      "loss": 1.2505,
+      "step": 49162
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.962228571428572e-07,
+      "loss": 1.4586,
+      "step": 49163
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.959542857142857e-07,
+      "loss": 1.2954,
+      "step": 49164
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.956857142857143e-07,
+      "loss": 0.9991,
+      "step": 49165
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.95417142857143e-07,
+      "loss": 0.7273,
+      "step": 49166
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.951485714285715e-07,
+      "loss": 0.6613,
+      "step": 49167
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.948800000000001e-07,
+      "loss": 0.9671,
+      "step": 49168
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.946114285714286e-07,
+      "loss": 0.9403,
+      "step": 49169
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.943428571428572e-07,
+      "loss": 0.7038,
+      "step": 49170
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.940742857142857e-07,
+      "loss": 1.0099,
+      "step": 49171
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.938057142857144e-07,
+      "loss": 1.3542,
+      "step": 49172
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.935371428571428e-07,
+      "loss": 1.3671,
+      "step": 49173
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.932685714285715e-07,
+      "loss": 1.1367,
+      "step": 49174
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.930000000000001e-07,
+      "loss": 1.0475,
+      "step": 49175
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.927314285714286e-07,
+      "loss": 1.1211,
+      "step": 49176
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.924628571428573e-07,
+      "loss": 1.2575,
+      "step": 49177
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.921942857142857e-07,
+      "loss": 1.182,
+      "step": 49178
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.919257142857144e-07,
+      "loss": 1.1736,
+      "step": 49179
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.916571428571429e-07,
+      "loss": 1.1099,
+      "step": 49180
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.913885714285715e-07,
+      "loss": 1.0941,
+      "step": 49181
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.911200000000002e-07,
+      "loss": 1.3161,
+      "step": 49182
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.908514285714285e-07,
+      "loss": 1.1531,
+      "step": 49183
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.905828571428573e-07,
+      "loss": 1.3351,
+      "step": 49184
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.903142857142857e-07,
+      "loss": 1.2197,
+      "step": 49185
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.900457142857143e-07,
+      "loss": 0.923,
+      "step": 49186
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.897771428571428e-07,
+      "loss": 1.2739,
+      "step": 49187
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.895085714285714e-07,
+      "loss": 0.8972,
+      "step": 49188
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.892399999999999e-07,
+      "loss": 0.9156,
+      "step": 49189
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.889714285714286e-07,
+      "loss": 0.786,
+      "step": 49190
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.887028571428572e-07,
+      "loss": 1.4989,
+      "step": 49191
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.884342857142857e-07,
+      "loss": 0.9423,
+      "step": 49192
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.881657142857143e-07,
+      "loss": 1.0932,
+      "step": 49193
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.878971428571428e-07,
+      "loss": 1.0114,
+      "step": 49194
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.876285714285715e-07,
+      "loss": 1.268,
+      "step": 49195
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.873599999999999e-07,
+      "loss": 0.7506,
+      "step": 49196
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.870914285714286e-07,
+      "loss": 1.4444,
+      "step": 49197
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.868228571428572e-07,
+      "loss": 1.3534,
+      "step": 49198
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.865542857142857e-07,
+      "loss": 1.1108,
+      "step": 49199
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.862857142857144e-07,
+      "loss": 1.1479,
+      "step": 49200
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.860171428571428e-07,
+      "loss": 1.3162,
+      "step": 49201
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.857485714285715e-07,
+      "loss": 1.0784,
+      "step": 49202
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.854799999999999e-07,
+      "loss": 1.267,
+      "step": 49203
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.852114285714286e-07,
+      "loss": 1.2256,
+      "step": 49204
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.849428571428573e-07,
+      "loss": 0.7598,
+      "step": 49205
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.846742857142857e-07,
+      "loss": 1.1883,
+      "step": 49206
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.844057142857144e-07,
+      "loss": 1.1565,
+      "step": 49207
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.841371428571429e-07,
+      "loss": 1.1822,
+      "step": 49208
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.838685714285715e-07,
+      "loss": 0.7348,
+      "step": 49209
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.836e-07,
+      "loss": 1.02,
+      "step": 49210
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.833314285714286e-07,
+      "loss": 0.8598,
+      "step": 49211
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.830628571428571e-07,
+      "loss": 0.9357,
+      "step": 49212
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.827942857142858e-07,
+      "loss": 1.2018,
+      "step": 49213
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.825257142857144e-07,
+      "loss": 0.9609,
+      "step": 49214
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.822571428571429e-07,
+      "loss": 1.282,
+      "step": 49215
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.819885714285715e-07,
+      "loss": 1.2286,
+      "step": 49216
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.8172e-07,
+      "loss": 1.0566,
+      "step": 49217
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.814514285714287e-07,
+      "loss": 0.7089,
+      "step": 49218
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.811828571428572e-07,
+      "loss": 1.0801,
+      "step": 49219
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.809142857142858e-07,
+      "loss": 0.7003,
+      "step": 49220
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.806457142857143e-07,
+      "loss": 1.2639,
+      "step": 49221
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.803771428571429e-07,
+      "loss": 1.1226,
+      "step": 49222
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.801085714285715e-07,
+      "loss": 1.0118,
+      "step": 49223
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.7984e-07,
+      "loss": 0.8237,
+      "step": 49224
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.795714285714287e-07,
+      "loss": 1.0618,
+      "step": 49225
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.793028571428572e-07,
+      "loss": 1.3062,
+      "step": 49226
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.790342857142858e-07,
+      "loss": 1.0836,
+      "step": 49227
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.787657142857144e-07,
+      "loss": 0.5114,
+      "step": 49228
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.784971428571429e-07,
+      "loss": 1.4278,
+      "step": 49229
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.782285714285715e-07,
+      "loss": 1.4994,
+      "step": 49230
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.7796e-07,
+      "loss": 1.3083,
+      "step": 49231
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.776914285714286e-07,
+      "loss": 0.9154,
+      "step": 49232
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.774228571428573e-07,
+      "loss": 0.8667,
+      "step": 49233
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.771542857142858e-07,
+      "loss": 1.0752,
+      "step": 49234
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.768857142857144e-07,
+      "loss": 1.2606,
+      "step": 49235
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.766171428571428e-07,
+      "loss": 1.0482,
+      "step": 49236
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.763485714285714e-07,
+      "loss": 1.2979,
+      "step": 49237
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.7608e-07,
+      "loss": 1.1577,
+      "step": 49238
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.758114285714285e-07,
+      "loss": 1.3622,
+      "step": 49239
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.755428571428571e-07,
+      "loss": 0.9874,
+      "step": 49240
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.752742857142857e-07,
+      "loss": 1.0273,
+      "step": 49241
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.750057142857143e-07,
+      "loss": 1.1737,
+      "step": 49242
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.747371428571429e-07,
+      "loss": 1.1454,
+      "step": 49243
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.744685714285714e-07,
+      "loss": 1.3866,
+      "step": 49244
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.742e-07,
+      "loss": 0.8415,
+      "step": 49245
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.739314285714286e-07,
+      "loss": 1.1293,
+      "step": 49246
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.736628571428571e-07,
+      "loss": 1.249,
+      "step": 49247
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.733942857142857e-07,
+      "loss": 0.8295,
+      "step": 49248
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.731257142857143e-07,
+      "loss": 0.6441,
+      "step": 49249
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.728571428571429e-07,
+      "loss": 0.9354,
+      "step": 49250
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.725885714285715e-07,
+      "loss": 0.9979,
+      "step": 49251
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.7232e-07,
+      "loss": 1.1961,
+      "step": 49252
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.720514285714286e-07,
+      "loss": 1.0314,
+      "step": 49253
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.717828571428571e-07,
+      "loss": 1.1034,
+      "step": 49254
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.715142857142857e-07,
+      "loss": 0.8431,
+      "step": 49255
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.712457142857143e-07,
+      "loss": 1.7261,
+      "step": 49256
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.709771428571429e-07,
+      "loss": 1.0709,
+      "step": 49257
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.707085714285715e-07,
+      "loss": 1.2228,
+      "step": 49258
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.7044e-07,
+      "loss": 0.7328,
+      "step": 49259
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.701714285714286e-07,
+      "loss": 1.1413,
+      "step": 49260
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.699028571428572e-07,
+      "loss": 1.1972,
+      "step": 49261
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.696342857142857e-07,
+      "loss": 0.6894,
+      "step": 49262
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.693657142857143e-07,
+      "loss": 1.2357,
+      "step": 49263
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.690971428571429e-07,
+      "loss": 1.0529,
+      "step": 49264
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.688285714285715e-07,
+      "loss": 1.0808,
+      "step": 49265
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.685600000000001e-07,
+      "loss": 0.9088,
+      "step": 49266
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.682914285714286e-07,
+      "loss": 1.0727,
+      "step": 49267
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.680228571428572e-07,
+      "loss": 0.7461,
+      "step": 49268
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.677542857142858e-07,
+      "loss": 1.0633,
+      "step": 49269
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.674857142857143e-07,
+      "loss": 1.1085,
+      "step": 49270
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.672171428571429e-07,
+      "loss": 1.1572,
+      "step": 49271
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.669485714285714e-07,
+      "loss": 0.9198,
+      "step": 49272
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.666800000000001e-07,
+      "loss": 1.1984,
+      "step": 49273
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.664114285714287e-07,
+      "loss": 1.3684,
+      "step": 49274
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.661428571428572e-07,
+      "loss": 0.8654,
+      "step": 49275
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.658742857142858e-07,
+      "loss": 0.9702,
+      "step": 49276
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.656057142857143e-07,
+      "loss": 0.9689,
+      "step": 49277
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.653371428571429e-07,
+      "loss": 1.0972,
+      "step": 49278
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.650685714285715e-07,
+      "loss": 0.7916,
+      "step": 49279
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.648e-07,
+      "loss": 0.8675,
+      "step": 49280
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.645314285714287e-07,
+      "loss": 1.2485,
+      "step": 49281
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.642628571428572e-07,
+      "loss": 0.9323,
+      "step": 49282
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.639942857142858e-07,
+      "loss": 0.9443,
+      "step": 49283
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.637257142857144e-07,
+      "loss": 1.4367,
+      "step": 49284
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.634571428571429e-07,
+      "loss": 1.7423,
+      "step": 49285
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.631885714285715e-07,
+      "loss": 1.3035,
+      "step": 49286
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.6292e-07,
+      "loss": 0.8241,
+      "step": 49287
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.626514285714285e-07,
+      "loss": 1.0345,
+      "step": 49288
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.623828571428573e-07,
+      "loss": 0.7518,
+      "step": 49289
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.621142857142857e-07,
+      "loss": 1.1522,
+      "step": 49290
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.618457142857143e-07,
+      "loss": 1.4016,
+      "step": 49291
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.615771428571428e-07,
+      "loss": 0.9123,
+      "step": 49292
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.613085714285714e-07,
+      "loss": 1.1524,
+      "step": 49293
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.6104e-07,
+      "loss": 1.0537,
+      "step": 49294
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.607714285714285e-07,
+      "loss": 0.8781,
+      "step": 49295
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.605028571428572e-07,
+      "loss": 1.5157,
+      "step": 49296
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.602342857142857e-07,
+      "loss": 0.9031,
+      "step": 49297
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.599657142857143e-07,
+      "loss": 0.7934,
+      "step": 49298
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.596971428571429e-07,
+      "loss": 1.0593,
+      "step": 49299
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.594285714285714e-07,
+      "loss": 0.8858,
+      "step": 49300
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.5916e-07,
+      "loss": 0.9484,
+      "step": 49301
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.588914285714286e-07,
+      "loss": 1.0648,
+      "step": 49302
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.586228571428571e-07,
+      "loss": 1.3639,
+      "step": 49303
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.583542857142858e-07,
+      "loss": 1.0978,
+      "step": 49304
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.580857142857143e-07,
+      "loss": 0.93,
+      "step": 49305
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.578171428571429e-07,
+      "loss": 0.9952,
+      "step": 49306
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.575485714285715e-07,
+      "loss": 0.9297,
+      "step": 49307
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.5728e-07,
+      "loss": 1.1032,
+      "step": 49308
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.570114285714286e-07,
+      "loss": 1.2129,
+      "step": 49309
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.567428571428571e-07,
+      "loss": 1.3285,
+      "step": 49310
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.564742857142857e-07,
+      "loss": 1.3931,
+      "step": 49311
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.562057142857144e-07,
+      "loss": 1.377,
+      "step": 49312
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.559371428571429e-07,
+      "loss": 0.9992,
+      "step": 49313
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.556685714285715e-07,
+      "loss": 0.7363,
+      "step": 49314
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.554e-07,
+      "loss": 1.372,
+      "step": 49315
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.551314285714286e-07,
+      "loss": 0.6917,
+      "step": 49316
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.548628571428572e-07,
+      "loss": 1.1444,
+      "step": 49317
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.545942857142857e-07,
+      "loss": 1.0392,
+      "step": 49318
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.543257142857143e-07,
+      "loss": 1.0333,
+      "step": 49319
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.540571428571429e-07,
+      "loss": 1.0114,
+      "step": 49320
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.537885714285715e-07,
+      "loss": 1.3485,
+      "step": 49321
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.535200000000001e-07,
+      "loss": 1.3155,
+      "step": 49322
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.532514285714286e-07,
+      "loss": 1.0101,
+      "step": 49323
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.529828571428572e-07,
+      "loss": 1.3065,
+      "step": 49324
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.527142857142858e-07,
+      "loss": 1.0719,
+      "step": 49325
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.524457142857143e-07,
+      "loss": 1.1332,
+      "step": 49326
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.521771428571429e-07,
+      "loss": 0.9785,
+      "step": 49327
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.519085714285715e-07,
+      "loss": 1.3901,
+      "step": 49328
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.516400000000001e-07,
+      "loss": 1.3845,
+      "step": 49329
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.513714285714287e-07,
+      "loss": 1.0341,
+      "step": 49330
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.511028571428572e-07,
+      "loss": 1.389,
+      "step": 49331
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.508342857142858e-07,
+      "loss": 1.3219,
+      "step": 49332
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.505657142857143e-07,
+      "loss": 1.0622,
+      "step": 49333
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.502971428571429e-07,
+      "loss": 1.2173,
+      "step": 49334
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.500285714285715e-07,
+      "loss": 0.8865,
+      "step": 49335
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.497600000000001e-07,
+      "loss": 1.0376,
+      "step": 49336
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.494914285714287e-07,
+      "loss": 1.3402,
+      "step": 49337
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.492228571428572e-07,
+      "loss": 0.8991,
+      "step": 49338
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.489542857142858e-07,
+      "loss": 1.1582,
+      "step": 49339
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.486857142857144e-07,
+      "loss": 1.3348,
+      "step": 49340
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.484171428571428e-07,
+      "loss": 1.3368,
+      "step": 49341
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.481485714285714e-07,
+      "loss": 1.3566,
+      "step": 49342
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.478799999999999e-07,
+      "loss": 1.066,
+      "step": 49343
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.476114285714286e-07,
+      "loss": 1.2074,
+      "step": 49344
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.473428571428572e-07,
+      "loss": 1.018,
+      "step": 49345
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.470742857142857e-07,
+      "loss": 1.1297,
+      "step": 49346
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.468057142857143e-07,
+      "loss": 0.6289,
+      "step": 49347
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.465371428571428e-07,
+      "loss": 0.8057,
+      "step": 49348
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.462685714285714e-07,
+      "loss": 1.2946,
+      "step": 49349
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.46e-07,
+      "loss": 1.1387,
+      "step": 49350
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.457314285714285e-07,
+      "loss": 1.513,
+      "step": 49351
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.454628571428572e-07,
+      "loss": 1.1635,
+      "step": 49352
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.451942857142857e-07,
+      "loss": 1.0374,
+      "step": 49353
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.449257142857143e-07,
+      "loss": 1.4773,
+      "step": 49354
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.446571428571429e-07,
+      "loss": 1.2084,
+      "step": 49355
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.443885714285714e-07,
+      "loss": 0.9178,
+      "step": 49356
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.4412e-07,
+      "loss": 1.2447,
+      "step": 49357
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.438514285714286e-07,
+      "loss": 1.329,
+      "step": 49358
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.435828571428571e-07,
+      "loss": 1.3625,
+      "step": 49359
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.433142857142858e-07,
+      "loss": 0.9959,
+      "step": 49360
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.430457142857143e-07,
+      "loss": 1.3509,
+      "step": 49361
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.427771428571429e-07,
+      "loss": 0.9551,
+      "step": 49362
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.425085714285715e-07,
+      "loss": 1.1439,
+      "step": 49363
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.4224e-07,
+      "loss": 1.0985,
+      "step": 49364
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.419714285714286e-07,
+      "loss": 1.2479,
+      "step": 49365
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.417028571428571e-07,
+      "loss": 1.1532,
+      "step": 49366
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.414342857142857e-07,
+      "loss": 0.986,
+      "step": 49367
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.411657142857144e-07,
+      "loss": 1.139,
+      "step": 49368
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.408971428571429e-07,
+      "loss": 1.2481,
+      "step": 49369
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.406285714285715e-07,
+      "loss": 1.1999,
+      "step": 49370
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.4036e-07,
+      "loss": 0.9469,
+      "step": 49371
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.400914285714286e-07,
+      "loss": 1.1802,
+      "step": 49372
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.398228571428572e-07,
+      "loss": 0.8616,
+      "step": 49373
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.395542857142857e-07,
+      "loss": 0.8385,
+      "step": 49374
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.392857142857143e-07,
+      "loss": 1.2037,
+      "step": 49375
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.390171428571429e-07,
+      "loss": 1.0025,
+      "step": 49376
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.387485714285715e-07,
+      "loss": 1.4246,
+      "step": 49377
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.384800000000001e-07,
+      "loss": 0.9735,
+      "step": 49378
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.382114285714286e-07,
+      "loss": 0.7067,
+      "step": 49379
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.379428571428572e-07,
+      "loss": 1.0819,
+      "step": 49380
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.376742857142858e-07,
+      "loss": 0.8571,
+      "step": 49381
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.374057142857143e-07,
+      "loss": 1.1086,
+      "step": 49382
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.371371428571429e-07,
+      "loss": 1.2871,
+      "step": 49383
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.368685714285715e-07,
+      "loss": 0.8488,
+      "step": 49384
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.366000000000001e-07,
+      "loss": 1.2036,
+      "step": 49385
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.363314285714287e-07,
+      "loss": 1.3675,
+      "step": 49386
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.360628571428572e-07,
+      "loss": 1.1855,
+      "step": 49387
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.357942857142858e-07,
+      "loss": 1.0977,
+      "step": 49388
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.355257142857143e-07,
+      "loss": 1.0898,
+      "step": 49389
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.352571428571429e-07,
+      "loss": 1.0419,
+      "step": 49390
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.349885714285715e-07,
+      "loss": 1.1192,
+      "step": 49391
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.3472e-07,
+      "loss": 1.1912,
+      "step": 49392
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.344514285714287e-07,
+      "loss": 1.1542,
+      "step": 49393
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.341828571428571e-07,
+      "loss": 1.1148,
+      "step": 49394
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.339142857142857e-07,
+      "loss": 1.4467,
+      "step": 49395
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.336457142857143e-07,
+      "loss": 1.5071,
+      "step": 49396
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.333771428571428e-07,
+      "loss": 0.7386,
+      "step": 49397
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.331085714285714e-07,
+      "loss": 1.4731,
+      "step": 49398
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.3284e-07,
+      "loss": 0.964,
+      "step": 49399
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.325714285714286e-07,
+      "loss": 0.3944,
+      "step": 49400
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.323028571428572e-07,
+      "loss": 1.0426,
+      "step": 49401
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.320342857142857e-07,
+      "loss": 0.9567,
+      "step": 49402
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.317657142857143e-07,
+      "loss": 1.082,
+      "step": 49403
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.314971428571428e-07,
+      "loss": 1.3038,
+      "step": 49404
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.312285714285714e-07,
+      "loss": 0.8511,
+      "step": 49405
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.3096e-07,
+      "loss": 0.8191,
+      "step": 49406
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.306914285714286e-07,
+      "loss": 1.0046,
+      "step": 49407
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.304228571428572e-07,
+      "loss": 1.3366,
+      "step": 49408
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.301542857142857e-07,
+      "loss": 1.1287,
+      "step": 49409
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.298857142857143e-07,
+      "loss": 1.3622,
+      "step": 49410
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.296171428571429e-07,
+      "loss": 1.0832,
+      "step": 49411
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.293485714285714e-07,
+      "loss": 1.9765,
+      "step": 49412
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.2908e-07,
+      "loss": 0.9137,
+      "step": 49413
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.288114285714286e-07,
+      "loss": 0.9878,
+      "step": 49414
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.285428571428572e-07,
+      "loss": 1.6032,
+      "step": 49415
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.282742857142858e-07,
+      "loss": 1.0565,
+      "step": 49416
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.280057142857143e-07,
+      "loss": 0.8023,
+      "step": 49417
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.277371428571429e-07,
+      "loss": 1.285,
+      "step": 49418
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.274685714285715e-07,
+      "loss": 1.1491,
+      "step": 49419
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.272e-07,
+      "loss": 1.1494,
+      "step": 49420
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.269314285714286e-07,
+      "loss": 1.0588,
+      "step": 49421
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.266628571428571e-07,
+      "loss": 0.9236,
+      "step": 49422
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.263942857142858e-07,
+      "loss": 0.908,
+      "step": 49423
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.261257142857144e-07,
+      "loss": 1.2492,
+      "step": 49424
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.258571428571429e-07,
+      "loss": 1.2847,
+      "step": 49425
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.255885714285715e-07,
+      "loss": 0.9739,
+      "step": 49426
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.2532e-07,
+      "loss": 1.0832,
+      "step": 49427
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.250514285714286e-07,
+      "loss": 1.1913,
+      "step": 49428
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.247828571428572e-07,
+      "loss": 0.7745,
+      "step": 49429
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.245142857142857e-07,
+      "loss": 1.2028,
+      "step": 49430
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.242457142857144e-07,
+      "loss": 1.1297,
+      "step": 49431
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.239771428571429e-07,
+      "loss": 1.5304,
+      "step": 49432
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.237085714285715e-07,
+      "loss": 1.2222,
+      "step": 49433
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.234400000000001e-07,
+      "loss": 1.2008,
+      "step": 49434
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.231714285714286e-07,
+      "loss": 0.5153,
+      "step": 49435
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.229028571428572e-07,
+      "loss": 1.1285,
+      "step": 49436
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.226342857142858e-07,
+      "loss": 1.292,
+      "step": 49437
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.223657142857143e-07,
+      "loss": 0.7518,
+      "step": 49438
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.22097142857143e-07,
+      "loss": 1.0903,
+      "step": 49439
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.218285714285715e-07,
+      "loss": 1.1458,
+      "step": 49440
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.215600000000001e-07,
+      "loss": 1.1663,
+      "step": 49441
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.212914285714287e-07,
+      "loss": 1.1779,
+      "step": 49442
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.210228571428572e-07,
+      "loss": 0.8706,
+      "step": 49443
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.207542857142858e-07,
+      "loss": 1.5244,
+      "step": 49444
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.204857142857142e-07,
+      "loss": 0.7202,
+      "step": 49445
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.202171428571428e-07,
+      "loss": 1.1431,
+      "step": 49446
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.199485714285716e-07,
+      "loss": 1.4161,
+      "step": 49447
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.1968e-07,
+      "loss": 1.3683,
+      "step": 49448
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.194114285714286e-07,
+      "loss": 0.7485,
+      "step": 49449
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.191428571428571e-07,
+      "loss": 0.8641,
+      "step": 49450
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.188742857142857e-07,
+      "loss": 1.0722,
+      "step": 49451
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.186057142857143e-07,
+      "loss": 1.0879,
+      "step": 49452
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.183371428571428e-07,
+      "loss": 0.8315,
+      "step": 49453
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.180685714285714e-07,
+      "loss": 1.2735,
+      "step": 49454
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.178e-07,
+      "loss": 1.262,
+      "step": 49455
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.175314285714286e-07,
+      "loss": 0.7842,
+      "step": 49456
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.172628571428572e-07,
+      "loss": 0.965,
+      "step": 49457
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.169942857142857e-07,
+      "loss": 1.6126,
+      "step": 49458
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.167257142857143e-07,
+      "loss": 1.2748,
+      "step": 49459
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.164571428571428e-07,
+      "loss": 1.1457,
+      "step": 49460
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.161885714285714e-07,
+      "loss": 0.7513,
+      "step": 49461
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.1592e-07,
+      "loss": 1.1796,
+      "step": 49462
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.156514285714286e-07,
+      "loss": 1.0646,
+      "step": 49463
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.153828571428572e-07,
+      "loss": 0.7418,
+      "step": 49464
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.151142857142857e-07,
+      "loss": 1.2903,
+      "step": 49465
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.148457142857143e-07,
+      "loss": 1.2677,
+      "step": 49466
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.145771428571429e-07,
+      "loss": 1.168,
+      "step": 49467
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.143085714285714e-07,
+      "loss": 0.9726,
+      "step": 49468
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.1404e-07,
+      "loss": 1.1631,
+      "step": 49469
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.137714285714286e-07,
+      "loss": 1.2604,
+      "step": 49470
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.135028571428572e-07,
+      "loss": 1.6895,
+      "step": 49471
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.132342857142858e-07,
+      "loss": 0.7766,
+      "step": 49472
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.129657142857143e-07,
+      "loss": 1.1479,
+      "step": 49473
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.126971428571429e-07,
+      "loss": 1.0949,
+      "step": 49474
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.124285714285715e-07,
+      "loss": 1.0164,
+      "step": 49475
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.1216e-07,
+      "loss": 1.2258,
+      "step": 49476
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.118914285714286e-07,
+      "loss": 1.2836,
+      "step": 49477
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.116228571428571e-07,
+      "loss": 0.9737,
+      "step": 49478
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.113542857142858e-07,
+      "loss": 1.4099,
+      "step": 49479
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.110857142857144e-07,
+      "loss": 0.9968,
+      "step": 49480
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.108171428571429e-07,
+      "loss": 0.9203,
+      "step": 49481
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.105485714285715e-07,
+      "loss": 0.9788,
+      "step": 49482
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.1028e-07,
+      "loss": 1.0482,
+      "step": 49483
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.100114285714286e-07,
+      "loss": 0.788,
+      "step": 49484
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.097428571428572e-07,
+      "loss": 1.1069,
+      "step": 49485
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.094742857142857e-07,
+      "loss": 1.0827,
+      "step": 49486
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.092057142857144e-07,
+      "loss": 1.1881,
+      "step": 49487
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.089371428571429e-07,
+      "loss": 1.0491,
+      "step": 49488
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.086685714285715e-07,
+      "loss": 1.5239,
+      "step": 49489
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.084000000000001e-07,
+      "loss": 1.4349,
+      "step": 49490
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.081314285714286e-07,
+      "loss": 1.4899,
+      "step": 49491
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.078628571428572e-07,
+      "loss": 0.8607,
+      "step": 49492
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.075942857142858e-07,
+      "loss": 1.2025,
+      "step": 49493
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.073257142857143e-07,
+      "loss": 1.1238,
+      "step": 49494
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.07057142857143e-07,
+      "loss": 1.17,
+      "step": 49495
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.067885714285715e-07,
+      "loss": 0.8864,
+      "step": 49496
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.0652e-07,
+      "loss": 0.6935,
+      "step": 49497
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.062514285714287e-07,
+      "loss": 1.4804,
+      "step": 49498
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.059828571428571e-07,
+      "loss": 1.0487,
+      "step": 49499
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.057142857142857e-07,
+      "loss": 1.3326,
+      "step": 49500
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.054457142857142e-07,
+      "loss": 1.2286,
+      "step": 49501
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.051771428571429e-07,
+      "loss": 0.826,
+      "step": 49502
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.049085714285715e-07,
+      "loss": 1.2095,
+      "step": 49503
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.0464e-07,
+      "loss": 1.1438,
+      "step": 49504
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.043714285714286e-07,
+      "loss": 0.9932,
+      "step": 49505
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.041028571428571e-07,
+      "loss": 1.1675,
+      "step": 49506
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.038342857142857e-07,
+      "loss": 0.919,
+      "step": 49507
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.035657142857143e-07,
+      "loss": 1.0773,
+      "step": 49508
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.032971428571428e-07,
+      "loss": 1.207,
+      "step": 49509
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.030285714285715e-07,
+      "loss": 1.0842,
+      "step": 49510
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.0276e-07,
+      "loss": 1.3552,
+      "step": 49511
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.024914285714286e-07,
+      "loss": 1.1342,
+      "step": 49512
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.022228571428572e-07,
+      "loss": 1.1493,
+      "step": 49513
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.019542857142857e-07,
+      "loss": 1.2416,
+      "step": 49514
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.016857142857143e-07,
+      "loss": 0.9556,
+      "step": 49515
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.014171428571428e-07,
+      "loss": 1.3877,
+      "step": 49516
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.011485714285714e-07,
+      "loss": 1.0048,
+      "step": 49517
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.008800000000001e-07,
+      "loss": 0.964,
+      "step": 49518
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.006114285714286e-07,
+      "loss": 0.87,
+      "step": 49519
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.003428571428572e-07,
+      "loss": 1.1738,
+      "step": 49520
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.000742857142857e-07,
+      "loss": 1.0998,
+      "step": 49521
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.998057142857143e-07,
+      "loss": 1.0974,
+      "step": 49522
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.995371428571429e-07,
+      "loss": 0.9208,
+      "step": 49523
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.992685714285714e-07,
+      "loss": 1.4129,
+      "step": 49524
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.99e-07,
+      "loss": 0.9595,
+      "step": 49525
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.987314285714286e-07,
+      "loss": 1.1737,
+      "step": 49526
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.984628571428572e-07,
+      "loss": 1.2427,
+      "step": 49527
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.981942857142858e-07,
+      "loss": 0.8289,
+      "step": 49528
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.979257142857143e-07,
+      "loss": 0.9396,
+      "step": 49529
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.976571428571429e-07,
+      "loss": 1.0025,
+      "step": 49530
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.973885714285715e-07,
+      "loss": 1.3063,
+      "step": 49531
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.9712e-07,
+      "loss": 0.8509,
+      "step": 49532
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.968514285714286e-07,
+      "loss": 1.0942,
+      "step": 49533
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.965828571428572e-07,
+      "loss": 1.2109,
+      "step": 49534
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.963142857142858e-07,
+      "loss": 0.8905,
+      "step": 49535
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.960457142857144e-07,
+      "loss": 0.8516,
+      "step": 49536
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.957771428571429e-07,
+      "loss": 0.8602,
+      "step": 49537
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.955085714285715e-07,
+      "loss": 1.0397,
+      "step": 49538
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.9524e-07,
+      "loss": 1.002,
+      "step": 49539
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.949714285714286e-07,
+      "loss": 1.134,
+      "step": 49540
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.947028571428572e-07,
+      "loss": 0.8954,
+      "step": 49541
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.944342857142858e-07,
+      "loss": 0.6416,
+      "step": 49542
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.941657142857144e-07,
+      "loss": 0.7632,
+      "step": 49543
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.938971428571429e-07,
+      "loss": 0.8316,
+      "step": 49544
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.936285714285715e-07,
+      "loss": 1.296,
+      "step": 49545
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.933600000000001e-07,
+      "loss": 1.2132,
+      "step": 49546
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.930914285714286e-07,
+      "loss": 1.0225,
+      "step": 49547
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.928228571428572e-07,
+      "loss": 1.4242,
+      "step": 49548
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.925542857142858e-07,
+      "loss": 1.1465,
+      "step": 49549
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.922857142857144e-07,
+      "loss": 1.4392,
+      "step": 49550
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.92017142857143e-07,
+      "loss": 1.5071,
+      "step": 49551
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.917485714285714e-07,
+      "loss": 1.101,
+      "step": 49552
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.9148e-07,
+      "loss": 1.1114,
+      "step": 49553
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.912114285714286e-07,
+      "loss": 0.9115,
+      "step": 49554
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.909428571428571e-07,
+      "loss": 0.993,
+      "step": 49555
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.906742857142857e-07,
+      "loss": 1.1793,
+      "step": 49556
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.904057142857142e-07,
+      "loss": 1.4556,
+      "step": 49557
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.901371428571429e-07,
+      "loss": 1.1216,
+      "step": 49558
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.898685714285715e-07,
+      "loss": 1.2595,
+      "step": 49559
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.896e-07,
+      "loss": 1.1792,
+      "step": 49560
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.893314285714286e-07,
+      "loss": 1.2053,
+      "step": 49561
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.890628571428571e-07,
+      "loss": 0.8329,
+      "step": 49562
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.887942857142857e-07,
+      "loss": 1.1161,
+      "step": 49563
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.885257142857143e-07,
+      "loss": 0.8532,
+      "step": 49564
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.882571428571428e-07,
+      "loss": 1.3079,
+      "step": 49565
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.879885714285715e-07,
+      "loss": 0.754,
+      "step": 49566
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.8772e-07,
+      "loss": 1.0505,
+      "step": 49567
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.874514285714286e-07,
+      "loss": 1.0149,
+      "step": 49568
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.871828571428572e-07,
+      "loss": 1.119,
+      "step": 49569
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.869142857142857e-07,
+      "loss": 0.8586,
+      "step": 49570
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.866457142857143e-07,
+      "loss": 0.8164,
+      "step": 49571
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.863771428571428e-07,
+      "loss": 0.7771,
+      "step": 49572
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.861085714285714e-07,
+      "loss": 1.0674,
+      "step": 49573
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.858400000000001e-07,
+      "loss": 1.011,
+      "step": 49574
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.855714285714286e-07,
+      "loss": 1.317,
+      "step": 49575
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.853028571428572e-07,
+      "loss": 1.0502,
+      "step": 49576
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.850342857142857e-07,
+      "loss": 1.6292,
+      "step": 49577
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.847657142857143e-07,
+      "loss": 1.2167,
+      "step": 49578
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.844971428571429e-07,
+      "loss": 1.3871,
+      "step": 49579
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.842285714285714e-07,
+      "loss": 0.773,
+      "step": 49580
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.8396e-07,
+      "loss": 1.493,
+      "step": 49581
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.836914285714286e-07,
+      "loss": 1.1404,
+      "step": 49582
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.834228571428572e-07,
+      "loss": 0.9712,
+      "step": 49583
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.831542857142858e-07,
+      "loss": 0.8722,
+      "step": 49584
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.828857142857143e-07,
+      "loss": 1.1991,
+      "step": 49585
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.826171428571429e-07,
+      "loss": 1.051,
+      "step": 49586
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.823485714285715e-07,
+      "loss": 1.1668,
+      "step": 49587
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.8208e-07,
+      "loss": 0.8116,
+      "step": 49588
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.818114285714286e-07,
+      "loss": 1.1256,
+      "step": 49589
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.815428571428572e-07,
+      "loss": 0.887,
+      "step": 49590
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.812742857142858e-07,
+      "loss": 1.0243,
+      "step": 49591
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.810057142857144e-07,
+      "loss": 1.1285,
+      "step": 49592
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.807371428571429e-07,
+      "loss": 1.2858,
+      "step": 49593
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.804685714285715e-07,
+      "loss": 1.626,
+      "step": 49594
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.802e-07,
+      "loss": 1.288,
+      "step": 49595
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.799314285714286e-07,
+      "loss": 1.4806,
+      "step": 49596
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.796628571428573e-07,
+      "loss": 0.9766,
+      "step": 49597
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.793942857142858e-07,
+      "loss": 0.791,
+      "step": 49598
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.791257142857144e-07,
+      "loss": 1.3454,
+      "step": 49599
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.788571428571429e-07,
+      "loss": 1.2473,
+      "step": 49600
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.785885714285715e-07,
+      "loss": 1.3733,
+      "step": 49601
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.7832e-07,
+      "loss": 0.7257,
+      "step": 49602
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.780514285714285e-07,
+      "loss": 1.1898,
+      "step": 49603
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.777828571428571e-07,
+      "loss": 1.2181,
+      "step": 49604
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.775142857142857e-07,
+      "loss": 0.9844,
+      "step": 49605
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.772457142857143e-07,
+      "loss": 1.2412,
+      "step": 49606
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.769771428571429e-07,
+      "loss": 1.3237,
+      "step": 49607
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.767085714285714e-07,
+      "loss": 0.7715,
+      "step": 49608
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.7644e-07,
+      "loss": 1.282,
+      "step": 49609
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.761714285714286e-07,
+      "loss": 1.3065,
+      "step": 49610
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.759028571428571e-07,
+      "loss": 1.1302,
+      "step": 49611
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.756342857142857e-07,
+      "loss": 1.1177,
+      "step": 49612
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.753657142857143e-07,
+      "loss": 0.782,
+      "step": 49613
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.750971428571429e-07,
+      "loss": 1.0669,
+      "step": 49614
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.748285714285715e-07,
+      "loss": 1.3334,
+      "step": 49615
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.7456e-07,
+      "loss": 1.4216,
+      "step": 49616
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.742914285714286e-07,
+      "loss": 0.7216,
+      "step": 49617
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.740228571428571e-07,
+      "loss": 0.9377,
+      "step": 49618
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.737542857142857e-07,
+      "loss": 0.9065,
+      "step": 49619
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.734857142857143e-07,
+      "loss": 0.9385,
+      "step": 49620
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.732171428571429e-07,
+      "loss": 0.7765,
+      "step": 49621
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.729485714285715e-07,
+      "loss": 1.1698,
+      "step": 49622
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.7268e-07,
+      "loss": 0.9035,
+      "step": 49623
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.724114285714286e-07,
+      "loss": 1.2803,
+      "step": 49624
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.721428571428572e-07,
+      "loss": 0.6313,
+      "step": 49625
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.718742857142857e-07,
+      "loss": 0.7836,
+      "step": 49626
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.716057142857143e-07,
+      "loss": 1.092,
+      "step": 49627
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.713371428571428e-07,
+      "loss": 1.1753,
+      "step": 49628
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.710685714285715e-07,
+      "loss": 0.9073,
+      "step": 49629
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.708000000000001e-07,
+      "loss": 1.0283,
+      "step": 49630
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.705314285714286e-07,
+      "loss": 0.9114,
+      "step": 49631
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.702628571428572e-07,
+      "loss": 0.6411,
+      "step": 49632
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.699942857142857e-07,
+      "loss": 1.166,
+      "step": 49633
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.697257142857143e-07,
+      "loss": 1.1507,
+      "step": 49634
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.694571428571429e-07,
+      "loss": 1.0862,
+      "step": 49635
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.691885714285714e-07,
+      "loss": 1.0749,
+      "step": 49636
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.689200000000001e-07,
+      "loss": 1.0929,
+      "step": 49637
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.686514285714286e-07,
+      "loss": 1.216,
+      "step": 49638
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.683828571428572e-07,
+      "loss": 1.4284,
+      "step": 49639
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.681142857142858e-07,
+      "loss": 1.0275,
+      "step": 49640
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.678457142857143e-07,
+      "loss": 1.5044,
+      "step": 49641
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.675771428571429e-07,
+      "loss": 0.8611,
+      "step": 49642
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.673085714285715e-07,
+      "loss": 1.2774,
+      "step": 49643
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.6704e-07,
+      "loss": 0.9489,
+      "step": 49644
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.667714285714287e-07,
+      "loss": 1.1328,
+      "step": 49645
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.665028571428572e-07,
+      "loss": 1.2006,
+      "step": 49646
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.662342857142858e-07,
+      "loss": 1.341,
+      "step": 49647
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.659657142857144e-07,
+      "loss": 0.9458,
+      "step": 49648
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.656971428571429e-07,
+      "loss": 1.0473,
+      "step": 49649
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.654285714285715e-07,
+      "loss": 1.0325,
+      "step": 49650
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.6516e-07,
+      "loss": 1.0244,
+      "step": 49651
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.648914285714286e-07,
+      "loss": 1.1233,
+      "step": 49652
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.646228571428573e-07,
+      "loss": 1.4154,
+      "step": 49653
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.643542857142857e-07,
+      "loss": 1.2731,
+      "step": 49654
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.640857142857143e-07,
+      "loss": 0.8816,
+      "step": 49655
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.638171428571428e-07,
+      "loss": 0.6447,
+      "step": 49656
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.635485714285714e-07,
+      "loss": 1.131,
+      "step": 49657
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.6328e-07,
+      "loss": 1.0573,
+      "step": 49658
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.630114285714285e-07,
+      "loss": 1.4388,
+      "step": 49659
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.627428571428571e-07,
+      "loss": 0.8779,
+      "step": 49660
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.624742857142857e-07,
+      "loss": 1.0951,
+      "step": 49661
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.622057142857143e-07,
+      "loss": 0.9026,
+      "step": 49662
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.619371428571429e-07,
+      "loss": 1.3078,
+      "step": 49663
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.616685714285714e-07,
+      "loss": 0.8512,
+      "step": 49664
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.614e-07,
+      "loss": 1.1944,
+      "step": 49665
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.611314285714286e-07,
+      "loss": 1.426,
+      "step": 49666
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.608628571428571e-07,
+      "loss": 1.1914,
+      "step": 49667
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.605942857142857e-07,
+      "loss": 1.1816,
+      "step": 49668
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.603257142857143e-07,
+      "loss": 0.9775,
+      "step": 49669
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.600571428571429e-07,
+      "loss": 1.3467,
+      "step": 49670
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.597885714285715e-07,
+      "loss": 0.8587,
+      "step": 49671
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.5952e-07,
+      "loss": 0.8946,
+      "step": 49672
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.592514285714286e-07,
+      "loss": 1.3947,
+      "step": 49673
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.589828571428571e-07,
+      "loss": 1.3768,
+      "step": 49674
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.587142857142857e-07,
+      "loss": 1.24,
+      "step": 49675
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.584457142857143e-07,
+      "loss": 0.682,
+      "step": 49676
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.581771428571429e-07,
+      "loss": 0.7937,
+      "step": 49677
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.579085714285715e-07,
+      "loss": 1.0252,
+      "step": 49678
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.5764e-07,
+      "loss": 1.4391,
+      "step": 49679
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.573714285714286e-07,
+      "loss": 0.9085,
+      "step": 49680
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.571028571428572e-07,
+      "loss": 0.9682,
+      "step": 49681
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.568342857142857e-07,
+      "loss": 1.6015,
+      "step": 49682
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.565657142857143e-07,
+      "loss": 1.1997,
+      "step": 49683
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.562971428571428e-07,
+      "loss": 1.3178,
+      "step": 49684
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.560285714285715e-07,
+      "loss": 1.2919,
+      "step": 49685
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.557600000000001e-07,
+      "loss": 0.4806,
+      "step": 49686
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.554914285714286e-07,
+      "loss": 0.8733,
+      "step": 49687
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.552228571428572e-07,
+      "loss": 1.3122,
+      "step": 49688
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.549542857142857e-07,
+      "loss": 0.8428,
+      "step": 49689
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.546857142857143e-07,
+      "loss": 0.9487,
+      "step": 49690
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.544171428571429e-07,
+      "loss": 0.9877,
+      "step": 49691
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.541485714285714e-07,
+      "loss": 1.1678,
+      "step": 49692
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.538800000000001e-07,
+      "loss": 0.9839,
+      "step": 49693
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.536114285714286e-07,
+      "loss": 1.2095,
+      "step": 49694
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.533428571428572e-07,
+      "loss": 1.1923,
+      "step": 49695
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.530742857142858e-07,
+      "loss": 1.1126,
+      "step": 49696
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.528057142857143e-07,
+      "loss": 1.1369,
+      "step": 49697
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.525371428571429e-07,
+      "loss": 1.1215,
+      "step": 49698
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.522685714285715e-07,
+      "loss": 1.0321,
+      "step": 49699
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.520000000000001e-07,
+      "loss": 1.21,
+      "step": 49700
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.517314285714287e-07,
+      "loss": 1.0447,
+      "step": 49701
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.514628571428572e-07,
+      "loss": 0.8009,
+      "step": 49702
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.511942857142858e-07,
+      "loss": 1.2199,
+      "step": 49703
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.509257142857144e-07,
+      "loss": 1.1076,
+      "step": 49704
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.506571428571428e-07,
+      "loss": 1.3146,
+      "step": 49705
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.503885714285714e-07,
+      "loss": 1.1749,
+      "step": 49706
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.501199999999999e-07,
+      "loss": 1.2834,
+      "step": 49707
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.498514285714286e-07,
+      "loss": 1.4103,
+      "step": 49708
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.495828571428572e-07,
+      "loss": 0.8043,
+      "step": 49709
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.493142857142857e-07,
+      "loss": 0.7908,
+      "step": 49710
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.490457142857143e-07,
+      "loss": 1.1421,
+      "step": 49711
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.487771428571428e-07,
+      "loss": 1.5318,
+      "step": 49712
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.485085714285714e-07,
+      "loss": 1.0255,
+      "step": 49713
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.4824e-07,
+      "loss": 0.7789,
+      "step": 49714
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.479714285714285e-07,
+      "loss": 1.5533,
+      "step": 49715
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.477028571428572e-07,
+      "loss": 1.6169,
+      "step": 49716
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.474342857142857e-07,
+      "loss": 1.3245,
+      "step": 49717
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.471657142857143e-07,
+      "loss": 1.1234,
+      "step": 49718
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.468971428571429e-07,
+      "loss": 1.2272,
+      "step": 49719
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.466285714285714e-07,
+      "loss": 0.7728,
+      "step": 49720
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.4636e-07,
+      "loss": 1.025,
+      "step": 49721
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.460914285714286e-07,
+      "loss": 1.4505,
+      "step": 49722
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.458228571428571e-07,
+      "loss": 1.0152,
+      "step": 49723
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.455542857142858e-07,
+      "loss": 1.2302,
+      "step": 49724
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.452857142857143e-07,
+      "loss": 0.7582,
+      "step": 49725
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.450171428571429e-07,
+      "loss": 1.2081,
+      "step": 49726
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.447485714285715e-07,
+      "loss": 1.1353,
+      "step": 49727
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.4448e-07,
+      "loss": 0.9309,
+      "step": 49728
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.442114285714286e-07,
+      "loss": 1.076,
+      "step": 49729
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.439428571428571e-07,
+      "loss": 1.7061,
+      "step": 49730
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.436742857142857e-07,
+      "loss": 0.948,
+      "step": 49731
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.434057142857144e-07,
+      "loss": 1.0022,
+      "step": 49732
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.431371428571429e-07,
+      "loss": 1.2038,
+      "step": 49733
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.428685714285715e-07,
+      "loss": 0.9628,
+      "step": 49734
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.426e-07,
+      "loss": 1.535,
+      "step": 49735
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.423314285714286e-07,
+      "loss": 1.1404,
+      "step": 49736
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.420628571428572e-07,
+      "loss": 0.8936,
+      "step": 49737
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.417942857142857e-07,
+      "loss": 1.0817,
+      "step": 49738
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.415257142857143e-07,
+      "loss": 0.864,
+      "step": 49739
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.412571428571429e-07,
+      "loss": 0.9903,
+      "step": 49740
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.409885714285715e-07,
+      "loss": 0.9451,
+      "step": 49741
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.407200000000001e-07,
+      "loss": 1.4699,
+      "step": 49742
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.404514285714286e-07,
+      "loss": 1.4324,
+      "step": 49743
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.401828571428572e-07,
+      "loss": 1.2855,
+      "step": 49744
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.399142857142857e-07,
+      "loss": 1.0017,
+      "step": 49745
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.396457142857143e-07,
+      "loss": 1.0307,
+      "step": 49746
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.393771428571429e-07,
+      "loss": 0.7583,
+      "step": 49747
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.391085714285715e-07,
+      "loss": 0.9281,
+      "step": 49748
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.388400000000001e-07,
+      "loss": 1.3191,
+      "step": 49749
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.385714285714287e-07,
+      "loss": 1.4876,
+      "step": 49750
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.383028571428572e-07,
+      "loss": 1.3395,
+      "step": 49751
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.380342857142858e-07,
+      "loss": 0.6444,
+      "step": 49752
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.377657142857143e-07,
+      "loss": 0.6819,
+      "step": 49753
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.374971428571429e-07,
+      "loss": 0.8179,
+      "step": 49754
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.372285714285715e-07,
+      "loss": 1.0956,
+      "step": 49755
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.369600000000001e-07,
+      "loss": 1.2978,
+      "step": 49756
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.366914285714287e-07,
+      "loss": 1.1726,
+      "step": 49757
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.364228571428571e-07,
+      "loss": 1.2184,
+      "step": 49758
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.361542857142857e-07,
+      "loss": 0.7556,
+      "step": 49759
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.358857142857143e-07,
+      "loss": 1.0935,
+      "step": 49760
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.356171428571428e-07,
+      "loss": 1.2786,
+      "step": 49761
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.353485714285714e-07,
+      "loss": 1.7653,
+      "step": 49762
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.350799999999999e-07,
+      "loss": 1.6498,
+      "step": 49763
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.348114285714286e-07,
+      "loss": 1.0867,
+      "step": 49764
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.345428571428572e-07,
+      "loss": 1.2215,
+      "step": 49765
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.342742857142857e-07,
+      "loss": 0.9374,
+      "step": 49766
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.340057142857143e-07,
+      "loss": 1.0061,
+      "step": 49767
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.337371428571428e-07,
+      "loss": 1.179,
+      "step": 49768
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.334685714285714e-07,
+      "loss": 1.2968,
+      "step": 49769
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.332e-07,
+      "loss": 1.3443,
+      "step": 49770
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.329314285714285e-07,
+      "loss": 1.4203,
+      "step": 49771
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.326628571428572e-07,
+      "loss": 1.3891,
+      "step": 49772
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.323942857142857e-07,
+      "loss": 1.2157,
+      "step": 49773
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.321257142857143e-07,
+      "loss": 0.9131,
+      "step": 49774
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.318571428571429e-07,
+      "loss": 1.159,
+      "step": 49775
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.315885714285714e-07,
+      "loss": 1.1083,
+      "step": 49776
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.3132e-07,
+      "loss": 1.5753,
+      "step": 49777
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.310514285714286e-07,
+      "loss": 1.1371,
+      "step": 49778
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.307828571428571e-07,
+      "loss": 0.7419,
+      "step": 49779
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.305142857142858e-07,
+      "loss": 1.375,
+      "step": 49780
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.302457142857143e-07,
+      "loss": 1.0828,
+      "step": 49781
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.299771428571429e-07,
+      "loss": 1.3379,
+      "step": 49782
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.297085714285715e-07,
+      "loss": 1.4686,
+      "step": 49783
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.2944e-07,
+      "loss": 1.4944,
+      "step": 49784
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.291714285714286e-07,
+      "loss": 1.1769,
+      "step": 49785
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.289028571428571e-07,
+      "loss": 1.1152,
+      "step": 49786
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.286342857142857e-07,
+      "loss": 1.438,
+      "step": 49787
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.283657142857144e-07,
+      "loss": 1.277,
+      "step": 49788
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.280971428571429e-07,
+      "loss": 1.1574,
+      "step": 49789
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.278285714285715e-07,
+      "loss": 1.1031,
+      "step": 49790
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.2756e-07,
+      "loss": 0.8637,
+      "step": 49791
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.272914285714286e-07,
+      "loss": 1.314,
+      "step": 49792
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.270228571428572e-07,
+      "loss": 1.3222,
+      "step": 49793
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.267542857142857e-07,
+      "loss": 0.784,
+      "step": 49794
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.264857142857144e-07,
+      "loss": 1.108,
+      "step": 49795
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.262171428571429e-07,
+      "loss": 0.8516,
+      "step": 49796
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.259485714285715e-07,
+      "loss": 1.6386,
+      "step": 49797
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.256800000000001e-07,
+      "loss": 1.4224,
+      "step": 49798
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.254114285714286e-07,
+      "loss": 1.0861,
+      "step": 49799
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.251428571428572e-07,
+      "loss": 0.9603,
+      "step": 49800
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.248742857142857e-07,
+      "loss": 1.039,
+      "step": 49801
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.246057142857143e-07,
+      "loss": 0.9759,
+      "step": 49802
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.24337142857143e-07,
+      "loss": 1.1966,
+      "step": 49803
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.240685714285715e-07,
+      "loss": 1.1726,
+      "step": 49804
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.238000000000001e-07,
+      "loss": 1.5737,
+      "step": 49805
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.235314285714287e-07,
+      "loss": 0.6762,
+      "step": 49806
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.232628571428572e-07,
+      "loss": 1.0263,
+      "step": 49807
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.229942857142858e-07,
+      "loss": 0.9969,
+      "step": 49808
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.227257142857142e-07,
+      "loss": 1.5659,
+      "step": 49809
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.224571428571428e-07,
+      "loss": 0.7129,
+      "step": 49810
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.221885714285716e-07,
+      "loss": 1.2157,
+      "step": 49811
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.2192e-07,
+      "loss": 0.8579,
+      "step": 49812
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.216514285714286e-07,
+      "loss": 0.9296,
+      "step": 49813
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.213828571428571e-07,
+      "loss": 1.0028,
+      "step": 49814
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.211142857142857e-07,
+      "loss": 1.1397,
+      "step": 49815
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.208457142857143e-07,
+      "loss": 1.1689,
+      "step": 49816
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.205771428571428e-07,
+      "loss": 1.0289,
+      "step": 49817
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.203085714285714e-07,
+      "loss": 1.0487,
+      "step": 49818
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.2004e-07,
+      "loss": 1.0113,
+      "step": 49819
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.197714285714286e-07,
+      "loss": 1.3019,
+      "step": 49820
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.195028571428572e-07,
+      "loss": 1.1854,
+      "step": 49821
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.192342857142857e-07,
+      "loss": 1.2982,
+      "step": 49822
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.189657142857143e-07,
+      "loss": 0.7764,
+      "step": 49823
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.186971428571428e-07,
+      "loss": 1.355,
+      "step": 49824
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.184285714285714e-07,
+      "loss": 1.323,
+      "step": 49825
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.1816e-07,
+      "loss": 1.032,
+      "step": 49826
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.178914285714286e-07,
+      "loss": 1.2873,
+      "step": 49827
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.176228571428572e-07,
+      "loss": 1.3161,
+      "step": 49828
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.173542857142857e-07,
+      "loss": 1.0372,
+      "step": 49829
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.170857142857143e-07,
+      "loss": 0.9694,
+      "step": 49830
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.168171428571429e-07,
+      "loss": 1.0886,
+      "step": 49831
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.165485714285714e-07,
+      "loss": 0.9531,
+      "step": 49832
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.1628e-07,
+      "loss": 0.9594,
+      "step": 49833
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.160114285714286e-07,
+      "loss": 0.7105,
+      "step": 49834
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.157428571428572e-07,
+      "loss": 1.7142,
+      "step": 49835
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.154742857142858e-07,
+      "loss": 1.4614,
+      "step": 49836
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.152057142857143e-07,
+      "loss": 1.2272,
+      "step": 49837
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.149371428571429e-07,
+      "loss": 1.5382,
+      "step": 49838
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.146685714285715e-07,
+      "loss": 1.115,
+      "step": 49839
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.144e-07,
+      "loss": 0.9087,
+      "step": 49840
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.141314285714286e-07,
+      "loss": 1.0139,
+      "step": 49841
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.138628571428571e-07,
+      "loss": 1.1905,
+      "step": 49842
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.135942857142858e-07,
+      "loss": 0.7517,
+      "step": 49843
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.133257142857144e-07,
+      "loss": 0.8516,
+      "step": 49844
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.130571428571429e-07,
+      "loss": 1.0313,
+      "step": 49845
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.127885714285715e-07,
+      "loss": 1.0423,
+      "step": 49846
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.1252e-07,
+      "loss": 1.3132,
+      "step": 49847
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.122514285714286e-07,
+      "loss": 0.8088,
+      "step": 49848
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.119828571428572e-07,
+      "loss": 1.489,
+      "step": 49849
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.117142857142857e-07,
+      "loss": 1.0856,
+      "step": 49850
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.114457142857144e-07,
+      "loss": 1.157,
+      "step": 49851
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.111771428571429e-07,
+      "loss": 1.1317,
+      "step": 49852
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.109085714285715e-07,
+      "loss": 1.2124,
+      "step": 49853
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.106400000000001e-07,
+      "loss": 1.4309,
+      "step": 49854
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.103714285714286e-07,
+      "loss": 1.0924,
+      "step": 49855
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.101028571428572e-07,
+      "loss": 1.2319,
+      "step": 49856
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.098342857142857e-07,
+      "loss": 1.0481,
+      "step": 49857
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.095657142857143e-07,
+      "loss": 0.6872,
+      "step": 49858
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.09297142857143e-07,
+      "loss": 1.3391,
+      "step": 49859
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.090285714285715e-07,
+      "loss": 1.1481,
+      "step": 49860
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.087600000000001e-07,
+      "loss": 1.0848,
+      "step": 49861
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.084914285714287e-07,
+      "loss": 0.9834,
+      "step": 49862
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.082228571428571e-07,
+      "loss": 1.2002,
+      "step": 49863
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.079542857142857e-07,
+      "loss": 0.8602,
+      "step": 49864
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.076857142857142e-07,
+      "loss": 1.2814,
+      "step": 49865
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.074171428571428e-07,
+      "loss": 1.0036,
+      "step": 49866
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.071485714285715e-07,
+      "loss": 0.7733,
+      "step": 49867
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.0688e-07,
+      "loss": 0.9541,
+      "step": 49868
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.066114285714286e-07,
+      "loss": 1.1411,
+      "step": 49869
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.063428571428571e-07,
+      "loss": 0.8281,
+      "step": 49870
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.060742857142857e-07,
+      "loss": 0.9133,
+      "step": 49871
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.058057142857143e-07,
+      "loss": 1.0267,
+      "step": 49872
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.055371428571428e-07,
+      "loss": 1.2689,
+      "step": 49873
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.052685714285714e-07,
+      "loss": 0.7738,
+      "step": 49874
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.05e-07,
+      "loss": 1.4003,
+      "step": 49875
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.047314285714286e-07,
+      "loss": 0.8866,
+      "step": 49876
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.044628571428572e-07,
+      "loss": 0.7762,
+      "step": 49877
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.041942857142857e-07,
+      "loss": 1.2522,
+      "step": 49878
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.039257142857143e-07,
+      "loss": 1.0296,
+      "step": 49879
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.036571428571428e-07,
+      "loss": 1.4179,
+      "step": 49880
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.033885714285714e-07,
+      "loss": 1.1566,
+      "step": 49881
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.0312e-07,
+      "loss": 0.9519,
+      "step": 49882
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.028514285714286e-07,
+      "loss": 1.4362,
+      "step": 49883
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.025828571428572e-07,
+      "loss": 1.2289,
+      "step": 49884
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.023142857142857e-07,
+      "loss": 0.6965,
+      "step": 49885
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.020457142857143e-07,
+      "loss": 1.1948,
+      "step": 49886
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.017771428571429e-07,
+      "loss": 1.1839,
+      "step": 49887
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.015085714285714e-07,
+      "loss": 1.2066,
+      "step": 49888
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.0124e-07,
+      "loss": 1.5485,
+      "step": 49889
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.009714285714286e-07,
+      "loss": 1.0454,
+      "step": 49890
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.007028571428572e-07,
+      "loss": 0.937,
+      "step": 49891
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.004342857142858e-07,
+      "loss": 1.1372,
+      "step": 49892
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.001657142857143e-07,
+      "loss": 0.939,
+      "step": 49893
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.998971428571429e-07,
+      "loss": 0.9568,
+      "step": 49894
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.996285714285715e-07,
+      "loss": 1.2026,
+      "step": 49895
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.9936e-07,
+      "loss": 0.9838,
+      "step": 49896
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.990914285714286e-07,
+      "loss": 0.8787,
+      "step": 49897
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.988228571428572e-07,
+      "loss": 1.2529,
+      "step": 49898
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.985542857142858e-07,
+      "loss": 1.263,
+      "step": 49899
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.982857142857144e-07,
+      "loss": 0.9827,
+      "step": 49900
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.980171428571429e-07,
+      "loss": 0.8149,
+      "step": 49901
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.977485714285715e-07,
+      "loss": 1.2074,
+      "step": 49902
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.9748e-07,
+      "loss": 1.1026,
+      "step": 49903
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.972114285714286e-07,
+      "loss": 1.4352,
+      "step": 49904
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.969428571428572e-07,
+      "loss": 1.2828,
+      "step": 49905
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.966742857142858e-07,
+      "loss": 1.1772,
+      "step": 49906
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.964057142857144e-07,
+      "loss": 1.262,
+      "step": 49907
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.961371428571429e-07,
+      "loss": 0.7208,
+      "step": 49908
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.958685714285715e-07,
+      "loss": 1.235,
+      "step": 49909
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.956000000000001e-07,
+      "loss": 0.8553,
+      "step": 49910
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.953314285714286e-07,
+      "loss": 1.2632,
+      "step": 49911
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.950628571428572e-07,
+      "loss": 1.7688,
+      "step": 49912
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.947942857142856e-07,
+      "loss": 0.7168,
+      "step": 49913
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.945257142857144e-07,
+      "loss": 0.93,
+      "step": 49914
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.94257142857143e-07,
+      "loss": 0.6159,
+      "step": 49915
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.939885714285714e-07,
+      "loss": 0.88,
+      "step": 49916
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.9372e-07,
+      "loss": 1.3797,
+      "step": 49917
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.934514285714286e-07,
+      "loss": 1.0443,
+      "step": 49918
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.931828571428571e-07,
+      "loss": 0.712,
+      "step": 49919
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.929142857142857e-07,
+      "loss": 1.039,
+      "step": 49920
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.926457142857142e-07,
+      "loss": 1.2698,
+      "step": 49921
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.923771428571429e-07,
+      "loss": 0.6874,
+      "step": 49922
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.921085714285715e-07,
+      "loss": 1.2373,
+      "step": 49923
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.9184e-07,
+      "loss": 1.1733,
+      "step": 49924
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.915714285714286e-07,
+      "loss": 1.089,
+      "step": 49925
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.913028571428571e-07,
+      "loss": 0.9067,
+      "step": 49926
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.910342857142857e-07,
+      "loss": 0.7171,
+      "step": 49927
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.907657142857143e-07,
+      "loss": 1.0966,
+      "step": 49928
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.904971428571428e-07,
+      "loss": 1.1174,
+      "step": 49929
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.902285714285715e-07,
+      "loss": 0.9174,
+      "step": 49930
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.8996e-07,
+      "loss": 0.9728,
+      "step": 49931
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.896914285714286e-07,
+      "loss": 1.0654,
+      "step": 49932
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.894228571428572e-07,
+      "loss": 1.3952,
+      "step": 49933
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.891542857142857e-07,
+      "loss": 1.1383,
+      "step": 49934
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.888857142857143e-07,
+      "loss": 1.145,
+      "step": 49935
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.886171428571428e-07,
+      "loss": 1.2334,
+      "step": 49936
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.883485714285714e-07,
+      "loss": 1.5021,
+      "step": 49937
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.880800000000001e-07,
+      "loss": 0.9527,
+      "step": 49938
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.878114285714286e-07,
+      "loss": 1.3546,
+      "step": 49939
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.875428571428572e-07,
+      "loss": 1.1792,
+      "step": 49940
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.872742857142857e-07,
+      "loss": 0.9476,
+      "step": 49941
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.870057142857143e-07,
+      "loss": 1.1213,
+      "step": 49942
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.867371428571429e-07,
+      "loss": 1.1519,
+      "step": 49943
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.864685714285714e-07,
+      "loss": 1.4285,
+      "step": 49944
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.862e-07,
+      "loss": 1.6094,
+      "step": 49945
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.859314285714286e-07,
+      "loss": 1.1531,
+      "step": 49946
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.856628571428572e-07,
+      "loss": 0.9469,
+      "step": 49947
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.853942857142858e-07,
+      "loss": 0.99,
+      "step": 49948
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.851257142857143e-07,
+      "loss": 1.5323,
+      "step": 49949
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.848571428571429e-07,
+      "loss": 0.8393,
+      "step": 49950
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.845885714285715e-07,
+      "loss": 1.3657,
+      "step": 49951
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.8432e-07,
+      "loss": 1.295,
+      "step": 49952
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.840514285714286e-07,
+      "loss": 1.1026,
+      "step": 49953
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.837828571428572e-07,
+      "loss": 1.1889,
+      "step": 49954
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.835142857142858e-07,
+      "loss": 0.9621,
+      "step": 49955
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.832457142857144e-07,
+      "loss": 0.9279,
+      "step": 49956
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.829771428571429e-07,
+      "loss": 0.9493,
+      "step": 49957
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.827085714285715e-07,
+      "loss": 1.2511,
+      "step": 49958
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.8244e-07,
+      "loss": 1.2531,
+      "step": 49959
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.821714285714286e-07,
+      "loss": 1.0288,
+      "step": 49960
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.819028571428572e-07,
+      "loss": 0.978,
+      "step": 49961
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.816342857142858e-07,
+      "loss": 1.0141,
+      "step": 49962
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.813657142857144e-07,
+      "loss": 0.7929,
+      "step": 49963
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.810971428571429e-07,
+      "loss": 0.9502,
+      "step": 49964
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.808285714285715e-07,
+      "loss": 1.3431,
+      "step": 49965
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.805600000000001e-07,
+      "loss": 0.8455,
+      "step": 49966
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.802914285714285e-07,
+      "loss": 0.9061,
+      "step": 49967
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.800228571428571e-07,
+      "loss": 0.8653,
+      "step": 49968
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.797542857142857e-07,
+      "loss": 1.0218,
+      "step": 49969
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.794857142857143e-07,
+      "loss": 1.1252,
+      "step": 49970
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.792171428571429e-07,
+      "loss": 1.4323,
+      "step": 49971
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.789485714285714e-07,
+      "loss": 1.1417,
+      "step": 49972
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.7868e-07,
+      "loss": 0.84,
+      "step": 49973
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.784114285714286e-07,
+      "loss": 1.0871,
+      "step": 49974
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.781428571428571e-07,
+      "loss": 0.846,
+      "step": 49975
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.778742857142857e-07,
+      "loss": 1.4093,
+      "step": 49976
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.776057142857142e-07,
+      "loss": 1.157,
+      "step": 49977
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.773371428571429e-07,
+      "loss": 1.0693,
+      "step": 49978
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.770685714285715e-07,
+      "loss": 1.2236,
+      "step": 49979
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.768e-07,
+      "loss": 1.4455,
+      "step": 49980
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.765314285714286e-07,
+      "loss": 0.9526,
+      "step": 49981
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.762628571428571e-07,
+      "loss": 0.879,
+      "step": 49982
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.759942857142857e-07,
+      "loss": 1.1185,
+      "step": 49983
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.757257142857143e-07,
+      "loss": 1.4709,
+      "step": 49984
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.754571428571428e-07,
+      "loss": 0.6725,
+      "step": 49985
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.751885714285715e-07,
+      "loss": 1.0399,
+      "step": 49986
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.7492e-07,
+      "loss": 1.12,
+      "step": 49987
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.746514285714286e-07,
+      "loss": 0.9232,
+      "step": 49988
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.743828571428572e-07,
+      "loss": 1.4205,
+      "step": 49989
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.741142857142857e-07,
+      "loss": 0.8379,
+      "step": 49990
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.738457142857143e-07,
+      "loss": 1.7692,
+      "step": 49991
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.735771428571428e-07,
+      "loss": 0.9608,
+      "step": 49992
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.733085714285714e-07,
+      "loss": 1.1688,
+      "step": 49993
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.730400000000001e-07,
+      "loss": 0.9987,
+      "step": 49994
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.727714285714286e-07,
+      "loss": 0.8344,
+      "step": 49995
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.725028571428572e-07,
+      "loss": 0.9471,
+      "step": 49996
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.722342857142857e-07,
+      "loss": 1.4495,
+      "step": 49997
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.719657142857143e-07,
+      "loss": 0.9421,
+      "step": 49998
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.716971428571429e-07,
+      "loss": 0.8382,
+      "step": 49999
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.714285714285714e-07,
+      "loss": 1.2937,
+      "step": 50000
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.711600000000001e-07,
+      "loss": 1.0386,
+      "step": 50001
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.708914285714286e-07,
+      "loss": 0.9538,
+      "step": 50002
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.706228571428572e-07,
+      "loss": 0.9079,
+      "step": 50003
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.703542857142858e-07,
+      "loss": 1.1928,
+      "step": 50004
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.700857142857143e-07,
+      "loss": 1.3235,
+      "step": 50005
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.698171428571429e-07,
+      "loss": 1.1512,
+      "step": 50006
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.695485714285715e-07,
+      "loss": 1.3241,
+      "step": 50007
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.6928e-07,
+      "loss": 0.8487,
+      "step": 50008
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.690114285714287e-07,
+      "loss": 1.0667,
+      "step": 50009
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.687428571428572e-07,
+      "loss": 1.2671,
+      "step": 50010
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.684742857142858e-07,
+      "loss": 0.9252,
+      "step": 50011
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.682057142857144e-07,
+      "loss": 0.8597,
+      "step": 50012
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.679371428571429e-07,
+      "loss": 1.176,
+      "step": 50013
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.676685714285715e-07,
+      "loss": 1.1433,
+      "step": 50014
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.674e-07,
+      "loss": 0.8569,
+      "step": 50015
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.671314285714286e-07,
+      "loss": 0.7745,
+      "step": 50016
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.668628571428573e-07,
+      "loss": 1.189,
+      "step": 50017
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.665942857142858e-07,
+      "loss": 1.2906,
+      "step": 50018
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.663257142857144e-07,
+      "loss": 1.2288,
+      "step": 50019
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.660571428571428e-07,
+      "loss": 1.1175,
+      "step": 50020
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.657885714285714e-07,
+      "loss": 1.4229,
+      "step": 50021
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.6552e-07,
+      "loss": 0.9873,
+      "step": 50022
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.652514285714285e-07,
+      "loss": 1.0151,
+      "step": 50023
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.649828571428571e-07,
+      "loss": 0.8369,
+      "step": 50024
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.647142857142857e-07,
+      "loss": 1.0465,
+      "step": 50025
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.644457142857143e-07,
+      "loss": 0.7961,
+      "step": 50026
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.641771428571429e-07,
+      "loss": 1.2743,
+      "step": 50027
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.639085714285714e-07,
+      "loss": 1.0664,
+      "step": 50028
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.6364e-07,
+      "loss": 1.3167,
+      "step": 50029
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.633714285714286e-07,
+      "loss": 0.9818,
+      "step": 50030
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.631028571428571e-07,
+      "loss": 0.8972,
+      "step": 50031
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.628342857142857e-07,
+      "loss": 0.7399,
+      "step": 50032
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.625657142857143e-07,
+      "loss": 0.971,
+      "step": 50033
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.622971428571429e-07,
+      "loss": 1.3113,
+      "step": 50034
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.620285714285715e-07,
+      "loss": 1.1424,
+      "step": 50035
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.6176e-07,
+      "loss": 1.2227,
+      "step": 50036
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.614914285714286e-07,
+      "loss": 1.44,
+      "step": 50037
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.612228571428571e-07,
+      "loss": 0.8802,
+      "step": 50038
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.609542857142857e-07,
+      "loss": 1.3417,
+      "step": 50039
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.606857142857143e-07,
+      "loss": 1.1891,
+      "step": 50040
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.604171428571429e-07,
+      "loss": 1.2687,
+      "step": 50041
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.601485714285715e-07,
+      "loss": 1.217,
+      "step": 50042
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.5988e-07,
+      "loss": 1.0793,
+      "step": 50043
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.596114285714286e-07,
+      "loss": 1.3823,
+      "step": 50044
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.593428571428572e-07,
+      "loss": 1.2564,
+      "step": 50045
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.590742857142857e-07,
+      "loss": 1.0304,
+      "step": 50046
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.588057142857143e-07,
+      "loss": 0.8944,
+      "step": 50047
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.585371428571428e-07,
+      "loss": 1.0778,
+      "step": 50048
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.582685714285715e-07,
+      "loss": 1.3785,
+      "step": 50049
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.580000000000001e-07,
+      "loss": 1.029,
+      "step": 50050
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.577314285714286e-07,
+      "loss": 1.1608,
+      "step": 50051
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.574628571428572e-07,
+      "loss": 1.1913,
+      "step": 50052
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.571942857142857e-07,
+      "loss": 0.8507,
+      "step": 50053
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.569257142857143e-07,
+      "loss": 1.1183,
+      "step": 50054
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.566571428571429e-07,
+      "loss": 0.9581,
+      "step": 50055
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.563885714285714e-07,
+      "loss": 1.3134,
+      "step": 50056
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.561200000000001e-07,
+      "loss": 1.0055,
+      "step": 50057
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.558514285714286e-07,
+      "loss": 0.9044,
+      "step": 50058
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.555828571428572e-07,
+      "loss": 1.1793,
+      "step": 50059
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.553142857142858e-07,
+      "loss": 0.83,
+      "step": 50060
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.550457142857143e-07,
+      "loss": 0.9226,
+      "step": 50061
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.547771428571429e-07,
+      "loss": 1.1694,
+      "step": 50062
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.545085714285715e-07,
+      "loss": 1.0894,
+      "step": 50063
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.5424e-07,
+      "loss": 0.8845,
+      "step": 50064
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.539714285714287e-07,
+      "loss": 1.7226,
+      "step": 50065
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.537028571428572e-07,
+      "loss": 0.8126,
+      "step": 50066
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.534342857142858e-07,
+      "loss": 0.9886,
+      "step": 50067
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.531657142857144e-07,
+      "loss": 1.1213,
+      "step": 50068
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.528971428571429e-07,
+      "loss": 0.9836,
+      "step": 50069
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.526285714285715e-07,
+      "loss": 1.263,
+      "step": 50070
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.523599999999999e-07,
+      "loss": 0.977,
+      "step": 50071
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.520914285714285e-07,
+      "loss": 0.8017,
+      "step": 50072
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.518228571428573e-07,
+      "loss": 0.8722,
+      "step": 50073
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.515542857142857e-07,
+      "loss": 0.65,
+      "step": 50074
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.512857142857143e-07,
+      "loss": 1.1192,
+      "step": 50075
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.510171428571428e-07,
+      "loss": 1.304,
+      "step": 50076
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.507485714285714e-07,
+      "loss": 1.014,
+      "step": 50077
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.5048e-07,
+      "loss": 0.972,
+      "step": 50078
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.502114285714285e-07,
+      "loss": 1.3344,
+      "step": 50079
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.499428571428571e-07,
+      "loss": 1.5138,
+      "step": 50080
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.496742857142857e-07,
+      "loss": 1.1666,
+      "step": 50081
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.494057142857143e-07,
+      "loss": 0.8916,
+      "step": 50082
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.491371428571429e-07,
+      "loss": 1.4046,
+      "step": 50083
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.488685714285714e-07,
+      "loss": 1.1026,
+      "step": 50084
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.486e-07,
+      "loss": 1.0706,
+      "step": 50085
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.483314285714286e-07,
+      "loss": 1.3633,
+      "step": 50086
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.480628571428571e-07,
+      "loss": 1.3506,
+      "step": 50087
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.477942857142857e-07,
+      "loss": 0.6811,
+      "step": 50088
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.475257142857143e-07,
+      "loss": 0.98,
+      "step": 50089
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.472571428571429e-07,
+      "loss": 1.0101,
+      "step": 50090
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.469885714285715e-07,
+      "loss": 1.0373,
+      "step": 50091
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.4672e-07,
+      "loss": 0.9491,
+      "step": 50092
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.464514285714286e-07,
+      "loss": 0.7128,
+      "step": 50093
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.461828571428571e-07,
+      "loss": 1.2296,
+      "step": 50094
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.459142857142857e-07,
+      "loss": 1.3237,
+      "step": 50095
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.456457142857144e-07,
+      "loss": 1.2571,
+      "step": 50096
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.453771428571429e-07,
+      "loss": 0.7316,
+      "step": 50097
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.451085714285715e-07,
+      "loss": 1.0921,
+      "step": 50098
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.4484e-07,
+      "loss": 1.4389,
+      "step": 50099
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.445714285714286e-07,
+      "loss": 1.4298,
+      "step": 50100
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.443028571428572e-07,
+      "loss": 1.1634,
+      "step": 50101
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.440342857142857e-07,
+      "loss": 0.8912,
+      "step": 50102
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.437657142857143e-07,
+      "loss": 0.6761,
+      "step": 50103
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.434971428571429e-07,
+      "loss": 0.9749,
+      "step": 50104
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.432285714285715e-07,
+      "loss": 1.3749,
+      "step": 50105
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.429600000000001e-07,
+      "loss": 1.077,
+      "step": 50106
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.426914285714286e-07,
+      "loss": 0.9739,
+      "step": 50107
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.424228571428572e-07,
+      "loss": 1.0658,
+      "step": 50108
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.421542857142857e-07,
+      "loss": 0.8785,
+      "step": 50109
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.418857142857143e-07,
+      "loss": 0.9439,
+      "step": 50110
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.416171428571429e-07,
+      "loss": 1.3878,
+      "step": 50111
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.413485714285715e-07,
+      "loss": 1.2627,
+      "step": 50112
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.410800000000001e-07,
+      "loss": 0.8749,
+      "step": 50113
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.408114285714286e-07,
+      "loss": 1.1217,
+      "step": 50114
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.405428571428572e-07,
+      "loss": 1.1888,
+      "step": 50115
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.402742857142858e-07,
+      "loss": 0.8095,
+      "step": 50116
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.400057142857143e-07,
+      "loss": 0.9813,
+      "step": 50117
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.397371428571429e-07,
+      "loss": 1.6574,
+      "step": 50118
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.394685714285715e-07,
+      "loss": 1.0412,
+      "step": 50119
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.392000000000001e-07,
+      "loss": 1.3741,
+      "step": 50120
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.389314285714287e-07,
+      "loss": 0.9478,
+      "step": 50121
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.386628571428572e-07,
+      "loss": 0.8056,
+      "step": 50122
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.383942857142858e-07,
+      "loss": 1.0478,
+      "step": 50123
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.381257142857144e-07,
+      "loss": 1.0134,
+      "step": 50124
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.378571428571428e-07,
+      "loss": 1.2306,
+      "step": 50125
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.375885714285714e-07,
+      "loss": 1.1638,
+      "step": 50126
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.373199999999999e-07,
+      "loss": 0.8658,
+      "step": 50127
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.370514285714286e-07,
+      "loss": 1.6442,
+      "step": 50128
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.367828571428572e-07,
+      "loss": 1.2581,
+      "step": 50129
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.365142857142857e-07,
+      "loss": 1.224,
+      "step": 50130
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.362457142857143e-07,
+      "loss": 1.1776,
+      "step": 50131
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.359771428571428e-07,
+      "loss": 1.5702,
+      "step": 50132
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.357085714285714e-07,
+      "loss": 1.3638,
+      "step": 50133
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.3544e-07,
+      "loss": 0.9997,
+      "step": 50134
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.351714285714285e-07,
+      "loss": 1.1852,
+      "step": 50135
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.349028571428572e-07,
+      "loss": 0.983,
+      "step": 50136
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.346342857142857e-07,
+      "loss": 0.8449,
+      "step": 50137
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.343657142857143e-07,
+      "loss": 1.4095,
+      "step": 50138
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.340971428571429e-07,
+      "loss": 0.9713,
+      "step": 50139
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.338285714285714e-07,
+      "loss": 0.9019,
+      "step": 50140
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.3356e-07,
+      "loss": 1.2753,
+      "step": 50141
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.332914285714286e-07,
+      "loss": 0.9209,
+      "step": 50142
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.330228571428571e-07,
+      "loss": 1.3463,
+      "step": 50143
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.327542857142858e-07,
+      "loss": 1.3334,
+      "step": 50144
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.324857142857143e-07,
+      "loss": 0.8864,
+      "step": 50145
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.322171428571429e-07,
+      "loss": 1.3403,
+      "step": 50146
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.319485714285715e-07,
+      "loss": 1.2345,
+      "step": 50147
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.3168e-07,
+      "loss": 1.1514,
+      "step": 50148
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.314114285714286e-07,
+      "loss": 0.9877,
+      "step": 50149
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.311428571428571e-07,
+      "loss": 1.3964,
+      "step": 50150
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.308742857142857e-07,
+      "loss": 1.1652,
+      "step": 50151
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.306057142857144e-07,
+      "loss": 1.6737,
+      "step": 50152
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.303371428571429e-07,
+      "loss": 1.2824,
+      "step": 50153
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.300685714285715e-07,
+      "loss": 1.3061,
+      "step": 50154
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.298e-07,
+      "loss": 0.5577,
+      "step": 50155
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.295314285714286e-07,
+      "loss": 1.1031,
+      "step": 50156
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.292628571428572e-07,
+      "loss": 1.2178,
+      "step": 50157
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.289942857142857e-07,
+      "loss": 1.0014,
+      "step": 50158
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.287257142857143e-07,
+      "loss": 0.6288,
+      "step": 50159
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.284571428571429e-07,
+      "loss": 1.2434,
+      "step": 50160
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.281885714285715e-07,
+      "loss": 0.8309,
+      "step": 50161
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.279200000000001e-07,
+      "loss": 0.9099,
+      "step": 50162
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.276514285714286e-07,
+      "loss": 0.9855,
+      "step": 50163
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.273828571428572e-07,
+      "loss": 0.9271,
+      "step": 50164
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.271142857142857e-07,
+      "loss": 1.283,
+      "step": 50165
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.268457142857143e-07,
+      "loss": 0.8301,
+      "step": 50166
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.265771428571429e-07,
+      "loss": 0.7337,
+      "step": 50167
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.263085714285715e-07,
+      "loss": 1.4015,
+      "step": 50168
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.260400000000001e-07,
+      "loss": 1.2063,
+      "step": 50169
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.257714285714286e-07,
+      "loss": 0.9327,
+      "step": 50170
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.255028571428572e-07,
+      "loss": 0.9253,
+      "step": 50171
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.252342857142858e-07,
+      "loss": 0.4787,
+      "step": 50172
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.249657142857143e-07,
+      "loss": 1.1682,
+      "step": 50173
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.246971428571429e-07,
+      "loss": 1.2148,
+      "step": 50174
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.244285714285715e-07,
+      "loss": 1.022,
+      "step": 50175
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.2416e-07,
+      "loss": 0.9216,
+      "step": 50176
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.238914285714287e-07,
+      "loss": 0.7771,
+      "step": 50177
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.236228571428571e-07,
+      "loss": 0.7773,
+      "step": 50178
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.233542857142857e-07,
+      "loss": 0.9028,
+      "step": 50179
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.230857142857143e-07,
+      "loss": 0.9042,
+      "step": 50180
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.228171428571428e-07,
+      "loss": 1.2418,
+      "step": 50181
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.225485714285714e-07,
+      "loss": 0.831,
+      "step": 50182
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.222799999999999e-07,
+      "loss": 0.9631,
+      "step": 50183
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.220114285714286e-07,
+      "loss": 1.3457,
+      "step": 50184
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.217428571428572e-07,
+      "loss": 1.2898,
+      "step": 50185
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.214742857142857e-07,
+      "loss": 1.5253,
+      "step": 50186
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.212057142857143e-07,
+      "loss": 1.4053,
+      "step": 50187
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.209371428571428e-07,
+      "loss": 0.8773,
+      "step": 50188
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.206685714285714e-07,
+      "loss": 1.1705,
+      "step": 50189
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.204e-07,
+      "loss": 1.2032,
+      "step": 50190
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.201314285714285e-07,
+      "loss": 0.8464,
+      "step": 50191
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.198628571428572e-07,
+      "loss": 1.239,
+      "step": 50192
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.195942857142857e-07,
+      "loss": 0.728,
+      "step": 50193
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.193257142857143e-07,
+      "loss": 0.9228,
+      "step": 50194
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.190571428571429e-07,
+      "loss": 1.0241,
+      "step": 50195
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.187885714285714e-07,
+      "loss": 1.3751,
+      "step": 50196
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.1852e-07,
+      "loss": 1.4481,
+      "step": 50197
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.182514285714286e-07,
+      "loss": 1.3217,
+      "step": 50198
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.179828571428572e-07,
+      "loss": 1.0772,
+      "step": 50199
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.177142857142858e-07,
+      "loss": 1.3034,
+      "step": 50200
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.174457142857143e-07,
+      "loss": 1.3578,
+      "step": 50201
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.171771428571429e-07,
+      "loss": 1.3402,
+      "step": 50202
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.169085714285715e-07,
+      "loss": 1.0969,
+      "step": 50203
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.1664e-07,
+      "loss": 1.1542,
+      "step": 50204
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.163714285714286e-07,
+      "loss": 1.1159,
+      "step": 50205
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.161028571428571e-07,
+      "loss": 0.82,
+      "step": 50206
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.158342857142858e-07,
+      "loss": 1.1845,
+      "step": 50207
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.155657142857144e-07,
+      "loss": 1.3557,
+      "step": 50208
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.152971428571429e-07,
+      "loss": 0.8835,
+      "step": 50209
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.150285714285715e-07,
+      "loss": 1.5522,
+      "step": 50210
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.1476e-07,
+      "loss": 1.0989,
+      "step": 50211
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.144914285714286e-07,
+      "loss": 0.909,
+      "step": 50212
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.142228571428572e-07,
+      "loss": 1.3412,
+      "step": 50213
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.139542857142857e-07,
+      "loss": 0.9818,
+      "step": 50214
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.136857142857144e-07,
+      "loss": 1.0045,
+      "step": 50215
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.134171428571429e-07,
+      "loss": 1.2943,
+      "step": 50216
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.131485714285715e-07,
+      "loss": 1.2718,
+      "step": 50217
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.128800000000001e-07,
+      "loss": 1.0303,
+      "step": 50218
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.126114285714286e-07,
+      "loss": 0.9038,
+      "step": 50219
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.123428571428572e-07,
+      "loss": 0.9937,
+      "step": 50220
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.120742857142857e-07,
+      "loss": 1.3763,
+      "step": 50221
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.118057142857143e-07,
+      "loss": 1.0902,
+      "step": 50222
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.11537142857143e-07,
+      "loss": 1.2811,
+      "step": 50223
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.112685714285715e-07,
+      "loss": 1.1127,
+      "step": 50224
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.110000000000001e-07,
+      "loss": 1.1783,
+      "step": 50225
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.107314285714286e-07,
+      "loss": 1.3628,
+      "step": 50226
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.104628571428572e-07,
+      "loss": 1.1298,
+      "step": 50227
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.101942857142858e-07,
+      "loss": 0.9157,
+      "step": 50228
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.099257142857142e-07,
+      "loss": 1.1838,
+      "step": 50229
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.096571428571428e-07,
+      "loss": 0.9735,
+      "step": 50230
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.093885714285714e-07,
+      "loss": 1.3059,
+      "step": 50231
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.0912e-07,
+      "loss": 1.0853,
+      "step": 50232
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.088514285714286e-07,
+      "loss": 1.3117,
+      "step": 50233
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.085828571428571e-07,
+      "loss": 0.7221,
+      "step": 50234
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.083142857142857e-07,
+      "loss": 1.7644,
+      "step": 50235
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.080457142857143e-07,
+      "loss": 0.9583,
+      "step": 50236
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.077771428571428e-07,
+      "loss": 1.1693,
+      "step": 50237
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.075085714285714e-07,
+      "loss": 1.0234,
+      "step": 50238
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.0724e-07,
+      "loss": 0.9665,
+      "step": 50239
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.069714285714286e-07,
+      "loss": 0.8054,
+      "step": 50240
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.067028571428572e-07,
+      "loss": 1.3293,
+      "step": 50241
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.064342857142857e-07,
+      "loss": 1.2739,
+      "step": 50242
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.061657142857143e-07,
+      "loss": 1.0351,
+      "step": 50243
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.058971428571428e-07,
+      "loss": 1.508,
+      "step": 50244
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.056285714285714e-07,
+      "loss": 1.3351,
+      "step": 50245
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.0536e-07,
+      "loss": 1.2154,
+      "step": 50246
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.050914285714286e-07,
+      "loss": 0.8027,
+      "step": 50247
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.048228571428572e-07,
+      "loss": 0.7635,
+      "step": 50248
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.045542857142857e-07,
+      "loss": 1.1598,
+      "step": 50249
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.042857142857143e-07,
+      "loss": 1.2557,
+      "step": 50250
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.040171428571429e-07,
+      "loss": 1.1626,
+      "step": 50251
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.037485714285714e-07,
+      "loss": 0.9568,
+      "step": 50252
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.0348e-07,
+      "loss": 1.7829,
+      "step": 50253
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.032114285714286e-07,
+      "loss": 0.9465,
+      "step": 50254
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.029428571428572e-07,
+      "loss": 1.3794,
+      "step": 50255
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.026742857142858e-07,
+      "loss": 0.7412,
+      "step": 50256
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.024057142857143e-07,
+      "loss": 1.1304,
+      "step": 50257
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.021371428571429e-07,
+      "loss": 1.0844,
+      "step": 50258
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.018685714285715e-07,
+      "loss": 1.0697,
+      "step": 50259
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.016e-07,
+      "loss": 1.5245,
+      "step": 50260
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.013314285714286e-07,
+      "loss": 1.2181,
+      "step": 50261
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.010628571428571e-07,
+      "loss": 1.0554,
+      "step": 50262
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.007942857142858e-07,
+      "loss": 0.8979,
+      "step": 50263
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.005257142857144e-07,
+      "loss": 1.0173,
+      "step": 50264
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.002571428571429e-07,
+      "loss": 1.024,
+      "step": 50265
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.999885714285715e-07,
+      "loss": 1.0009,
+      "step": 50266
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.9972e-07,
+      "loss": 1.3898,
+      "step": 50267
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.994514285714286e-07,
+      "loss": 1.0448,
+      "step": 50268
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.991828571428572e-07,
+      "loss": 1.4983,
+      "step": 50269
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.989142857142857e-07,
+      "loss": 0.5676,
+      "step": 50270
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.986457142857144e-07,
+      "loss": 1.4784,
+      "step": 50271
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.983771428571429e-07,
+      "loss": 1.4801,
+      "step": 50272
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.981085714285715e-07,
+      "loss": 1.0449,
+      "step": 50273
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.978400000000001e-07,
+      "loss": 1.1814,
+      "step": 50274
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.975714285714286e-07,
+      "loss": 1.285,
+      "step": 50275
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.973028571428572e-07,
+      "loss": 1.3424,
+      "step": 50276
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.970342857142857e-07,
+      "loss": 0.9698,
+      "step": 50277
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.967657142857143e-07,
+      "loss": 1.3766,
+      "step": 50278
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.96497142857143e-07,
+      "loss": 1.0715,
+      "step": 50279
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.962285714285715e-07,
+      "loss": 1.6272,
+      "step": 50280
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.9596e-07,
+      "loss": 1.242,
+      "step": 50281
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.956914285714285e-07,
+      "loss": 1.1141,
+      "step": 50282
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.954228571428571e-07,
+      "loss": 0.9079,
+      "step": 50283
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.951542857142857e-07,
+      "loss": 0.9763,
+      "step": 50284
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.948857142857142e-07,
+      "loss": 1.0012,
+      "step": 50285
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.946171428571428e-07,
+      "loss": 0.9959,
+      "step": 50286
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.943485714285714e-07,
+      "loss": 0.7859,
+      "step": 50287
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.9408e-07,
+      "loss": 0.9044,
+      "step": 50288
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.938114285714286e-07,
+      "loss": 1.0337,
+      "step": 50289
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.935428571428571e-07,
+      "loss": 1.2531,
+      "step": 50290
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.932742857142857e-07,
+      "loss": 0.7525,
+      "step": 50291
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.930057142857143e-07,
+      "loss": 1.15,
+      "step": 50292
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.927371428571428e-07,
+      "loss": 1.0156,
+      "step": 50293
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.924685714285714e-07,
+      "loss": 0.9433,
+      "step": 50294
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.922e-07,
+      "loss": 1.1053,
+      "step": 50295
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.919314285714286e-07,
+      "loss": 0.7885,
+      "step": 50296
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.916628571428572e-07,
+      "loss": 1.2948,
+      "step": 50297
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.913942857142857e-07,
+      "loss": 1.0137,
+      "step": 50298
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.911257142857143e-07,
+      "loss": 0.6856,
+      "step": 50299
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.908571428571428e-07,
+      "loss": 1.4205,
+      "step": 50300
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.905885714285714e-07,
+      "loss": 0.9435,
+      "step": 50301
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.903200000000001e-07,
+      "loss": 1.5503,
+      "step": 50302
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.900514285714286e-07,
+      "loss": 1.4709,
+      "step": 50303
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.897828571428572e-07,
+      "loss": 1.0273,
+      "step": 50304
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.895142857142857e-07,
+      "loss": 0.8822,
+      "step": 50305
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.892457142857143e-07,
+      "loss": 1.2044,
+      "step": 50306
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.889771428571429e-07,
+      "loss": 1.5547,
+      "step": 50307
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.887085714285714e-07,
+      "loss": 1.1009,
+      "step": 50308
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.8844e-07,
+      "loss": 0.7121,
+      "step": 50309
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.881714285714286e-07,
+      "loss": 1.3943,
+      "step": 50310
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.879028571428572e-07,
+      "loss": 1.462,
+      "step": 50311
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 5.876342857142858e-07,
+      "loss": 0.74,
+      "step": 50312
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.873657142857143e-07,
+      "loss": 1.0387,
+      "step": 50313
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.870971428571429e-07,
+      "loss": 1.0657,
+      "step": 50314
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.868285714285715e-07,
+      "loss": 0.5699,
+      "step": 50315
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.8656e-07,
+      "loss": 1.6564,
+      "step": 50316
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.862914285714286e-07,
+      "loss": 1.036,
+      "step": 50317
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.860228571428572e-07,
+      "loss": 0.9421,
+      "step": 50318
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.857542857142858e-07,
+      "loss": 1.4143,
+      "step": 50319
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.854857142857144e-07,
+      "loss": 1.0577,
+      "step": 50320
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.852171428571429e-07,
+      "loss": 0.7933,
+      "step": 50321
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.849485714285715e-07,
+      "loss": 0.9225,
+      "step": 50322
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.8468e-07,
+      "loss": 0.826,
+      "step": 50323
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.844114285714286e-07,
+      "loss": 1.1789,
+      "step": 50324
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.841428571428572e-07,
+      "loss": 1.3509,
+      "step": 50325
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.838742857142858e-07,
+      "loss": 1.4083,
+      "step": 50326
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.836057142857144e-07,
+      "loss": 1.229,
+      "step": 50327
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.833371428571429e-07,
+      "loss": 1.2884,
+      "step": 50328
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.830685714285715e-07,
+      "loss": 1.4194,
+      "step": 50329
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.828000000000001e-07,
+      "loss": 0.7444,
+      "step": 50330
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.825314285714286e-07,
+      "loss": 1.0129,
+      "step": 50331
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.822628571428572e-07,
+      "loss": 1.1161,
+      "step": 50332
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.819942857142856e-07,
+      "loss": 1.2091,
+      "step": 50333
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.817257142857143e-07,
+      "loss": 1.5173,
+      "step": 50334
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.81457142857143e-07,
+      "loss": 1.3409,
+      "step": 50335
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.811885714285714e-07,
+      "loss": 0.8408,
+      "step": 50336
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.8092e-07,
+      "loss": 0.9617,
+      "step": 50337
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.806514285714285e-07,
+      "loss": 1.5306,
+      "step": 50338
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.803828571428571e-07,
+      "loss": 0.9228,
+      "step": 50339
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.801142857142857e-07,
+      "loss": 1.4893,
+      "step": 50340
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.798457142857142e-07,
+      "loss": 1.2211,
+      "step": 50341
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.795771428571429e-07,
+      "loss": 1.1351,
+      "step": 50342
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.793085714285714e-07,
+      "loss": 1.2264,
+      "step": 50343
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.7904e-07,
+      "loss": 1.2017,
+      "step": 50344
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.787714285714286e-07,
+      "loss": 1.0219,
+      "step": 50345
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.785028571428571e-07,
+      "loss": 1.4177,
+      "step": 50346
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.782342857142857e-07,
+      "loss": 0.9783,
+      "step": 50347
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.779657142857143e-07,
+      "loss": 1.1716,
+      "step": 50348
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.776971428571428e-07,
+      "loss": 1.0722,
+      "step": 50349
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.774285714285715e-07,
+      "loss": 0.894,
+      "step": 50350
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.7716e-07,
+      "loss": 1.3179,
+      "step": 50351
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.768914285714286e-07,
+      "loss": 1.1296,
+      "step": 50352
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.766228571428572e-07,
+      "loss": 0.9714,
+      "step": 50353
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.763542857142857e-07,
+      "loss": 1.246,
+      "step": 50354
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.760857142857143e-07,
+      "loss": 0.8898,
+      "step": 50355
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.758171428571428e-07,
+      "loss": 1.0881,
+      "step": 50356
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.755485714285714e-07,
+      "loss": 1.1457,
+      "step": 50357
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.752800000000001e-07,
+      "loss": 1.2594,
+      "step": 50358
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.750114285714286e-07,
+      "loss": 1.1151,
+      "step": 50359
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.747428571428572e-07,
+      "loss": 1.1369,
+      "step": 50360
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.744742857142857e-07,
+      "loss": 1.0247,
+      "step": 50361
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.742057142857143e-07,
+      "loss": 0.5531,
+      "step": 50362
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.739371428571429e-07,
+      "loss": 1.0571,
+      "step": 50363
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.736685714285714e-07,
+      "loss": 1.1491,
+      "step": 50364
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.734e-07,
+      "loss": 1.163,
+      "step": 50365
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.731314285714286e-07,
+      "loss": 1.0829,
+      "step": 50366
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.728628571428572e-07,
+      "loss": 0.8647,
+      "step": 50367
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.725942857142858e-07,
+      "loss": 1.1293,
+      "step": 50368
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.723257142857143e-07,
+      "loss": 1.1499,
+      "step": 50369
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.720571428571429e-07,
+      "loss": 1.5331,
+      "step": 50370
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.717885714285715e-07,
+      "loss": 1.0856,
+      "step": 50371
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.7152e-07,
+      "loss": 1.1972,
+      "step": 50372
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.712514285714286e-07,
+      "loss": 1.1165,
+      "step": 50373
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.709828571428572e-07,
+      "loss": 1.3162,
+      "step": 50374
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.707142857142858e-07,
+      "loss": 1.5778,
+      "step": 50375
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.704457142857144e-07,
+      "loss": 1.1339,
+      "step": 50376
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.701771428571429e-07,
+      "loss": 1.2071,
+      "step": 50377
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.699085714285715e-07,
+      "loss": 1.0065,
+      "step": 50378
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.6964e-07,
+      "loss": 1.3242,
+      "step": 50379
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.693714285714286e-07,
+      "loss": 1.0732,
+      "step": 50380
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.691028571428572e-07,
+      "loss": 0.8087,
+      "step": 50381
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.688342857142858e-07,
+      "loss": 1.1269,
+      "step": 50382
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.685657142857144e-07,
+      "loss": 0.8853,
+      "step": 50383
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.682971428571429e-07,
+      "loss": 1.2606,
+      "step": 50384
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.680285714285714e-07,
+      "loss": 1.0888,
+      "step": 50385
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.6776e-07,
+      "loss": 1.4411,
+      "step": 50386
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.674914285714285e-07,
+      "loss": 1.0636,
+      "step": 50387
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.672228571428571e-07,
+      "loss": 0.8992,
+      "step": 50388
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.669542857142856e-07,
+      "loss": 1.0126,
+      "step": 50389
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.666857142857143e-07,
+      "loss": 1.1875,
+      "step": 50390
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.664171428571429e-07,
+      "loss": 0.8436,
+      "step": 50391
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.661485714285714e-07,
+      "loss": 1.4117,
+      "step": 50392
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.6588e-07,
+      "loss": 0.8518,
+      "step": 50393
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.656114285714285e-07,
+      "loss": 1.1638,
+      "step": 50394
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.653428571428571e-07,
+      "loss": 1.6586,
+      "step": 50395
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.650742857142857e-07,
+      "loss": 1.1199,
+      "step": 50396
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.648057142857143e-07,
+      "loss": 1.3138,
+      "step": 50397
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.645371428571429e-07,
+      "loss": 0.8188,
+      "step": 50398
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.642685714285714e-07,
+      "loss": 0.9435,
+      "step": 50399
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.64e-07,
+      "loss": 1.0004,
+      "step": 50400
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.637314285714286e-07,
+      "loss": 1.3316,
+      "step": 50401
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.634628571428571e-07,
+      "loss": 0.8978,
+      "step": 50402
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.631942857142857e-07,
+      "loss": 1.0644,
+      "step": 50403
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.629257142857143e-07,
+      "loss": 1.4834,
+      "step": 50404
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.626571428571429e-07,
+      "loss": 1.4698,
+      "step": 50405
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.623885714285715e-07,
+      "loss": 0.8784,
+      "step": 50406
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.6212e-07,
+      "loss": 1.5344,
+      "step": 50407
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.618514285714286e-07,
+      "loss": 1.0349,
+      "step": 50408
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.615828571428572e-07,
+      "loss": 1.1198,
+      "step": 50409
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.613142857142857e-07,
+      "loss": 0.9835,
+      "step": 50410
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.610457142857143e-07,
+      "loss": 0.899,
+      "step": 50411
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.607771428571428e-07,
+      "loss": 1.3692,
+      "step": 50412
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.605085714285715e-07,
+      "loss": 1.2891,
+      "step": 50413
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.602400000000001e-07,
+      "loss": 1.2308,
+      "step": 50414
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.599714285714286e-07,
+      "loss": 1.3545,
+      "step": 50415
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.597028571428572e-07,
+      "loss": 1.1425,
+      "step": 50416
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.594342857142857e-07,
+      "loss": 0.8543,
+      "step": 50417
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.591657142857143e-07,
+      "loss": 0.9925,
+      "step": 50418
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.588971428571429e-07,
+      "loss": 0.9712,
+      "step": 50419
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.586285714285714e-07,
+      "loss": 1.0738,
+      "step": 50420
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.583600000000001e-07,
+      "loss": 1.4132,
+      "step": 50421
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.580914285714286e-07,
+      "loss": 1.0137,
+      "step": 50422
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.578228571428572e-07,
+      "loss": 1.5555,
+      "step": 50423
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.575542857142858e-07,
+      "loss": 1.1075,
+      "step": 50424
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.572857142857143e-07,
+      "loss": 0.929,
+      "step": 50425
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.570171428571429e-07,
+      "loss": 1.0237,
+      "step": 50426
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.567485714285715e-07,
+      "loss": 1.1585,
+      "step": 50427
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.5648e-07,
+      "loss": 1.4263,
+      "step": 50428
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.562114285714287e-07,
+      "loss": 1.202,
+      "step": 50429
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.559428571428572e-07,
+      "loss": 1.4633,
+      "step": 50430
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.556742857142858e-07,
+      "loss": 0.904,
+      "step": 50431
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.554057142857144e-07,
+      "loss": 1.352,
+      "step": 50432
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.551371428571429e-07,
+      "loss": 1.2459,
+      "step": 50433
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.548685714285715e-07,
+      "loss": 1.1296,
+      "step": 50434
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.546e-07,
+      "loss": 1.393,
+      "step": 50435
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.543314285714285e-07,
+      "loss": 1.2834,
+      "step": 50436
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.540628571428573e-07,
+      "loss": 0.9755,
+      "step": 50437
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.537942857142857e-07,
+      "loss": 1.3578,
+      "step": 50438
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.535257142857143e-07,
+      "loss": 1.1984,
+      "step": 50439
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.532571428571428e-07,
+      "loss": 1.6269,
+      "step": 50440
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.529885714285714e-07,
+      "loss": 1.2519,
+      "step": 50441
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.5272e-07,
+      "loss": 0.9506,
+      "step": 50442
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.524514285714285e-07,
+      "loss": 1.0634,
+      "step": 50443
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.521828571428571e-07,
+      "loss": 0.8806,
+      "step": 50444
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.519142857142857e-07,
+      "loss": 1.1735,
+      "step": 50445
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.516457142857143e-07,
+      "loss": 1.4121,
+      "step": 50446
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.513771428571429e-07,
+      "loss": 1.208,
+      "step": 50447
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.511085714285714e-07,
+      "loss": 1.6495,
+      "step": 50448
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.5084e-07,
+      "loss": 0.9681,
+      "step": 50449
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.505714285714285e-07,
+      "loss": 1.4816,
+      "step": 50450
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.503028571428571e-07,
+      "loss": 1.3171,
+      "step": 50451
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.500342857142857e-07,
+      "loss": 0.9697,
+      "step": 50452
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.497657142857143e-07,
+      "loss": 1.4184,
+      "step": 50453
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.494971428571429e-07,
+      "loss": 1.4357,
+      "step": 50454
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.492285714285714e-07,
+      "loss": 0.7588,
+      "step": 50455
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.4896e-07,
+      "loss": 1.1068,
+      "step": 50456
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.486914285714286e-07,
+      "loss": 1.3973,
+      "step": 50457
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.484228571428571e-07,
+      "loss": 1.0426,
+      "step": 50458
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.481542857142857e-07,
+      "loss": 1.3684,
+      "step": 50459
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.478857142857143e-07,
+      "loss": 0.9143,
+      "step": 50460
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.476171428571429e-07,
+      "loss": 0.9945,
+      "step": 50461
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.473485714285715e-07,
+      "loss": 1.2876,
+      "step": 50462
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.4708e-07,
+      "loss": 1.0717,
+      "step": 50463
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.468114285714286e-07,
+      "loss": 1.172,
+      "step": 50464
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.465428571428572e-07,
+      "loss": 1.3739,
+      "step": 50465
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.462742857142857e-07,
+      "loss": 0.9434,
+      "step": 50466
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.460057142857143e-07,
+      "loss": 1.2123,
+      "step": 50467
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.457371428571428e-07,
+      "loss": 1.395,
+      "step": 50468
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.454685714285715e-07,
+      "loss": 1.2959,
+      "step": 50469
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.452000000000001e-07,
+      "loss": 0.9691,
+      "step": 50470
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.449314285714286e-07,
+      "loss": 1.776,
+      "step": 50471
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.446628571428572e-07,
+      "loss": 0.8611,
+      "step": 50472
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.443942857142857e-07,
+      "loss": 1.2885,
+      "step": 50473
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.441257142857143e-07,
+      "loss": 0.8561,
+      "step": 50474
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.438571428571429e-07,
+      "loss": 1.1394,
+      "step": 50475
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.435885714285714e-07,
+      "loss": 0.6498,
+      "step": 50476
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.433200000000001e-07,
+      "loss": 0.7244,
+      "step": 50477
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.430514285714286e-07,
+      "loss": 1.5204,
+      "step": 50478
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.427828571428572e-07,
+      "loss": 0.9932,
+      "step": 50479
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.425142857142858e-07,
+      "loss": 1.2759,
+      "step": 50480
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.422457142857143e-07,
+      "loss": 1.4693,
+      "step": 50481
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.419771428571429e-07,
+      "loss": 1.4751,
+      "step": 50482
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.417085714285715e-07,
+      "loss": 1.1137,
+      "step": 50483
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.4144e-07,
+      "loss": 1.5328,
+      "step": 50484
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.411714285714287e-07,
+      "loss": 1.2504,
+      "step": 50485
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.409028571428572e-07,
+      "loss": 1.1775,
+      "step": 50486
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 5.406342857142858e-07,
+      "loss": 1.1201,
+      "step": 50487
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.403657142857144e-07,
+      "loss": 1.0407,
+      "step": 50488
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.400971428571428e-07,
+      "loss": 0.9069,
+      "step": 50489
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.398285714285714e-07,
+      "loss": 0.9081,
+      "step": 50490
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.395599999999999e-07,
+      "loss": 1.3471,
+      "step": 50491
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.392914285714285e-07,
+      "loss": 1.2715,
+      "step": 50492
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.390228571428572e-07,
+      "loss": 0.9714,
+      "step": 50493
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.387542857142857e-07,
+      "loss": 1.674,
+      "step": 50494
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.384857142857143e-07,
+      "loss": 1.244,
+      "step": 50495
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.382171428571428e-07,
+      "loss": 1.0627,
+      "step": 50496
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.379485714285714e-07,
+      "loss": 1.4847,
+      "step": 50497
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.3768e-07,
+      "loss": 1.2692,
+      "step": 50498
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.374114285714285e-07,
+      "loss": 1.1692,
+      "step": 50499
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.371428571428572e-07,
+      "loss": 1.0152,
+      "step": 50500
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.368742857142857e-07,
+      "loss": 1.0121,
+      "step": 50501
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.366057142857143e-07,
+      "loss": 0.6654,
+      "step": 50502
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.363371428571429e-07,
+      "loss": 0.9141,
+      "step": 50503
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.360685714285714e-07,
+      "loss": 1.24,
+      "step": 50504
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.358e-07,
+      "loss": 1.0952,
+      "step": 50505
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.355314285714285e-07,
+      "loss": 1.6088,
+      "step": 50506
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.352628571428571e-07,
+      "loss": 0.8365,
+      "step": 50507
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.349942857142858e-07,
+      "loss": 0.6982,
+      "step": 50508
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.347257142857143e-07,
+      "loss": 1.6444,
+      "step": 50509
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.344571428571429e-07,
+      "loss": 1.3657,
+      "step": 50510
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.341885714285714e-07,
+      "loss": 1.0004,
+      "step": 50511
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.3392e-07,
+      "loss": 0.871,
+      "step": 50512
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.336514285714286e-07,
+      "loss": 1.1212,
+      "step": 50513
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.333828571428571e-07,
+      "loss": 1.1486,
+      "step": 50514
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.331142857142857e-07,
+      "loss": 1.2724,
+      "step": 50515
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.328457142857144e-07,
+      "loss": 0.961,
+      "step": 50516
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.325771428571429e-07,
+      "loss": 1.2653,
+      "step": 50517
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.323085714285715e-07,
+      "loss": 0.8589,
+      "step": 50518
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.3204e-07,
+      "loss": 1.3467,
+      "step": 50519
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.317714285714286e-07,
+      "loss": 0.902,
+      "step": 50520
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.315028571428572e-07,
+      "loss": 1.1172,
+      "step": 50521
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.312342857142857e-07,
+      "loss": 1.3426,
+      "step": 50522
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.309657142857143e-07,
+      "loss": 1.4874,
+      "step": 50523
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.306971428571429e-07,
+      "loss": 1.2939,
+      "step": 50524
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.304285714285715e-07,
+      "loss": 1.3071,
+      "step": 50525
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.301600000000001e-07,
+      "loss": 1.0338,
+      "step": 50526
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.298914285714286e-07,
+      "loss": 1.3354,
+      "step": 50527
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.296228571428572e-07,
+      "loss": 1.0562,
+      "step": 50528
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.293542857142857e-07,
+      "loss": 1.3155,
+      "step": 50529
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.290857142857143e-07,
+      "loss": 1.3528,
+      "step": 50530
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.288171428571429e-07,
+      "loss": 1.1506,
+      "step": 50531
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.285485714285715e-07,
+      "loss": 1.1815,
+      "step": 50532
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.282800000000001e-07,
+      "loss": 0.8906,
+      "step": 50533
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.280114285714286e-07,
+      "loss": 1.2575,
+      "step": 50534
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.277428571428572e-07,
+      "loss": 1.0648,
+      "step": 50535
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.274742857142858e-07,
+      "loss": 1.022,
+      "step": 50536
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.272057142857143e-07,
+      "loss": 1.0633,
+      "step": 50537
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.269371428571429e-07,
+      "loss": 1.2549,
+      "step": 50538
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.266685714285715e-07,
+      "loss": 0.9015,
+      "step": 50539
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.264000000000001e-07,
+      "loss": 0.7143,
+      "step": 50540
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.261314285714287e-07,
+      "loss": 1.0652,
+      "step": 50541
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.258628571428571e-07,
+      "loss": 1.1372,
+      "step": 50542
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.255942857142857e-07,
+      "loss": 1.4211,
+      "step": 50543
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.253257142857143e-07,
+      "loss": 1.0701,
+      "step": 50544
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.250571428571428e-07,
+      "loss": 0.871,
+      "step": 50545
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.247885714285714e-07,
+      "loss": 1.0355,
+      "step": 50546
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.245199999999999e-07,
+      "loss": 1.3113,
+      "step": 50547
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.242514285714286e-07,
+      "loss": 1.1763,
+      "step": 50548
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.239828571428572e-07,
+      "loss": 1.2676,
+      "step": 50549
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.237142857142857e-07,
+      "loss": 1.0079,
+      "step": 50550
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.234457142857143e-07,
+      "loss": 1.0149,
+      "step": 50551
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.231771428571428e-07,
+      "loss": 1.4998,
+      "step": 50552
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.229085714285714e-07,
+      "loss": 1.2559,
+      "step": 50553
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.2264e-07,
+      "loss": 0.9998,
+      "step": 50554
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.223714285714285e-07,
+      "loss": 0.9655,
+      "step": 50555
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.221028571428572e-07,
+      "loss": 1.3138,
+      "step": 50556
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.218342857142857e-07,
+      "loss": 1.7698,
+      "step": 50557
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.215657142857143e-07,
+      "loss": 0.7974,
+      "step": 50558
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.212971428571429e-07,
+      "loss": 1.1482,
+      "step": 50559
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.210285714285714e-07,
+      "loss": 1.1861,
+      "step": 50560
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.2076e-07,
+      "loss": 1.1222,
+      "step": 50561
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.204914285714285e-07,
+      "loss": 0.9971,
+      "step": 50562
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.202228571428571e-07,
+      "loss": 0.7464,
+      "step": 50563
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.199542857142858e-07,
+      "loss": 1.52,
+      "step": 50564
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.196857142857143e-07,
+      "loss": 1.0844,
+      "step": 50565
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.194171428571429e-07,
+      "loss": 1.5322,
+      "step": 50566
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.191485714285715e-07,
+      "loss": 1.0137,
+      "step": 50567
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.1888e-07,
+      "loss": 0.9363,
+      "step": 50568
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.186114285714286e-07,
+      "loss": 1.5829,
+      "step": 50569
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.183428571428571e-07,
+      "loss": 1.3319,
+      "step": 50570
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.180742857142857e-07,
+      "loss": 1.4773,
+      "step": 50571
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.178057142857144e-07,
+      "loss": 1.2442,
+      "step": 50572
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.175371428571429e-07,
+      "loss": 0.805,
+      "step": 50573
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.172685714285715e-07,
+      "loss": 0.9145,
+      "step": 50574
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.17e-07,
+      "loss": 0.9958,
+      "step": 50575
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.167314285714286e-07,
+      "loss": 1.3666,
+      "step": 50576
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.164628571428572e-07,
+      "loss": 1.1139,
+      "step": 50577
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.161942857142857e-07,
+      "loss": 0.9873,
+      "step": 50578
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.159257142857143e-07,
+      "loss": 1.2284,
+      "step": 50579
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.156571428571429e-07,
+      "loss": 1.199,
+      "step": 50580
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.153885714285715e-07,
+      "loss": 1.4276,
+      "step": 50581
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.151200000000001e-07,
+      "loss": 1.1515,
+      "step": 50582
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.148514285714286e-07,
+      "loss": 1.4138,
+      "step": 50583
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.145828571428572e-07,
+      "loss": 0.9841,
+      "step": 50584
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.143142857142857e-07,
+      "loss": 1.4448,
+      "step": 50585
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.140457142857143e-07,
+      "loss": 1.2262,
+      "step": 50586
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.137771428571429e-07,
+      "loss": 1.2347,
+      "step": 50587
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.135085714285715e-07,
+      "loss": 0.6845,
+      "step": 50588
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.132400000000001e-07,
+      "loss": 1.6465,
+      "step": 50589
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.129714285714286e-07,
+      "loss": 0.7289,
+      "step": 50590
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.127028571428572e-07,
+      "loss": 1.0525,
+      "step": 50591
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.124342857142858e-07,
+      "loss": 1.0018,
+      "step": 50592
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.121657142857142e-07,
+      "loss": 1.1649,
+      "step": 50593
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.118971428571428e-07,
+      "loss": 1.0212,
+      "step": 50594
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.116285714285714e-07,
+      "loss": 0.8107,
+      "step": 50595
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.1136e-07,
+      "loss": 0.8959,
+      "step": 50596
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.110914285714286e-07,
+      "loss": 1.0225,
+      "step": 50597
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.108228571428571e-07,
+      "loss": 1.1618,
+      "step": 50598
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.105542857142857e-07,
+      "loss": 1.5841,
+      "step": 50599
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.102857142857143e-07,
+      "loss": 1.3324,
+      "step": 50600
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.100171428571428e-07,
+      "loss": 1.4361,
+      "step": 50601
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.097485714285714e-07,
+      "loss": 0.9602,
+      "step": 50602
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.0948e-07,
+      "loss": 1.301,
+      "step": 50603
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.092114285714286e-07,
+      "loss": 1.5803,
+      "step": 50604
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.089428571428572e-07,
+      "loss": 1.2913,
+      "step": 50605
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.086742857142857e-07,
+      "loss": 0.8313,
+      "step": 50606
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.084057142857143e-07,
+      "loss": 1.2768,
+      "step": 50607
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.081371428571428e-07,
+      "loss": 1.188,
+      "step": 50608
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.078685714285714e-07,
+      "loss": 1.1168,
+      "step": 50609
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.076e-07,
+      "loss": 0.9621,
+      "step": 50610
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.073314285714286e-07,
+      "loss": 1.2039,
+      "step": 50611
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.070628571428572e-07,
+      "loss": 1.3121,
+      "step": 50612
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.067942857142857e-07,
+      "loss": 1.022,
+      "step": 50613
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.065257142857143e-07,
+      "loss": 1.2894,
+      "step": 50614
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.062571428571429e-07,
+      "loss": 1.0452,
+      "step": 50615
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.059885714285714e-07,
+      "loss": 0.7405,
+      "step": 50616
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.0572e-07,
+      "loss": 1.2309,
+      "step": 50617
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.054514285714285e-07,
+      "loss": 0.9616,
+      "step": 50618
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.051828571428572e-07,
+      "loss": 1.0912,
+      "step": 50619
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.049142857142858e-07,
+      "loss": 0.7618,
+      "step": 50620
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.046457142857143e-07,
+      "loss": 1.0217,
+      "step": 50621
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.043771428571429e-07,
+      "loss": 1.5091,
+      "step": 50622
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.041085714285715e-07,
+      "loss": 1.379,
+      "step": 50623
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.0384e-07,
+      "loss": 0.6121,
+      "step": 50624
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.035714285714286e-07,
+      "loss": 1.3561,
+      "step": 50625
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.033028571428571e-07,
+      "loss": 0.8481,
+      "step": 50626
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.030342857142858e-07,
+      "loss": 1.3858,
+      "step": 50627
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.027657142857144e-07,
+      "loss": 1.0715,
+      "step": 50628
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.024971428571429e-07,
+      "loss": 1.582,
+      "step": 50629
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.022285714285715e-07,
+      "loss": 1.4124,
+      "step": 50630
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.0196e-07,
+      "loss": 1.3661,
+      "step": 50631
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.016914285714286e-07,
+      "loss": 1.0431,
+      "step": 50632
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.014228571428572e-07,
+      "loss": 1.4559,
+      "step": 50633
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.011542857142857e-07,
+      "loss": 1.6961,
+      "step": 50634
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.008857142857144e-07,
+      "loss": 1.3633,
+      "step": 50635
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.006171428571429e-07,
+      "loss": 1.1246,
+      "step": 50636
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.003485714285715e-07,
+      "loss": 1.0677,
+      "step": 50637
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 5.000800000000001e-07,
+      "loss": 1.2153,
+      "step": 50638
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.998114285714286e-07,
+      "loss": 1.0081,
+      "step": 50639
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.995428571428572e-07,
+      "loss": 1.2152,
+      "step": 50640
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.992742857142857e-07,
+      "loss": 1.3449,
+      "step": 50641
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.990057142857143e-07,
+      "loss": 1.4328,
+      "step": 50642
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.98737142857143e-07,
+      "loss": 1.0742,
+      "step": 50643
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.984685714285715e-07,
+      "loss": 1.1265,
+      "step": 50644
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.982000000000001e-07,
+      "loss": 1.4414,
+      "step": 50645
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.979314285714285e-07,
+      "loss": 1.2261,
+      "step": 50646
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.976628571428571e-07,
+      "loss": 1.2105,
+      "step": 50647
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.973942857142857e-07,
+      "loss": 0.8951,
+      "step": 50648
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.971257142857142e-07,
+      "loss": 1.0377,
+      "step": 50649
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.968571428571428e-07,
+      "loss": 1.4189,
+      "step": 50650
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.965885714285714e-07,
+      "loss": 0.8715,
+      "step": 50651
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.9632e-07,
+      "loss": 1.23,
+      "step": 50652
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.960514285714286e-07,
+      "loss": 1.0142,
+      "step": 50653
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.957828571428571e-07,
+      "loss": 1.0753,
+      "step": 50654
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.955142857142857e-07,
+      "loss": 1.1933,
+      "step": 50655
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.952457142857143e-07,
+      "loss": 0.9282,
+      "step": 50656
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.949771428571428e-07,
+      "loss": 1.4452,
+      "step": 50657
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.947085714285714e-07,
+      "loss": 0.9746,
+      "step": 50658
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.9444e-07,
+      "loss": 1.4115,
+      "step": 50659
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.941714285714286e-07,
+      "loss": 0.9314,
+      "step": 50660
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.939028571428572e-07,
+      "loss": 1.2549,
+      "step": 50661
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.936342857142857e-07,
+      "loss": 1.0324,
+      "step": 50662
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.933657142857143e-07,
+      "loss": 0.94,
+      "step": 50663
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.930971428571428e-07,
+      "loss": 1.3565,
+      "step": 50664
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.928285714285714e-07,
+      "loss": 1.4212,
+      "step": 50665
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.9256e-07,
+      "loss": 1.1525,
+      "step": 50666
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.922914285714286e-07,
+      "loss": 0.9284,
+      "step": 50667
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.920228571428572e-07,
+      "loss": 0.7914,
+      "step": 50668
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.917542857142857e-07,
+      "loss": 1.1296,
+      "step": 50669
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.914857142857143e-07,
+      "loss": 1.5667,
+      "step": 50670
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.912171428571429e-07,
+      "loss": 1.2953,
+      "step": 50671
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.909485714285714e-07,
+      "loss": 1.0866,
+      "step": 50672
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.9068e-07,
+      "loss": 0.5036,
+      "step": 50673
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.904114285714285e-07,
+      "loss": 1.6196,
+      "step": 50674
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.901428571428572e-07,
+      "loss": 1.368,
+      "step": 50675
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.898742857142858e-07,
+      "loss": 1.2071,
+      "step": 50676
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.896057142857143e-07,
+      "loss": 1.1003,
+      "step": 50677
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.893371428571429e-07,
+      "loss": 0.8996,
+      "step": 50678
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.890685714285715e-07,
+      "loss": 0.9712,
+      "step": 50679
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.888e-07,
+      "loss": 0.871,
+      "step": 50680
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.885314285714286e-07,
+      "loss": 0.5126,
+      "step": 50681
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.882628571428571e-07,
+      "loss": 1.182,
+      "step": 50682
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.879942857142858e-07,
+      "loss": 0.9595,
+      "step": 50683
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.877257142857144e-07,
+      "loss": 0.636,
+      "step": 50684
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.874571428571429e-07,
+      "loss": 1.0604,
+      "step": 50685
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.871885714285715e-07,
+      "loss": 1.2816,
+      "step": 50686
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.8692e-07,
+      "loss": 1.2932,
+      "step": 50687
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.866514285714286e-07,
+      "loss": 0.8446,
+      "step": 50688
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.863828571428572e-07,
+      "loss": 1.2778,
+      "step": 50689
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.861142857142857e-07,
+      "loss": 1.0714,
+      "step": 50690
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.858457142857144e-07,
+      "loss": 0.9386,
+      "step": 50691
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.855771428571429e-07,
+      "loss": 0.7499,
+      "step": 50692
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.853085714285715e-07,
+      "loss": 0.8931,
+      "step": 50693
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.850400000000001e-07,
+      "loss": 0.9896,
+      "step": 50694
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.847714285714286e-07,
+      "loss": 0.8143,
+      "step": 50695
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.845028571428572e-07,
+      "loss": 0.8564,
+      "step": 50696
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.842342857142856e-07,
+      "loss": 0.9549,
+      "step": 50697
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.839657142857144e-07,
+      "loss": 0.6569,
+      "step": 50698
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.83697142857143e-07,
+      "loss": 0.9649,
+      "step": 50699
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.834285714285714e-07,
+      "loss": 1.3409,
+      "step": 50700
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.8316e-07,
+      "loss": 1.2184,
+      "step": 50701
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.828914285714285e-07,
+      "loss": 1.4882,
+      "step": 50702
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.826228571428571e-07,
+      "loss": 1.492,
+      "step": 50703
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.823542857142857e-07,
+      "loss": 0.9395,
+      "step": 50704
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.820857142857142e-07,
+      "loss": 1.0102,
+      "step": 50705
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.818171428571429e-07,
+      "loss": 1.1341,
+      "step": 50706
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.815485714285714e-07,
+      "loss": 0.8219,
+      "step": 50707
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.8128e-07,
+      "loss": 1.4662,
+      "step": 50708
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.810114285714286e-07,
+      "loss": 0.7517,
+      "step": 50709
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.807428571428571e-07,
+      "loss": 1.4739,
+      "step": 50710
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.804742857142857e-07,
+      "loss": 0.8658,
+      "step": 50711
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.802057142857143e-07,
+      "loss": 1.1583,
+      "step": 50712
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.799371428571428e-07,
+      "loss": 1.0709,
+      "step": 50713
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.796685714285715e-07,
+      "loss": 1.1595,
+      "step": 50714
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.794e-07,
+      "loss": 0.7887,
+      "step": 50715
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.791314285714286e-07,
+      "loss": 1.0955,
+      "step": 50716
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.788628571428572e-07,
+      "loss": 0.8112,
+      "step": 50717
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.785942857142857e-07,
+      "loss": 0.9082,
+      "step": 50718
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.783257142857143e-07,
+      "loss": 1.1827,
+      "step": 50719
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.780571428571428e-07,
+      "loss": 0.7153,
+      "step": 50720
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.777885714285714e-07,
+      "loss": 1.3099,
+      "step": 50721
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.775200000000001e-07,
+      "loss": 1.2451,
+      "step": 50722
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.772514285714286e-07,
+      "loss": 1.1799,
+      "step": 50723
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.769828571428572e-07,
+      "loss": 0.9458,
+      "step": 50724
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7671428571428575e-07,
+      "loss": 1.1737,
+      "step": 50725
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.764457142857143e-07,
+      "loss": 0.9413,
+      "step": 50726
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7617714285714284e-07,
+      "loss": 1.2054,
+      "step": 50727
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7590857142857144e-07,
+      "loss": 0.9725,
+      "step": 50728
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7564e-07,
+      "loss": 1.1694,
+      "step": 50729
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7537142857142864e-07,
+      "loss": 0.8372,
+      "step": 50730
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.751028571428572e-07,
+      "loss": 0.8501,
+      "step": 50731
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7483428571428574e-07,
+      "loss": 0.7901,
+      "step": 50732
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7456571428571434e-07,
+      "loss": 0.9303,
+      "step": 50733
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.742971428571429e-07,
+      "loss": 0.9727,
+      "step": 50734
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7402857142857143e-07,
+      "loss": 1.361,
+      "step": 50735
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7376e-07,
+      "loss": 0.721,
+      "step": 50736
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.734914285714285e-07,
+      "loss": 0.9543,
+      "step": 50737
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7322285714285723e-07,
+      "loss": 1.1771,
+      "step": 50738
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.729542857142858e-07,
+      "loss": 1.7271,
+      "step": 50739
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.726857142857143e-07,
+      "loss": 1.067,
+      "step": 50740
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7241714285714287e-07,
+      "loss": 0.9463,
+      "step": 50741
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.721485714285714e-07,
+      "loss": 1.3367,
+      "step": 50742
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7188e-07,
+      "loss": 1.3088,
+      "step": 50743
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7161142857142856e-07,
+      "loss": 1.187,
+      "step": 50744
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.713428571428571e-07,
+      "loss": 0.7598,
+      "step": 50745
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7107428571428576e-07,
+      "loss": 1.1848,
+      "step": 50746
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7080571428571436e-07,
+      "loss": 0.9819,
+      "step": 50747
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.705371428571429e-07,
+      "loss": 0.705,
+      "step": 50748
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7026857142857145e-07,
+      "loss": 1.1052,
+      "step": 50749
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.7e-07,
+      "loss": 0.98,
+      "step": 50750
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6973142857142855e-07,
+      "loss": 0.8267,
+      "step": 50751
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6946285714285715e-07,
+      "loss": 0.9008,
+      "step": 50752
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.691942857142857e-07,
+      "loss": 1.1633,
+      "step": 50753
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6892571428571435e-07,
+      "loss": 1.0755,
+      "step": 50754
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.686571428571429e-07,
+      "loss": 1.2503,
+      "step": 50755
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6838857142857144e-07,
+      "loss": 1.0201,
+      "step": 50756
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6812000000000004e-07,
+      "loss": 1.1784,
+      "step": 50757
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.678514285714286e-07,
+      "loss": 0.9779,
+      "step": 50758
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6758285714285713e-07,
+      "loss": 1.393,
+      "step": 50759
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.673142857142857e-07,
+      "loss": 1.4275,
+      "step": 50760
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.670457142857143e-07,
+      "loss": 1.0786,
+      "step": 50761
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6677714285714293e-07,
+      "loss": 0.7549,
+      "step": 50762
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.665085714285715e-07,
+      "loss": 1.0246,
+      "step": 50763
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6624e-07,
+      "loss": 0.8708,
+      "step": 50764
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6597142857142857e-07,
+      "loss": 1.0753,
+      "step": 50765
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6570285714285717e-07,
+      "loss": 0.8217,
+      "step": 50766
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.654342857142857e-07,
+      "loss": 0.9915,
+      "step": 50767
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6516571428571426e-07,
+      "loss": 1.0134,
+      "step": 50768
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.648971428571428e-07,
+      "loss": 1.2118,
+      "step": 50769
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6462857142857146e-07,
+      "loss": 0.7796,
+      "step": 50770
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6436000000000006e-07,
+      "loss": 0.7751,
+      "step": 50771
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.640914285714286e-07,
+      "loss": 1.0815,
+      "step": 50772
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6382285714285716e-07,
+      "loss": 0.8747,
+      "step": 50773
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.635542857142857e-07,
+      "loss": 1.0092,
+      "step": 50774
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.632857142857143e-07,
+      "loss": 0.9816,
+      "step": 50775
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6301714285714285e-07,
+      "loss": 1.2932,
+      "step": 50776
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.627485714285714e-07,
+      "loss": 0.8215,
+      "step": 50777
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6248000000000005e-07,
+      "loss": 0.6565,
+      "step": 50778
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.622114285714286e-07,
+      "loss": 1.3481,
+      "step": 50779
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.619428571428572e-07,
+      "loss": 1.2908,
+      "step": 50780
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6167428571428574e-07,
+      "loss": 0.8248,
+      "step": 50781
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.614057142857143e-07,
+      "loss": 1.2425,
+      "step": 50782
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6113714285714284e-07,
+      "loss": 0.9684,
+      "step": 50783
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6086857142857144e-07,
+      "loss": 0.5968,
+      "step": 50784
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.606e-07,
+      "loss": 1.4566,
+      "step": 50785
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.6033142857142863e-07,
+      "loss": 1.3522,
+      "step": 50786
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.600628571428572e-07,
+      "loss": 1.2102,
+      "step": 50787
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5979428571428573e-07,
+      "loss": 1.8424,
+      "step": 50788
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5952571428571433e-07,
+      "loss": 1.0499,
+      "step": 50789
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.592571428571429e-07,
+      "loss": 1.3185,
+      "step": 50790
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.589885714285714e-07,
+      "loss": 1.5631,
+      "step": 50791
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5871999999999997e-07,
+      "loss": 1.2279,
+      "step": 50792
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.584514285714285e-07,
+      "loss": 1.4274,
+      "step": 50793
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.581828571428572e-07,
+      "loss": 0.9725,
+      "step": 50794
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5791428571428577e-07,
+      "loss": 0.9985,
+      "step": 50795
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.576457142857143e-07,
+      "loss": 1.4359,
+      "step": 50796
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5737714285714286e-07,
+      "loss": 1.3915,
+      "step": 50797
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.571085714285714e-07,
+      "loss": 0.84,
+      "step": 50798
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5684e-07,
+      "loss": 1.2685,
+      "step": 50799
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5657142857142855e-07,
+      "loss": 1.0991,
+      "step": 50800
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.563028571428572e-07,
+      "loss": 1.3554,
+      "step": 50801
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5603428571428575e-07,
+      "loss": 1.1591,
+      "step": 50802
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5576571428571435e-07,
+      "loss": 0.6292,
+      "step": 50803
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.554971428571429e-07,
+      "loss": 0.6037,
+      "step": 50804
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5522857142857145e-07,
+      "loss": 0.9496,
+      "step": 50805
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5496e-07,
+      "loss": 1.1962,
+      "step": 50806
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5469142857142854e-07,
+      "loss": 1.1104,
+      "step": 50807
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5442285714285714e-07,
+      "loss": 1.133,
+      "step": 50808
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.541542857142858e-07,
+      "loss": 0.8587,
+      "step": 50809
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5388571428571434e-07,
+      "loss": 1.1185,
+      "step": 50810
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.536171428571429e-07,
+      "loss": 1.035,
+      "step": 50811
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5334857142857143e-07,
+      "loss": 1.3737,
+      "step": 50812
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5308000000000003e-07,
+      "loss": 0.6725,
+      "step": 50813
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.528114285714286e-07,
+      "loss": 1.146,
+      "step": 50814
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.525428571428571e-07,
+      "loss": 0.9847,
+      "step": 50815
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5227428571428567e-07,
+      "loss": 0.9899,
+      "step": 50816
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.520057142857143e-07,
+      "loss": 1.3876,
+      "step": 50817
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.517371428571429e-07,
+      "loss": 1.3961,
+      "step": 50818
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5146857142857147e-07,
+      "loss": 0.8375,
+      "step": 50819
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.512e-07,
+      "loss": 1.2988,
+      "step": 50820
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5093142857142856e-07,
+      "loss": 0.8354,
+      "step": 50821
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5066285714285716e-07,
+      "loss": 1.1343,
+      "step": 50822
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.503942857142857e-07,
+      "loss": 1.5478,
+      "step": 50823
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.5012571428571426e-07,
+      "loss": 1.4405,
+      "step": 50824
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.498571428571429e-07,
+      "loss": 1.1386,
+      "step": 50825
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.4958857142857146e-07,
+      "loss": 1.2341,
+      "step": 50826
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.4932000000000006e-07,
+      "loss": 1.2048,
+      "step": 50827
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.490514285714286e-07,
+      "loss": 0.809,
+      "step": 50828
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.4878285714285715e-07,
+      "loss": 0.9444,
+      "step": 50829
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.485142857142857e-07,
+      "loss": 1.2312,
+      "step": 50830
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.482457142857143e-07,
+      "loss": 1.3822,
+      "step": 50831
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.4797714285714284e-07,
+      "loss": 1.089,
+      "step": 50832
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.477085714285715e-07,
+      "loss": 0.6258,
+      "step": 50833
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.4744000000000004e-07,
+      "loss": 1.303,
+      "step": 50834
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.471714285714286e-07,
+      "loss": 1.0618,
+      "step": 50835
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.469028571428572e-07,
+      "loss": 1.2121,
+      "step": 50836
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.4663428571428573e-07,
+      "loss": 0.9867,
+      "step": 50837
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.463657142857143e-07,
+      "loss": 0.8957,
+      "step": 50838
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4609714285714283e-07,
+      "loss": 0.6911,
+      "step": 50839
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4582857142857143e-07,
+      "loss": 0.8211,
+      "step": 50840
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.455600000000001e-07,
+      "loss": 0.9739,
+      "step": 50841
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4529142857142863e-07,
+      "loss": 0.9759,
+      "step": 50842
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.450228571428572e-07,
+      "loss": 1.0938,
+      "step": 50843
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.447542857142857e-07,
+      "loss": 1.3042,
+      "step": 50844
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.444857142857143e-07,
+      "loss": 0.8535,
+      "step": 50845
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4421714285714287e-07,
+      "loss": 1.0112,
+      "step": 50846
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.439485714285714e-07,
+      "loss": 0.9983,
+      "step": 50847
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4367999999999996e-07,
+      "loss": 1.1655,
+      "step": 50848
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.434114285714286e-07,
+      "loss": 1.3986,
+      "step": 50849
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.431428571428572e-07,
+      "loss": 1.2978,
+      "step": 50850
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4287428571428576e-07,
+      "loss": 1.2892,
+      "step": 50851
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.426057142857143e-07,
+      "loss": 1.188,
+      "step": 50852
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4233714285714285e-07,
+      "loss": 0.8882,
+      "step": 50853
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4206857142857145e-07,
+      "loss": 1.4027,
+      "step": 50854
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.418e-07,
+      "loss": 0.935,
+      "step": 50855
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4153142857142855e-07,
+      "loss": 1.0983,
+      "step": 50856
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.412628571428572e-07,
+      "loss": 0.9703,
+      "step": 50857
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4099428571428574e-07,
+      "loss": 1.1093,
+      "step": 50858
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4072571428571434e-07,
+      "loss": 1.2148,
+      "step": 50859
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.404571428571429e-07,
+      "loss": 1.5291,
+      "step": 50860
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4018857142857144e-07,
+      "loss": 0.9504,
+      "step": 50861
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3992e-07,
+      "loss": 0.8682,
+      "step": 50862
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.396514285714286e-07,
+      "loss": 1.6191,
+      "step": 50863
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.393828571428572e-07,
+      "loss": 1.2973,
+      "step": 50864
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3911428571428573e-07,
+      "loss": 0.9324,
+      "step": 50865
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.388457142857143e-07,
+      "loss": 1.2109,
+      "step": 50866
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.385771428571429e-07,
+      "loss": 1.1119,
+      "step": 50867
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.383085714285714e-07,
+      "loss": 1.1757,
+      "step": 50868
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3804e-07,
+      "loss": 0.6149,
+      "step": 50869
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3777142857142857e-07,
+      "loss": 0.9556,
+      "step": 50870
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3750285714285717e-07,
+      "loss": 1.3341,
+      "step": 50871
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.372342857142857e-07,
+      "loss": 1.2436,
+      "step": 50872
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.369657142857143e-07,
+      "loss": 1.3804,
+      "step": 50873
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3669714285714286e-07,
+      "loss": 0.9523,
+      "step": 50874
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3642857142857146e-07,
+      "loss": 1.1994,
+      "step": 50875
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3616e-07,
+      "loss": 1.2879,
+      "step": 50876
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3589142857142856e-07,
+      "loss": 1.1406,
+      "step": 50877
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3562285714285716e-07,
+      "loss": 1.1067,
+      "step": 50878
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3535428571428576e-07,
+      "loss": 1.2268,
+      "step": 50879
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.350857142857143e-07,
+      "loss": 1.0997,
+      "step": 50880
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3481714285714285e-07,
+      "loss": 0.879,
+      "step": 50881
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3454857142857145e-07,
+      "loss": 1.442,
+      "step": 50882
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3428000000000005e-07,
+      "loss": 1.0269,
+      "step": 50883
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.340114285714286e-07,
+      "loss": 0.9109,
+      "step": 50884
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3374285714285714e-07,
+      "loss": 1.3688,
+      "step": 50885
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.334742857142857e-07,
+      "loss": 1.1054,
+      "step": 50886
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3320571428571434e-07,
+      "loss": 0.8736,
+      "step": 50887
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.329371428571429e-07,
+      "loss": 1.1379,
+      "step": 50888
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3266857142857143e-07,
+      "loss": 1.245,
+      "step": 50889
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.324e-07,
+      "loss": 1.5707,
+      "step": 50890
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.321314285714286e-07,
+      "loss": 1.0909,
+      "step": 50891
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.318628571428572e-07,
+      "loss": 1.5383,
+      "step": 50892
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3159428571428573e-07,
+      "loss": 1.3287,
+      "step": 50893
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.313257142857143e-07,
+      "loss": 0.9625,
+      "step": 50894
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3105714285714287e-07,
+      "loss": 1.4048,
+      "step": 50895
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.307885714285714e-07,
+      "loss": 1.3181,
+      "step": 50896
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3052e-07,
+      "loss": 0.9955,
+      "step": 50897
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.302514285714286e-07,
+      "loss": 1.5124,
+      "step": 50898
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2998285714285717e-07,
+      "loss": 1.2693,
+      "step": 50899
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.297142857142857e-07,
+      "loss": 1.3847,
+      "step": 50900
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.294457142857143e-07,
+      "loss": 1.2516,
+      "step": 50901
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.291771428571429e-07,
+      "loss": 0.9386,
+      "step": 50902
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2890857142857146e-07,
+      "loss": 1.0373,
+      "step": 50903
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2864e-07,
+      "loss": 1.3659,
+      "step": 50904
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2837142857142855e-07,
+      "loss": 1.0368,
+      "step": 50905
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.281028571428572e-07,
+      "loss": 1.1624,
+      "step": 50906
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2783428571428575e-07,
+      "loss": 0.9056,
+      "step": 50907
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.275657142857143e-07,
+      "loss": 1.1377,
+      "step": 50908
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2729714285714284e-07,
+      "loss": 1.1235,
+      "step": 50909
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2702857142857144e-07,
+      "loss": 0.8862,
+      "step": 50910
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2676000000000004e-07,
+      "loss": 0.9744,
+      "step": 50911
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.264914285714286e-07,
+      "loss": 1.2047,
+      "step": 50912
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2622285714285714e-07,
+      "loss": 1.2813,
+      "step": 50913
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2595428571428574e-07,
+      "loss": 0.6737,
+      "step": 50914
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2568571428571434e-07,
+      "loss": 1.0441,
+      "step": 50915
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.254171428571429e-07,
+      "loss": 1.3336,
+      "step": 50916
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2514857142857143e-07,
+      "loss": 1.1269,
+      "step": 50917
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2488000000000003e-07,
+      "loss": 0.8991,
+      "step": 50918
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.246114285714286e-07,
+      "loss": 1.0835,
+      "step": 50919
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.243428571428572e-07,
+      "loss": 1.1692,
+      "step": 50920
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.240742857142857e-07,
+      "loss": 1.2883,
+      "step": 50921
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.238057142857143e-07,
+      "loss": 1.1738,
+      "step": 50922
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2353714285714287e-07,
+      "loss": 1.0769,
+      "step": 50923
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.232685714285714e-07,
+      "loss": 0.8273,
+      "step": 50924
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.23e-07,
+      "loss": 1.482,
+      "step": 50925
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.227314285714286e-07,
+      "loss": 1.1917,
+      "step": 50926
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2246285714285716e-07,
+      "loss": 1.3743,
+      "step": 50927
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.221942857142857e-07,
+      "loss": 1.1891,
+      "step": 50928
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.219257142857143e-07,
+      "loss": 0.7769,
+      "step": 50929
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.216571428571429e-07,
+      "loss": 1.236,
+      "step": 50930
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2138857142857145e-07,
+      "loss": 1.0567,
+      "step": 50931
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2112e-07,
+      "loss": 1.1082,
+      "step": 50932
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2085142857142855e-07,
+      "loss": 1.1312,
+      "step": 50933
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.205828571428572e-07,
+      "loss": 0.9591,
+      "step": 50934
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2031428571428575e-07,
+      "loss": 1.1577,
+      "step": 50935
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.200457142857143e-07,
+      "loss": 1.1612,
+      "step": 50936
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1977714285714284e-07,
+      "loss": 0.839,
+      "step": 50937
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1950857142857144e-07,
+      "loss": 1.2691,
+      "step": 50938
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1924000000000004e-07,
+      "loss": 1.4641,
+      "step": 50939
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.189714285714286e-07,
+      "loss": 1.148,
+      "step": 50940
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1870285714285713e-07,
+      "loss": 0.921,
+      "step": 50941
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1843428571428573e-07,
+      "loss": 1.407,
+      "step": 50942
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1816571428571433e-07,
+      "loss": 1.0771,
+      "step": 50943
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.178971428571429e-07,
+      "loss": 1.2436,
+      "step": 50944
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1762857142857143e-07,
+      "loss": 1.3084,
+      "step": 50945
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1736e-07,
+      "loss": 0.9398,
+      "step": 50946
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1709142857142857e-07,
+      "loss": 1.0454,
+      "step": 50947
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1682285714285717e-07,
+      "loss": 1.2363,
+      "step": 50948
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.165542857142857e-07,
+      "loss": 0.9256,
+      "step": 50949
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.162857142857143e-07,
+      "loss": 0.8685,
+      "step": 50950
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1601714285714287e-07,
+      "loss": 1.143,
+      "step": 50951
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.157485714285714e-07,
+      "loss": 1.0043,
+      "step": 50952
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1548e-07,
+      "loss": 0.9727,
+      "step": 50953
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.152114285714286e-07,
+      "loss": 1.0492,
+      "step": 50954
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1494285714285716e-07,
+      "loss": 0.7116,
+      "step": 50955
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.146742857142857e-07,
+      "loss": 0.8938,
+      "step": 50956
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.144057142857143e-07,
+      "loss": 0.6572,
+      "step": 50957
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.141371428571429e-07,
+      "loss": 1.3422,
+      "step": 50958
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1386857142857145e-07,
+      "loss": 0.8059,
+      "step": 50959
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.136e-07,
+      "loss": 1.4924,
+      "step": 50960
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1333142857142854e-07,
+      "loss": 1.1076,
+      "step": 50961
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.130628571428572e-07,
+      "loss": 1.2428,
+      "step": 50962
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1279428571428574e-07,
+      "loss": 0.9799,
+      "step": 50963
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.125257142857143e-07,
+      "loss": 1.0986,
+      "step": 50964
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1225714285714284e-07,
+      "loss": 1.2567,
+      "step": 50965
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1198857142857144e-07,
+      "loss": 1.2017,
+      "step": 50966
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1172000000000004e-07,
+      "loss": 1.1693,
+      "step": 50967
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.114514285714286e-07,
+      "loss": 1.2898,
+      "step": 50968
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1118285714285713e-07,
+      "loss": 0.6531,
+      "step": 50969
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1091428571428573e-07,
+      "loss": 1.4488,
+      "step": 50970
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1064571428571433e-07,
+      "loss": 1.2674,
+      "step": 50971
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.103771428571429e-07,
+      "loss": 1.1357,
+      "step": 50972
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.101085714285714e-07,
+      "loss": 1.1006,
+      "step": 50973
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0984e-07,
+      "loss": 1.1359,
+      "step": 50974
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0957142857142857e-07,
+      "loss": 1.1039,
+      "step": 50975
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0930285714285717e-07,
+      "loss": 1.2039,
+      "step": 50976
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.090342857142857e-07,
+      "loss": 1.2046,
+      "step": 50977
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.087657142857143e-07,
+      "loss": 0.7654,
+      "step": 50978
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0849714285714286e-07,
+      "loss": 1.2304,
+      "step": 50979
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.082285714285714e-07,
+      "loss": 1.5202,
+      "step": 50980
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0796e-07,
+      "loss": 0.9634,
+      "step": 50981
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.076914285714286e-07,
+      "loss": 1.1076,
+      "step": 50982
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0742285714285715e-07,
+      "loss": 0.9806,
+      "step": 50983
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.071542857142857e-07,
+      "loss": 0.903,
+      "step": 50984
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.068857142857143e-07,
+      "loss": 1.1114,
+      "step": 50985
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.066171428571429e-07,
+      "loss": 1.4774,
+      "step": 50986
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0634857142857145e-07,
+      "loss": 1.1484,
+      "step": 50987
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0608e-07,
+      "loss": 1.1768,
+      "step": 50988
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0581142857142854e-07,
+      "loss": 0.9474,
+      "step": 50989
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.055428571428572e-07,
+      "loss": 1.1136,
+      "step": 50990
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0527428571428574e-07,
+      "loss": 0.9436,
+      "step": 50991
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.050057142857143e-07,
+      "loss": 0.8042,
+      "step": 50992
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0473714285714283e-07,
+      "loss": 1.0501,
+      "step": 50993
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0446857142857143e-07,
+      "loss": 0.9285,
+      "step": 50994
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0420000000000003e-07,
+      "loss": 0.8965,
+      "step": 50995
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.039314285714286e-07,
+      "loss": 0.8649,
+      "step": 50996
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.036628571428571e-07,
+      "loss": 1.1679,
+      "step": 50997
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.033942857142857e-07,
+      "loss": 0.902,
+      "step": 50998
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.031257142857143e-07,
+      "loss": 0.9869,
+      "step": 50999
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0285714285714287e-07,
+      "loss": 0.7282,
+      "step": 51000
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0258857142857147e-07,
+      "loss": 1.1768,
+      "step": 51001
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0232e-07,
+      "loss": 1.3168,
+      "step": 51002
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0205142857142857e-07,
+      "loss": 0.847,
+      "step": 51003
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0178285714285716e-07,
+      "loss": 0.9909,
+      "step": 51004
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0151428571428576e-07,
+      "loss": 1.0422,
+      "step": 51005
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.012457142857143e-07,
+      "loss": 0.773,
+      "step": 51006
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0097714285714286e-07,
+      "loss": 0.9956,
+      "step": 51007
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.007085714285714e-07,
+      "loss": 1.3293,
+      "step": 51008
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0044000000000006e-07,
+      "loss": 1.0427,
+      "step": 51009
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.001714285714286e-07,
+      "loss": 1.2725,
+      "step": 51010
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.9990285714285715e-07,
+      "loss": 0.8121,
+      "step": 51011
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.996342857142857e-07,
+      "loss": 1.058,
+      "step": 51012
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.993657142857143e-07,
+      "loss": 1.2074,
+      "step": 51013
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.990971428571429e-07,
+      "loss": 1.1262,
+      "step": 51014
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9882857142857144e-07,
+      "loss": 1.4798,
+      "step": 51015
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9856e-07,
+      "loss": 0.6683,
+      "step": 51016
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.982914285714286e-07,
+      "loss": 0.9626,
+      "step": 51017
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.980228571428572e-07,
+      "loss": 1.3135,
+      "step": 51018
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9775428571428574e-07,
+      "loss": 1.2259,
+      "step": 51019
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.974857142857143e-07,
+      "loss": 1.373,
+      "step": 51020
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.972171428571429e-07,
+      "loss": 1.0899,
+      "step": 51021
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9694857142857143e-07,
+      "loss": 1.0559,
+      "step": 51022
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9668000000000003e-07,
+      "loss": 1.6342,
+      "step": 51023
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.964114285714286e-07,
+      "loss": 0.5265,
+      "step": 51024
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.961428571428572e-07,
+      "loss": 1.1123,
+      "step": 51025
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.958742857142857e-07,
+      "loss": 1.3347,
+      "step": 51026
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.956057142857143e-07,
+      "loss": 1.0573,
+      "step": 51027
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9533714285714287e-07,
+      "loss": 1.1035,
+      "step": 51028
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9506857142857147e-07,
+      "loss": 0.765,
+      "step": 51029
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.948e-07,
+      "loss": 1.0527,
+      "step": 51030
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9453142857142856e-07,
+      "loss": 0.7014,
+      "step": 51031
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9426285714285716e-07,
+      "loss": 1.337,
+      "step": 51032
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9399428571428576e-07,
+      "loss": 1.2093,
+      "step": 51033
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.937257142857143e-07,
+      "loss": 1.4059,
+      "step": 51034
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9345714285714285e-07,
+      "loss": 0.7812,
+      "step": 51035
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.931885714285714e-07,
+      "loss": 1.1036,
+      "step": 51036
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9292000000000005e-07,
+      "loss": 1.1143,
+      "step": 51037
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.926514285714286e-07,
+      "loss": 0.9674,
+      "step": 51038
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9238285714285715e-07,
+      "loss": 1.0771,
+      "step": 51039
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.921142857142857e-07,
+      "loss": 1.7037,
+      "step": 51040
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.918457142857143e-07,
+      "loss": 1.0713,
+      "step": 51041
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.915771428571429e-07,
+      "loss": 1.4446,
+      "step": 51042
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9130857142857144e-07,
+      "loss": 1.1072,
+      "step": 51043
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9104e-07,
+      "loss": 1.2213,
+      "step": 51044
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.907714285714286e-07,
+      "loss": 0.998,
+      "step": 51045
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.905028571428572e-07,
+      "loss": 1.272,
+      "step": 51046
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9023428571428573e-07,
+      "loss": 1.0177,
+      "step": 51047
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.899657142857143e-07,
+      "loss": 1.1981,
+      "step": 51048
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.896971428571429e-07,
+      "loss": 1.1598,
+      "step": 51049
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.894285714285714e-07,
+      "loss": 1.0539,
+      "step": 51050
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8916e-07,
+      "loss": 1.0002,
+      "step": 51051
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8889142857142857e-07,
+      "loss": 1.3755,
+      "step": 51052
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8862285714285717e-07,
+      "loss": 1.518,
+      "step": 51053
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.883542857142857e-07,
+      "loss": 1.3636,
+      "step": 51054
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.880857142857143e-07,
+      "loss": 0.7408,
+      "step": 51055
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8781714285714286e-07,
+      "loss": 0.8148,
+      "step": 51056
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8754857142857146e-07,
+      "loss": 1.3914,
+      "step": 51057
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8728e-07,
+      "loss": 0.9349,
+      "step": 51058
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8701142857142856e-07,
+      "loss": 1.3162,
+      "step": 51059
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8674285714285716e-07,
+      "loss": 1.1877,
+      "step": 51060
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8647428571428576e-07,
+      "loss": 1.148,
+      "step": 51061
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.862057142857143e-07,
+      "loss": 0.9428,
+      "step": 51062
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8593714285714285e-07,
+      "loss": 0.7658,
+      "step": 51063
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.856685714285714e-07,
+      "loss": 1.3769,
+      "step": 51064
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8540000000000005e-07,
+      "loss": 1.2748,
+      "step": 51065
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.851314285714286e-07,
+      "loss": 1.1357,
+      "step": 51066
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8486285714285714e-07,
+      "loss": 1.23,
+      "step": 51067
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.845942857142857e-07,
+      "loss": 1.0083,
+      "step": 51068
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.843257142857143e-07,
+      "loss": 1.3197,
+      "step": 51069
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.840571428571429e-07,
+      "loss": 0.9644,
+      "step": 51070
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8378857142857144e-07,
+      "loss": 0.9006,
+      "step": 51071
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8352e-07,
+      "loss": 1.4278,
+      "step": 51072
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.832514285714286e-07,
+      "loss": 0.8657,
+      "step": 51073
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.829828571428572e-07,
+      "loss": 1.6148,
+      "step": 51074
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8271428571428573e-07,
+      "loss": 0.8828,
+      "step": 51075
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.824457142857143e-07,
+      "loss": 1.2575,
+      "step": 51076
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.821771428571429e-07,
+      "loss": 1.187,
+      "step": 51077
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.819085714285714e-07,
+      "loss": 1.3317,
+      "step": 51078
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8164e-07,
+      "loss": 0.945,
+      "step": 51079
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8137142857142857e-07,
+      "loss": 0.9823,
+      "step": 51080
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8110285714285717e-07,
+      "loss": 1.3118,
+      "step": 51081
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.808342857142857e-07,
+      "loss": 1.568,
+      "step": 51082
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.805657142857143e-07,
+      "loss": 0.9631,
+      "step": 51083
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8029714285714286e-07,
+      "loss": 1.4072,
+      "step": 51084
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8002857142857146e-07,
+      "loss": 0.8721,
+      "step": 51085
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7976e-07,
+      "loss": 1.058,
+      "step": 51086
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7949142857142855e-07,
+      "loss": 1.0615,
+      "step": 51087
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7922285714285715e-07,
+      "loss": 0.9792,
+      "step": 51088
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7895428571428575e-07,
+      "loss": 1.327,
+      "step": 51089
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.786857142857143e-07,
+      "loss": 1.3412,
+      "step": 51090
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7841714285714285e-07,
+      "loss": 1.1989,
+      "step": 51091
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.781485714285714e-07,
+      "loss": 1.0543,
+      "step": 51092
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7788000000000005e-07,
+      "loss": 0.8804,
+      "step": 51093
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.776114285714286e-07,
+      "loss": 1.3973,
+      "step": 51094
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7734285714285714e-07,
+      "loss": 1.3161,
+      "step": 51095
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.770742857142857e-07,
+      "loss": 0.9342,
+      "step": 51096
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.768057142857143e-07,
+      "loss": 1.7021,
+      "step": 51097
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.765371428571429e-07,
+      "loss": 0.5261,
+      "step": 51098
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7626857142857143e-07,
+      "loss": 0.8983,
+      "step": 51099
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7600000000000003e-07,
+      "loss": 0.8228,
+      "step": 51100
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.757314285714286e-07,
+      "loss": 1.7809,
+      "step": 51101
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.754628571428572e-07,
+      "loss": 1.2842,
+      "step": 51102
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.751942857142857e-07,
+      "loss": 0.8378,
+      "step": 51103
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.749257142857143e-07,
+      "loss": 0.9453,
+      "step": 51104
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7465714285714287e-07,
+      "loss": 1.5483,
+      "step": 51105
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.743885714285714e-07,
+      "loss": 0.7221,
+      "step": 51106
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7412e-07,
+      "loss": 1.1186,
+      "step": 51107
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.738514285714286e-07,
+      "loss": 0.9029,
+      "step": 51108
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7358285714285716e-07,
+      "loss": 1.0175,
+      "step": 51109
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.733142857142857e-07,
+      "loss": 1.0467,
+      "step": 51110
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.730457142857143e-07,
+      "loss": 1.052,
+      "step": 51111
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.727771428571429e-07,
+      "loss": 1.21,
+      "step": 51112
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7250857142857146e-07,
+      "loss": 0.8881,
+      "step": 51113
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7224e-07,
+      "loss": 1.3267,
+      "step": 51114
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7197142857142855e-07,
+      "loss": 0.8886,
+      "step": 51115
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.717028571428572e-07,
+      "loss": 0.9801,
+      "step": 51116
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7143428571428575e-07,
+      "loss": 1.4698,
+      "step": 51117
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.711657142857143e-07,
+      "loss": 1.1204,
+      "step": 51118
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7089714285714284e-07,
+      "loss": 1.2227,
+      "step": 51119
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7062857142857144e-07,
+      "loss": 1.6036,
+      "step": 51120
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7036000000000004e-07,
+      "loss": 1.1269,
+      "step": 51121
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.700914285714286e-07,
+      "loss": 1.2764,
+      "step": 51122
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6982285714285714e-07,
+      "loss": 1.1188,
+      "step": 51123
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6955428571428574e-07,
+      "loss": 1.0149,
+      "step": 51124
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6928571428571433e-07,
+      "loss": 1.2106,
+      "step": 51125
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.690171428571429e-07,
+      "loss": 1.29,
+      "step": 51126
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6874857142857143e-07,
+      "loss": 1.4559,
+      "step": 51127
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6848000000000003e-07,
+      "loss": 1.1094,
+      "step": 51128
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.682114285714286e-07,
+      "loss": 0.9276,
+      "step": 51129
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.679428571428572e-07,
+      "loss": 1.1654,
+      "step": 51130
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.676742857142857e-07,
+      "loss": 1.0061,
+      "step": 51131
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.674057142857143e-07,
+      "loss": 1.5274,
+      "step": 51132
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6713714285714287e-07,
+      "loss": 1.2483,
+      "step": 51133
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.668685714285714e-07,
+      "loss": 0.9867,
+      "step": 51134
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.666e-07,
+      "loss": 0.963,
+      "step": 51135
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.663314285714286e-07,
+      "loss": 1.0294,
+      "step": 51136
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6606285714285716e-07,
+      "loss": 1.0272,
+      "step": 51137
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.657942857142857e-07,
+      "loss": 1.2742,
+      "step": 51138
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.655257142857143e-07,
+      "loss": 1.18,
+      "step": 51139
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.652571428571429e-07,
+      "loss": 1.2328,
+      "step": 51140
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6498857142857145e-07,
+      "loss": 1.1329,
+      "step": 51141
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6472e-07,
+      "loss": 1.1008,
+      "step": 51142
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6445142857142855e-07,
+      "loss": 1.513,
+      "step": 51143
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.641828571428572e-07,
+      "loss": 1.0151,
+      "step": 51144
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6391428571428575e-07,
+      "loss": 1.508,
+      "step": 51145
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.636457142857143e-07,
+      "loss": 1.0202,
+      "step": 51146
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6337714285714284e-07,
+      "loss": 0.983,
+      "step": 51147
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6310857142857144e-07,
+      "loss": 1.1017,
+      "step": 51148
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6284000000000004e-07,
+      "loss": 1.1059,
+      "step": 51149
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.625714285714286e-07,
+      "loss": 0.8607,
+      "step": 51150
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6230285714285713e-07,
+      "loss": 1.3288,
+      "step": 51151
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6203428571428573e-07,
+      "loss": 0.7508,
+      "step": 51152
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6176571428571433e-07,
+      "loss": 0.8937,
+      "step": 51153
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.614971428571429e-07,
+      "loss": 1.2236,
+      "step": 51154
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.612285714285714e-07,
+      "loss": 1.298,
+      "step": 51155
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6096e-07,
+      "loss": 1.15,
+      "step": 51156
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6069142857142857e-07,
+      "loss": 1.3938,
+      "step": 51157
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6042285714285717e-07,
+      "loss": 0.8211,
+      "step": 51158
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.601542857142857e-07,
+      "loss": 1.0466,
+      "step": 51159
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.598857142857143e-07,
+      "loss": 1.192,
+      "step": 51160
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5961714285714286e-07,
+      "loss": 0.9575,
+      "step": 51161
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.593485714285714e-07,
+      "loss": 1.412,
+      "step": 51162
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5908e-07,
+      "loss": 1.3167,
+      "step": 51163
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.588114285714286e-07,
+      "loss": 1.6535,
+      "step": 51164
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5854285714285716e-07,
+      "loss": 0.9664,
+      "step": 51165
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.582742857142857e-07,
+      "loss": 1.2812,
+      "step": 51166
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.580057142857143e-07,
+      "loss": 1.0155,
+      "step": 51167
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.577371428571429e-07,
+      "loss": 0.882,
+      "step": 51168
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5746857142857145e-07,
+      "loss": 1.1575,
+      "step": 51169
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.572e-07,
+      "loss": 0.9849,
+      "step": 51170
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5693142857142854e-07,
+      "loss": 1.071,
+      "step": 51171
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.566628571428572e-07,
+      "loss": 0.8093,
+      "step": 51172
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5639428571428574e-07,
+      "loss": 1.4417,
+      "step": 51173
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.561257142857143e-07,
+      "loss": 1.3833,
+      "step": 51174
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5585714285714284e-07,
+      "loss": 1.4286,
+      "step": 51175
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5558857142857143e-07,
+      "loss": 0.9694,
+      "step": 51176
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5532000000000003e-07,
+      "loss": 1.3861,
+      "step": 51177
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.550514285714286e-07,
+      "loss": 0.7949,
+      "step": 51178
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5478285714285713e-07,
+      "loss": 1.2447,
+      "step": 51179
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5451428571428573e-07,
+      "loss": 0.9293,
+      "step": 51180
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5424571428571433e-07,
+      "loss": 1.1384,
+      "step": 51181
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.539771428571429e-07,
+      "loss": 1.0314,
+      "step": 51182
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.537085714285714e-07,
+      "loss": 1.0038,
+      "step": 51183
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5344e-07,
+      "loss": 0.994,
+      "step": 51184
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5317142857142857e-07,
+      "loss": 1.2404,
+      "step": 51185
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5290285714285717e-07,
+      "loss": 0.8244,
+      "step": 51186
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.526342857142857e-07,
+      "loss": 0.9529,
+      "step": 51187
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.523657142857143e-07,
+      "loss": 1.6464,
+      "step": 51188
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.5209714285714286e-07,
+      "loss": 1.0377,
+      "step": 51189
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.518285714285714e-07,
+      "loss": 0.8473,
+      "step": 51190
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.5156e-07,
+      "loss": 1.1036,
+      "step": 51191
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.512914285714286e-07,
+      "loss": 1.3967,
+      "step": 51192
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.5102285714285715e-07,
+      "loss": 0.8697,
+      "step": 51193
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.507542857142857e-07,
+      "loss": 1.2208,
+      "step": 51194
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.504857142857143e-07,
+      "loss": 1.4493,
+      "step": 51195
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.502171428571429e-07,
+      "loss": 0.724,
+      "step": 51196
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4994857142857145e-07,
+      "loss": 1.4976,
+      "step": 51197
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4968e-07,
+      "loss": 1.2225,
+      "step": 51198
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.494114285714286e-07,
+      "loss": 1.3057,
+      "step": 51199
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.491428571428572e-07,
+      "loss": 1.5886,
+      "step": 51200
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4887428571428574e-07,
+      "loss": 1.0385,
+      "step": 51201
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.486057142857143e-07,
+      "loss": 1.3829,
+      "step": 51202
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.483371428571429e-07,
+      "loss": 1.2697,
+      "step": 51203
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4806857142857143e-07,
+      "loss": 0.7832,
+      "step": 51204
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4780000000000003e-07,
+      "loss": 1.6276,
+      "step": 51205
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.475314285714286e-07,
+      "loss": 1.2639,
+      "step": 51206
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.472628571428572e-07,
+      "loss": 0.7421,
+      "step": 51207
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.469942857142857e-07,
+      "loss": 0.8965,
+      "step": 51208
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.467257142857143e-07,
+      "loss": 1.5475,
+      "step": 51209
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4645714285714287e-07,
+      "loss": 0.9292,
+      "step": 51210
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4618857142857147e-07,
+      "loss": 0.9717,
+      "step": 51211
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4592e-07,
+      "loss": 1.1732,
+      "step": 51212
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4565142857142856e-07,
+      "loss": 1.2753,
+      "step": 51213
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4538285714285716e-07,
+      "loss": 1.3491,
+      "step": 51214
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4511428571428576e-07,
+      "loss": 1.1894,
+      "step": 51215
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.448457142857143e-07,
+      "loss": 0.9455,
+      "step": 51216
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4457714285714286e-07,
+      "loss": 0.8482,
+      "step": 51217
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.443085714285714e-07,
+      "loss": 1.0789,
+      "step": 51218
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4404000000000006e-07,
+      "loss": 1.5247,
+      "step": 51219
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.437714285714286e-07,
+      "loss": 1.0237,
+      "step": 51220
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4350285714285715e-07,
+      "loss": 0.8601,
+      "step": 51221
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.432342857142857e-07,
+      "loss": 0.9648,
+      "step": 51222
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.429657142857143e-07,
+      "loss": 1.1981,
+      "step": 51223
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.426971428571429e-07,
+      "loss": 1.29,
+      "step": 51224
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4242857142857144e-07,
+      "loss": 1.0735,
+      "step": 51225
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4216e-07,
+      "loss": 1.2637,
+      "step": 51226
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.418914285714286e-07,
+      "loss": 0.8562,
+      "step": 51227
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.416228571428572e-07,
+      "loss": 1.1819,
+      "step": 51228
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4135428571428573e-07,
+      "loss": 0.9243,
+      "step": 51229
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.410857142857143e-07,
+      "loss": 1.5241,
+      "step": 51230
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.408171428571429e-07,
+      "loss": 0.98,
+      "step": 51231
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4054857142857143e-07,
+      "loss": 1.3018,
+      "step": 51232
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.4028000000000003e-07,
+      "loss": 1.4315,
+      "step": 51233
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.400114285714286e-07,
+      "loss": 0.996,
+      "step": 51234
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3974285714285717e-07,
+      "loss": 1.124,
+      "step": 51235
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.394742857142857e-07,
+      "loss": 1.6498,
+      "step": 51236
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.392057142857143e-07,
+      "loss": 1.0481,
+      "step": 51237
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3893714285714287e-07,
+      "loss": 0.943,
+      "step": 51238
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3866857142857147e-07,
+      "loss": 0.8874,
+      "step": 51239
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.384e-07,
+      "loss": 1.1989,
+      "step": 51240
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3813142857142856e-07,
+      "loss": 1.1207,
+      "step": 51241
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3786285714285716e-07,
+      "loss": 1.2594,
+      "step": 51242
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3759428571428576e-07,
+      "loss": 1.0294,
+      "step": 51243
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.373257142857143e-07,
+      "loss": 0.7544,
+      "step": 51244
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3705714285714285e-07,
+      "loss": 1.1149,
+      "step": 51245
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.367885714285714e-07,
+      "loss": 1.3606,
+      "step": 51246
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3652000000000005e-07,
+      "loss": 0.9632,
+      "step": 51247
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.362514285714286e-07,
+      "loss": 1.0568,
+      "step": 51248
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3598285714285714e-07,
+      "loss": 1.5216,
+      "step": 51249
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.357142857142857e-07,
+      "loss": 0.9646,
+      "step": 51250
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.354457142857143e-07,
+      "loss": 1.0367,
+      "step": 51251
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.351771428571429e-07,
+      "loss": 0.5948,
+      "step": 51252
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3490857142857144e-07,
+      "loss": 1.1751,
+      "step": 51253
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3464e-07,
+      "loss": 0.8581,
+      "step": 51254
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.343714285714286e-07,
+      "loss": 0.917,
+      "step": 51255
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.341028571428572e-07,
+      "loss": 1.0492,
+      "step": 51256
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3383428571428573e-07,
+      "loss": 1.3386,
+      "step": 51257
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.335657142857143e-07,
+      "loss": 0.747,
+      "step": 51258
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.332971428571429e-07,
+      "loss": 1.1707,
+      "step": 51259
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.330285714285714e-07,
+      "loss": 1.137,
+      "step": 51260
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3276e-07,
+      "loss": 1.1617,
+      "step": 51261
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3249142857142857e-07,
+      "loss": 1.1021,
+      "step": 51262
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3222285714285717e-07,
+      "loss": 1.2313,
+      "step": 51263
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.319542857142857e-07,
+      "loss": 1.3547,
+      "step": 51264
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.316857142857143e-07,
+      "loss": 0.9059,
+      "step": 51265
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3141714285714286e-07,
+      "loss": 1.1658,
+      "step": 51266
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3114857142857146e-07,
+      "loss": 0.854,
+      "step": 51267
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3088e-07,
+      "loss": 1.463,
+      "step": 51268
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3061142857142856e-07,
+      "loss": 0.9332,
+      "step": 51269
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3034285714285716e-07,
+      "loss": 1.052,
+      "step": 51270
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.3007428571428575e-07,
+      "loss": 0.8632,
+      "step": 51271
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.298057142857143e-07,
+      "loss": 1.0609,
+      "step": 51272
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2953714285714285e-07,
+      "loss": 1.5806,
+      "step": 51273
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.292685714285714e-07,
+      "loss": 0.9083,
+      "step": 51274
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2900000000000005e-07,
+      "loss": 1.1919,
+      "step": 51275
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.287314285714286e-07,
+      "loss": 0.9451,
+      "step": 51276
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2846285714285714e-07,
+      "loss": 1.5631,
+      "step": 51277
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.281942857142857e-07,
+      "loss": 1.1127,
+      "step": 51278
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.279257142857143e-07,
+      "loss": 0.6622,
+      "step": 51279
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.276571428571429e-07,
+      "loss": 1.2546,
+      "step": 51280
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2738857142857143e-07,
+      "loss": 1.2263,
+      "step": 51281
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2712e-07,
+      "loss": 1.2656,
+      "step": 51282
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.268514285714286e-07,
+      "loss": 1.2982,
+      "step": 51283
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.265828571428572e-07,
+      "loss": 0.8555,
+      "step": 51284
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2631428571428573e-07,
+      "loss": 0.9678,
+      "step": 51285
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2604571428571427e-07,
+      "loss": 1.3107,
+      "step": 51286
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2577714285714287e-07,
+      "loss": 1.1884,
+      "step": 51287
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.255085714285714e-07,
+      "loss": 1.3501,
+      "step": 51288
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2524e-07,
+      "loss": 1.5994,
+      "step": 51289
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2497142857142857e-07,
+      "loss": 1.0063,
+      "step": 51290
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2470285714285717e-07,
+      "loss": 1.0437,
+      "step": 51291
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.244342857142857e-07,
+      "loss": 1.4686,
+      "step": 51292
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.241657142857143e-07,
+      "loss": 1.174,
+      "step": 51293
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2389714285714286e-07,
+      "loss": 1.2974,
+      "step": 51294
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2362857142857146e-07,
+      "loss": 1.0817,
+      "step": 51295
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2336e-07,
+      "loss": 1.1046,
+      "step": 51296
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2309142857142855e-07,
+      "loss": 1.4321,
+      "step": 51297
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.228228571428572e-07,
+      "loss": 1.4837,
+      "step": 51298
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2255428571428575e-07,
+      "loss": 0.8809,
+      "step": 51299
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.222857142857143e-07,
+      "loss": 0.9568,
+      "step": 51300
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2201714285714284e-07,
+      "loss": 1.0991,
+      "step": 51301
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2174857142857144e-07,
+      "loss": 1.047,
+      "step": 51302
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2148000000000004e-07,
+      "loss": 0.6797,
+      "step": 51303
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.212114285714286e-07,
+      "loss": 1.6293,
+      "step": 51304
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2094285714285714e-07,
+      "loss": 1.2393,
+      "step": 51305
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.2067428571428574e-07,
+      "loss": 1.2689,
+      "step": 51306
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.204057142857143e-07,
+      "loss": 1.1813,
+      "step": 51307
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.201371428571429e-07,
+      "loss": 1.2534,
+      "step": 51308
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1986857142857143e-07,
+      "loss": 0.8476,
+      "step": 51309
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1960000000000003e-07,
+      "loss": 0.6194,
+      "step": 51310
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.193314285714286e-07,
+      "loss": 0.8831,
+      "step": 51311
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.190628571428572e-07,
+      "loss": 1.0937,
+      "step": 51312
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.187942857142857e-07,
+      "loss": 0.7781,
+      "step": 51313
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.185257142857143e-07,
+      "loss": 1.4527,
+      "step": 51314
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1825714285714287e-07,
+      "loss": 0.6956,
+      "step": 51315
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.179885714285714e-07,
+      "loss": 0.9135,
+      "step": 51316
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1772e-07,
+      "loss": 1.0583,
+      "step": 51317
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.174514285714286e-07,
+      "loss": 1.243,
+      "step": 51318
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1718285714285716e-07,
+      "loss": 1.2601,
+      "step": 51319
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.169142857142857e-07,
+      "loss": 0.9648,
+      "step": 51320
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.166457142857143e-07,
+      "loss": 1.0791,
+      "step": 51321
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.163771428571429e-07,
+      "loss": 1.2587,
+      "step": 51322
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1610857142857145e-07,
+      "loss": 1.0933,
+      "step": 51323
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1584e-07,
+      "loss": 1.4182,
+      "step": 51324
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1557142857142855e-07,
+      "loss": 0.8519,
+      "step": 51325
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.153028571428572e-07,
+      "loss": 0.7158,
+      "step": 51326
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1503428571428575e-07,
+      "loss": 1.3558,
+      "step": 51327
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.147657142857143e-07,
+      "loss": 1.2718,
+      "step": 51328
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1449714285714284e-07,
+      "loss": 1.2526,
+      "step": 51329
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1422857142857144e-07,
+      "loss": 1.1354,
+      "step": 51330
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1396000000000004e-07,
+      "loss": 1.3546,
+      "step": 51331
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.136914285714286e-07,
+      "loss": 1.1716,
+      "step": 51332
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1342285714285713e-07,
+      "loss": 1.1293,
+      "step": 51333
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1315428571428573e-07,
+      "loss": 1.5285,
+      "step": 51334
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.128857142857143e-07,
+      "loss": 1.1459,
+      "step": 51335
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.126171428571429e-07,
+      "loss": 1.2056,
+      "step": 51336
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.123485714285714e-07,
+      "loss": 1.0023,
+      "step": 51337
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1208e-07,
+      "loss": 1.5075,
+      "step": 51338
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1181142857142857e-07,
+      "loss": 0.8799,
+      "step": 51339
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1154285714285717e-07,
+      "loss": 0.8773,
+      "step": 51340
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.112742857142857e-07,
+      "loss": 1.1158,
+      "step": 51341
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.110057142857143e-07,
+      "loss": 0.9172,
+      "step": 51342
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1073714285714287e-07,
+      "loss": 0.8258,
+      "step": 51343
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.104685714285714e-07,
+      "loss": 0.7446,
+      "step": 51344
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.102e-07,
+      "loss": 0.9766,
+      "step": 51345
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.099314285714286e-07,
+      "loss": 1.2006,
+      "step": 51346
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0966285714285716e-07,
+      "loss": 1.1812,
+      "step": 51347
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.093942857142857e-07,
+      "loss": 0.9484,
+      "step": 51348
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.091257142857143e-07,
+      "loss": 0.8897,
+      "step": 51349
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.088571428571429e-07,
+      "loss": 1.2539,
+      "step": 51350
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0858857142857145e-07,
+      "loss": 1.0914,
+      "step": 51351
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0832e-07,
+      "loss": 0.7437,
+      "step": 51352
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0805142857142854e-07,
+      "loss": 1.0691,
+      "step": 51353
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.077828571428572e-07,
+      "loss": 1.2099,
+      "step": 51354
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0751428571428574e-07,
+      "loss": 1.1602,
+      "step": 51355
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.072457142857143e-07,
+      "loss": 1.178,
+      "step": 51356
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0697714285714284e-07,
+      "loss": 1.1297,
+      "step": 51357
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0670857142857144e-07,
+      "loss": 0.9512,
+      "step": 51358
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0644000000000004e-07,
+      "loss": 1.0553,
+      "step": 51359
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.061714285714286e-07,
+      "loss": 0.8259,
+      "step": 51360
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0590285714285713e-07,
+      "loss": 1.1692,
+      "step": 51361
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0563428571428573e-07,
+      "loss": 0.9925,
+      "step": 51362
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.053657142857143e-07,
+      "loss": 1.3549,
+      "step": 51363
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.050971428571429e-07,
+      "loss": 1.0154,
+      "step": 51364
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.048285714285714e-07,
+      "loss": 1.0559,
+      "step": 51365
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0456e-07,
+      "loss": 1.3942,
+      "step": 51366
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0429142857142857e-07,
+      "loss": 0.9124,
+      "step": 51367
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0402285714285717e-07,
+      "loss": 1.0738,
+      "step": 51368
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.037542857142857e-07,
+      "loss": 0.7368,
+      "step": 51369
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.034857142857143e-07,
+      "loss": 1.2919,
+      "step": 51370
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0321714285714286e-07,
+      "loss": 1.2461,
+      "step": 51371
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.029485714285714e-07,
+      "loss": 1.2099,
+      "step": 51372
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0268e-07,
+      "loss": 1.142,
+      "step": 51373
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.024114285714286e-07,
+      "loss": 1.2967,
+      "step": 51374
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0214285714285715e-07,
+      "loss": 1.3655,
+      "step": 51375
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.018742857142857e-07,
+      "loss": 1.165,
+      "step": 51376
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.016057142857143e-07,
+      "loss": 1.0219,
+      "step": 51377
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.013371428571429e-07,
+      "loss": 1.0609,
+      "step": 51378
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0106857142857145e-07,
+      "loss": 0.8342,
+      "step": 51379
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.008e-07,
+      "loss": 1.0672,
+      "step": 51380
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.0053142857142854e-07,
+      "loss": 1.3074,
+      "step": 51381
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.002628571428572e-07,
+      "loss": 0.9427,
+      "step": 51382
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9999428571428574e-07,
+      "loss": 1.0775,
+      "step": 51383
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.997257142857143e-07,
+      "loss": 0.617,
+      "step": 51384
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9945714285714283e-07,
+      "loss": 0.8416,
+      "step": 51385
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9918857142857143e-07,
+      "loss": 1.296,
+      "step": 51386
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9892000000000003e-07,
+      "loss": 1.2756,
+      "step": 51387
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.986514285714286e-07,
+      "loss": 1.0073,
+      "step": 51388
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.983828571428571e-07,
+      "loss": 1.6358,
+      "step": 51389
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.981142857142857e-07,
+      "loss": 1.1889,
+      "step": 51390
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9784571428571427e-07,
+      "loss": 1.2487,
+      "step": 51391
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9757714285714287e-07,
+      "loss": 0.9778,
+      "step": 51392
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.973085714285714e-07,
+      "loss": 0.9585,
+      "step": 51393
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9704e-07,
+      "loss": 0.8334,
+      "step": 51394
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9677142857142856e-07,
+      "loss": 1.1432,
+      "step": 51395
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9650285714285716e-07,
+      "loss": 0.9255,
+      "step": 51396
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.962342857142857e-07,
+      "loss": 1.0278,
+      "step": 51397
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.959657142857143e-07,
+      "loss": 1.0713,
+      "step": 51398
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9569714285714286e-07,
+      "loss": 1.1625,
+      "step": 51399
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.954285714285714e-07,
+      "loss": 1.0971,
+      "step": 51400
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9516000000000006e-07,
+      "loss": 0.9383,
+      "step": 51401
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.948914285714286e-07,
+      "loss": 1.2627,
+      "step": 51402
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9462285714285715e-07,
+      "loss": 1.4142,
+      "step": 51403
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.943542857142857e-07,
+      "loss": 1.3532,
+      "step": 51404
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.940857142857143e-07,
+      "loss": 0.688,
+      "step": 51405
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.938171428571429e-07,
+      "loss": 1.0356,
+      "step": 51406
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9354857142857144e-07,
+      "loss": 1.1772,
+      "step": 51407
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9328e-07,
+      "loss": 1.5439,
+      "step": 51408
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.930114285714286e-07,
+      "loss": 1.1078,
+      "step": 51409
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.927428571428572e-07,
+      "loss": 0.6779,
+      "step": 51410
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9247428571428574e-07,
+      "loss": 0.986,
+      "step": 51411
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.922057142857143e-07,
+      "loss": 1.4101,
+      "step": 51412
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.919371428571429e-07,
+      "loss": 1.3298,
+      "step": 51413
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9166857142857143e-07,
+      "loss": 1.2029,
+      "step": 51414
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9140000000000003e-07,
+      "loss": 1.0832,
+      "step": 51415
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.911314285714286e-07,
+      "loss": 1.1879,
+      "step": 51416
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.908628571428572e-07,
+      "loss": 0.9004,
+      "step": 51417
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.905942857142857e-07,
+      "loss": 1.0248,
+      "step": 51418
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9032571428571427e-07,
+      "loss": 1.3093,
+      "step": 51419
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.9005714285714287e-07,
+      "loss": 0.9986,
+      "step": 51420
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8978857142857147e-07,
+      "loss": 0.8374,
+      "step": 51421
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8952e-07,
+      "loss": 1.0092,
+      "step": 51422
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8925142857142856e-07,
+      "loss": 1.4384,
+      "step": 51423
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8898285714285716e-07,
+      "loss": 1.3009,
+      "step": 51424
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8871428571428576e-07,
+      "loss": 1.2569,
+      "step": 51425
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.884457142857143e-07,
+      "loss": 1.2104,
+      "step": 51426
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8817714285714285e-07,
+      "loss": 0.6572,
+      "step": 51427
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.879085714285714e-07,
+      "loss": 1.0626,
+      "step": 51428
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8764000000000005e-07,
+      "loss": 1.2213,
+      "step": 51429
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.873714285714286e-07,
+      "loss": 0.8267,
+      "step": 51430
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8710285714285715e-07,
+      "loss": 1.0248,
+      "step": 51431
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.868342857142857e-07,
+      "loss": 1.0907,
+      "step": 51432
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.865657142857143e-07,
+      "loss": 1.2196,
+      "step": 51433
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.862971428571429e-07,
+      "loss": 1.3178,
+      "step": 51434
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8602857142857144e-07,
+      "loss": 0.7947,
+      "step": 51435
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8576e-07,
+      "loss": 1.3396,
+      "step": 51436
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.854914285714286e-07,
+      "loss": 1.0265,
+      "step": 51437
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.852228571428572e-07,
+      "loss": 1.2273,
+      "step": 51438
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8495428571428573e-07,
+      "loss": 0.9081,
+      "step": 51439
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.846857142857143e-07,
+      "loss": 1.6429,
+      "step": 51440
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.844171428571429e-07,
+      "loss": 0.9084,
+      "step": 51441
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.841485714285714e-07,
+      "loss": 1.576,
+      "step": 51442
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8388e-07,
+      "loss": 0.9774,
+      "step": 51443
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8361142857142857e-07,
+      "loss": 1.2913,
+      "step": 51444
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8334285714285717e-07,
+      "loss": 1.4055,
+      "step": 51445
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.830742857142857e-07,
+      "loss": 0.8661,
+      "step": 51446
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8280571428571426e-07,
+      "loss": 0.8998,
+      "step": 51447
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8253714285714286e-07,
+      "loss": 1.236,
+      "step": 51448
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8226857142857146e-07,
+      "loss": 1.2496,
+      "step": 51449
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.82e-07,
+      "loss": 0.7277,
+      "step": 51450
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8173142857142856e-07,
+      "loss": 0.7964,
+      "step": 51451
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8146285714285716e-07,
+      "loss": 1.3813,
+      "step": 51452
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8119428571428576e-07,
+      "loss": 1.0929,
+      "step": 51453
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.809257142857143e-07,
+      "loss": 0.8642,
+      "step": 51454
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8065714285714285e-07,
+      "loss": 1.2874,
+      "step": 51455
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.803885714285714e-07,
+      "loss": 0.6134,
+      "step": 51456
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8012000000000005e-07,
+      "loss": 0.8792,
+      "step": 51457
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.798514285714286e-07,
+      "loss": 0.8441,
+      "step": 51458
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7958285714285714e-07,
+      "loss": 0.9329,
+      "step": 51459
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.793142857142857e-07,
+      "loss": 1.1049,
+      "step": 51460
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.790457142857143e-07,
+      "loss": 1.3692,
+      "step": 51461
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.787771428571429e-07,
+      "loss": 1.0078,
+      "step": 51462
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7850857142857144e-07,
+      "loss": 1.3374,
+      "step": 51463
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7824e-07,
+      "loss": 1.276,
+      "step": 51464
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.779714285714286e-07,
+      "loss": 1.116,
+      "step": 51465
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.777028571428572e-07,
+      "loss": 0.958,
+      "step": 51466
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7743428571428573e-07,
+      "loss": 1.1082,
+      "step": 51467
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.771657142857143e-07,
+      "loss": 0.5985,
+      "step": 51468
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.768971428571429e-07,
+      "loss": 0.9129,
+      "step": 51469
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.766285714285714e-07,
+      "loss": 1.3751,
+      "step": 51470
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7636e-07,
+      "loss": 1.0292,
+      "step": 51471
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7609142857142857e-07,
+      "loss": 1.5827,
+      "step": 51472
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7582285714285717e-07,
+      "loss": 0.6286,
+      "step": 51473
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.755542857142857e-07,
+      "loss": 0.9736,
+      "step": 51474
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7528571428571426e-07,
+      "loss": 0.887,
+      "step": 51475
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7501714285714286e-07,
+      "loss": 1.0596,
+      "step": 51476
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7474857142857146e-07,
+      "loss": 1.0467,
+      "step": 51477
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7448e-07,
+      "loss": 1.2862,
+      "step": 51478
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7421142857142855e-07,
+      "loss": 0.8466,
+      "step": 51479
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7394285714285715e-07,
+      "loss": 1.1678,
+      "step": 51480
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7367428571428575e-07,
+      "loss": 1.0852,
+      "step": 51481
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.734057142857143e-07,
+      "loss": 1.0691,
+      "step": 51482
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7313714285714285e-07,
+      "loss": 1.2566,
+      "step": 51483
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.728685714285714e-07,
+      "loss": 1.104,
+      "step": 51484
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7260000000000005e-07,
+      "loss": 1.1635,
+      "step": 51485
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.723314285714286e-07,
+      "loss": 1.3019,
+      "step": 51486
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7206285714285714e-07,
+      "loss": 0.5977,
+      "step": 51487
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.717942857142857e-07,
+      "loss": 1.1226,
+      "step": 51488
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.715257142857143e-07,
+      "loss": 0.8281,
+      "step": 51489
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.712571428571429e-07,
+      "loss": 0.6858,
+      "step": 51490
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7098857142857143e-07,
+      "loss": 1.1397,
+      "step": 51491
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.7072e-07,
+      "loss": 1.0948,
+      "step": 51492
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.704514285714286e-07,
+      "loss": 1.025,
+      "step": 51493
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.701828571428572e-07,
+      "loss": 1.2551,
+      "step": 51494
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.699142857142857e-07,
+      "loss": 1.156,
+      "step": 51495
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6964571428571427e-07,
+      "loss": 0.8334,
+      "step": 51496
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6937714285714287e-07,
+      "loss": 1.5353,
+      "step": 51497
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.691085714285714e-07,
+      "loss": 1.2869,
+      "step": 51498
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6884e-07,
+      "loss": 1.2729,
+      "step": 51499
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.685714285714286e-07,
+      "loss": 1.1433,
+      "step": 51500
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6830285714285716e-07,
+      "loss": 1.4648,
+      "step": 51501
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.680342857142857e-07,
+      "loss": 1.0908,
+      "step": 51502
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6776571428571426e-07,
+      "loss": 1.0585,
+      "step": 51503
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.674971428571429e-07,
+      "loss": 1.3006,
+      "step": 51504
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6722857142857146e-07,
+      "loss": 1.3028,
+      "step": 51505
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6696e-07,
+      "loss": 1.3133,
+      "step": 51506
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6669142857142855e-07,
+      "loss": 1.1183,
+      "step": 51507
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.664228571428572e-07,
+      "loss": 0.9124,
+      "step": 51508
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6615428571428575e-07,
+      "loss": 0.7314,
+      "step": 51509
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.658857142857143e-07,
+      "loss": 0.6943,
+      "step": 51510
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6561714285714284e-07,
+      "loss": 1.2094,
+      "step": 51511
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6534857142857144e-07,
+      "loss": 0.7616,
+      "step": 51512
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6508000000000004e-07,
+      "loss": 0.9609,
+      "step": 51513
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.648114285714286e-07,
+      "loss": 0.983,
+      "step": 51514
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6454285714285714e-07,
+      "loss": 1.1135,
+      "step": 51515
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6427428571428573e-07,
+      "loss": 1.6395,
+      "step": 51516
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.640057142857143e-07,
+      "loss": 1.2516,
+      "step": 51517
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.637371428571429e-07,
+      "loss": 1.1947,
+      "step": 51518
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6346857142857143e-07,
+      "loss": 1.228,
+      "step": 51519
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6320000000000003e-07,
+      "loss": 0.9225,
+      "step": 51520
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.629314285714286e-07,
+      "loss": 1.1878,
+      "step": 51521
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.626628571428572e-07,
+      "loss": 1.4517,
+      "step": 51522
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.623942857142857e-07,
+      "loss": 0.8231,
+      "step": 51523
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.621257142857143e-07,
+      "loss": 0.5906,
+      "step": 51524
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6185714285714287e-07,
+      "loss": 1.039,
+      "step": 51525
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.615885714285714e-07,
+      "loss": 1.115,
+      "step": 51526
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6132e-07,
+      "loss": 1.2791,
+      "step": 51527
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.610514285714286e-07,
+      "loss": 1.1748,
+      "step": 51528
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6078285714285716e-07,
+      "loss": 0.6715,
+      "step": 51529
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.605142857142857e-07,
+      "loss": 1.1325,
+      "step": 51530
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6024571428571425e-07,
+      "loss": 1.1609,
+      "step": 51531
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.599771428571429e-07,
+      "loss": 0.9774,
+      "step": 51532
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.5970857142857145e-07,
+      "loss": 0.8993,
+      "step": 51533
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.5944e-07,
+      "loss": 1.084,
+      "step": 51534
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.5917142857142855e-07,
+      "loss": 1.2855,
+      "step": 51535
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.589028571428572e-07,
+      "loss": 0.8818,
+      "step": 51536
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.5863428571428575e-07,
+      "loss": 1.0834,
+      "step": 51537
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.583657142857143e-07,
+      "loss": 1.3384,
+      "step": 51538
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5809714285714284e-07,
+      "loss": 1.2528,
+      "step": 51539
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5782857142857144e-07,
+      "loss": 1.034,
+      "step": 51540
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5756000000000004e-07,
+      "loss": 0.8271,
+      "step": 51541
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.572914285714286e-07,
+      "loss": 0.9693,
+      "step": 51542
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5702285714285713e-07,
+      "loss": 0.7621,
+      "step": 51543
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5675428571428573e-07,
+      "loss": 1.1577,
+      "step": 51544
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.564857142857143e-07,
+      "loss": 1.1026,
+      "step": 51545
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.562171428571429e-07,
+      "loss": 1.528,
+      "step": 51546
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.559485714285714e-07,
+      "loss": 1.5447,
+      "step": 51547
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5568e-07,
+      "loss": 1.4358,
+      "step": 51548
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5541142857142857e-07,
+      "loss": 1.1118,
+      "step": 51549
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5514285714285717e-07,
+      "loss": 1.165,
+      "step": 51550
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.548742857142857e-07,
+      "loss": 1.3009,
+      "step": 51551
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.546057142857143e-07,
+      "loss": 0.8198,
+      "step": 51552
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5433714285714286e-07,
+      "loss": 1.0238,
+      "step": 51553
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.540685714285714e-07,
+      "loss": 1.6448,
+      "step": 51554
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.538e-07,
+      "loss": 1.2145,
+      "step": 51555
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.535314285714286e-07,
+      "loss": 1.1658,
+      "step": 51556
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5326285714285716e-07,
+      "loss": 0.9245,
+      "step": 51557
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.529942857142857e-07,
+      "loss": 0.8566,
+      "step": 51558
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5272571428571425e-07,
+      "loss": 0.9621,
+      "step": 51559
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.524571428571429e-07,
+      "loss": 1.0765,
+      "step": 51560
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5218857142857145e-07,
+      "loss": 1.2303,
+      "step": 51561
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5192e-07,
+      "loss": 0.9364,
+      "step": 51562
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5165142857142854e-07,
+      "loss": 1.2574,
+      "step": 51563
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.513828571428572e-07,
+      "loss": 1.6099,
+      "step": 51564
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5111428571428574e-07,
+      "loss": 0.9423,
+      "step": 51565
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.508457142857143e-07,
+      "loss": 1.1872,
+      "step": 51566
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5057714285714283e-07,
+      "loss": 1.0905,
+      "step": 51567
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5030857142857143e-07,
+      "loss": 1.5954,
+      "step": 51568
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.5004000000000003e-07,
+      "loss": 1.1063,
+      "step": 51569
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.497714285714286e-07,
+      "loss": 0.7767,
+      "step": 51570
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4950285714285713e-07,
+      "loss": 1.2006,
+      "step": 51571
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4923428571428573e-07,
+      "loss": 1.382,
+      "step": 51572
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.489657142857143e-07,
+      "loss": 1.2075,
+      "step": 51573
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4869714285714287e-07,
+      "loss": 1.0424,
+      "step": 51574
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.484285714285714e-07,
+      "loss": 0.9348,
+      "step": 51575
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4816e-07,
+      "loss": 1.1828,
+      "step": 51576
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4789142857142857e-07,
+      "loss": 0.9871,
+      "step": 51577
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4762285714285717e-07,
+      "loss": 1.5652,
+      "step": 51578
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.473542857142857e-07,
+      "loss": 0.8948,
+      "step": 51579
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.470857142857143e-07,
+      "loss": 1.2163,
+      "step": 51580
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4681714285714286e-07,
+      "loss": 1.5534,
+      "step": 51581
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.465485714285714e-07,
+      "loss": 0.8302,
+      "step": 51582
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4628e-07,
+      "loss": 1.1241,
+      "step": 51583
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.460114285714286e-07,
+      "loss": 1.0547,
+      "step": 51584
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4574285714285715e-07,
+      "loss": 1.0935,
+      "step": 51585
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.454742857142857e-07,
+      "loss": 1.0944,
+      "step": 51586
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4520571428571425e-07,
+      "loss": 1.063,
+      "step": 51587
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.449371428571429e-07,
+      "loss": 1.2685,
+      "step": 51588
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4466857142857144e-07,
+      "loss": 1.2096,
+      "step": 51589
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.444e-07,
+      "loss": 1.5497,
+      "step": 51590
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4413142857142854e-07,
+      "loss": 1.3367,
+      "step": 51591
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.438628571428572e-07,
+      "loss": 1.1835,
+      "step": 51592
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4359428571428574e-07,
+      "loss": 1.1435,
+      "step": 51593
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.433257142857143e-07,
+      "loss": 0.9833,
+      "step": 51594
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4305714285714283e-07,
+      "loss": 1.2037,
+      "step": 51595
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4278857142857143e-07,
+      "loss": 0.9693,
+      "step": 51596
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4252000000000003e-07,
+      "loss": 1.159,
+      "step": 51597
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.422514285714286e-07,
+      "loss": 1.3867,
+      "step": 51598
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.419828571428572e-07,
+      "loss": 1.2432,
+      "step": 51599
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.417142857142857e-07,
+      "loss": 0.6948,
+      "step": 51600
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4144571428571427e-07,
+      "loss": 0.9017,
+      "step": 51601
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4117714285714287e-07,
+      "loss": 1.0858,
+      "step": 51602
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4090857142857147e-07,
+      "loss": 1.0858,
+      "step": 51603
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4064e-07,
+      "loss": 1.3557,
+      "step": 51604
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4037142857142856e-07,
+      "loss": 1.1107,
+      "step": 51605
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.4010285714285716e-07,
+      "loss": 0.9902,
+      "step": 51606
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3983428571428576e-07,
+      "loss": 1.02,
+      "step": 51607
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.395657142857143e-07,
+      "loss": 0.6013,
+      "step": 51608
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3929714285714286e-07,
+      "loss": 1.3615,
+      "step": 51609
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.390285714285714e-07,
+      "loss": 1.0596,
+      "step": 51610
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3876000000000005e-07,
+      "loss": 0.786,
+      "step": 51611
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.384914285714286e-07,
+      "loss": 1.1034,
+      "step": 51612
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3822285714285715e-07,
+      "loss": 1.3534,
+      "step": 51613
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3795428571428572e-07,
+      "loss": 0.837,
+      "step": 51614
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3768571428571432e-07,
+      "loss": 1.05,
+      "step": 51615
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3741714285714287e-07,
+      "loss": 0.9901,
+      "step": 51616
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3714857142857144e-07,
+      "loss": 1.0624,
+      "step": 51617
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3688e-07,
+      "loss": 1.0985,
+      "step": 51618
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3661142857142861e-07,
+      "loss": 1.439,
+      "step": 51619
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3634285714285716e-07,
+      "loss": 1.2228,
+      "step": 51620
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.360742857142857e-07,
+      "loss": 1.4126,
+      "step": 51621
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3580571428571428e-07,
+      "loss": 1.0316,
+      "step": 51622
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3553714285714288e-07,
+      "loss": 0.7009,
+      "step": 51623
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3526857142857145e-07,
+      "loss": 1.0881,
+      "step": 51624
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.35e-07,
+      "loss": 0.9579,
+      "step": 51625
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3473142857142857e-07,
+      "loss": 0.9195,
+      "step": 51626
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3446285714285717e-07,
+      "loss": 1.2159,
+      "step": 51627
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3419428571428572e-07,
+      "loss": 0.8745,
+      "step": 51628
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.339257142857143e-07,
+      "loss": 0.7726,
+      "step": 51629
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3365714285714284e-07,
+      "loss": 1.0306,
+      "step": 51630
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3338857142857147e-07,
+      "loss": 1.114,
+      "step": 51631
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3312e-07,
+      "loss": 0.7776,
+      "step": 51632
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3285142857142859e-07,
+      "loss": 1.1468,
+      "step": 51633
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3258285714285713e-07,
+      "loss": 1.2047,
+      "step": 51634
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3231428571428573e-07,
+      "loss": 1.0436,
+      "step": 51635
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.320457142857143e-07,
+      "loss": 1.2924,
+      "step": 51636
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3177714285714285e-07,
+      "loss": 0.6911,
+      "step": 51637
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3150857142857142e-07,
+      "loss": 1.106,
+      "step": 51638
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3124000000000002e-07,
+      "loss": 0.8004,
+      "step": 51639
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.309714285714286e-07,
+      "loss": 1.1745,
+      "step": 51640
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3070285714285714e-07,
+      "loss": 1.3273,
+      "step": 51641
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3043428571428572e-07,
+      "loss": 1.1832,
+      "step": 51642
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.3016571428571432e-07,
+      "loss": 0.7659,
+      "step": 51643
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2989714285714286e-07,
+      "loss": 1.0025,
+      "step": 51644
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2962857142857144e-07,
+      "loss": 0.8688,
+      "step": 51645
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2935999999999998e-07,
+      "loss": 1.3483,
+      "step": 51646
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.290914285714286e-07,
+      "loss": 1.1111,
+      "step": 51647
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2882285714285716e-07,
+      "loss": 1.3506,
+      "step": 51648
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.285542857142857e-07,
+      "loss": 0.969,
+      "step": 51649
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2828571428571428e-07,
+      "loss": 1.5099,
+      "step": 51650
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2801714285714288e-07,
+      "loss": 0.7747,
+      "step": 51651
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2774857142857145e-07,
+      "loss": 1.1344,
+      "step": 51652
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2748e-07,
+      "loss": 1.0924,
+      "step": 51653
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2721142857142857e-07,
+      "loss": 1.203,
+      "step": 51654
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2694285714285717e-07,
+      "loss": 0.9424,
+      "step": 51655
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2667428571428572e-07,
+      "loss": 1.0259,
+      "step": 51656
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.264057142857143e-07,
+      "loss": 1.1207,
+      "step": 51657
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2613714285714284e-07,
+      "loss": 0.9992,
+      "step": 51658
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2586857142857146e-07,
+      "loss": 0.7938,
+      "step": 51659
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.256e-07,
+      "loss": 0.6564,
+      "step": 51660
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2533142857142858e-07,
+      "loss": 0.7875,
+      "step": 51661
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2506285714285713e-07,
+      "loss": 1.1045,
+      "step": 51662
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2479428571428573e-07,
+      "loss": 0.7325,
+      "step": 51663
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.245257142857143e-07,
+      "loss": 0.9814,
+      "step": 51664
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2425714285714285e-07,
+      "loss": 0.7916,
+      "step": 51665
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2398857142857142e-07,
+      "loss": 1.1259,
+      "step": 51666
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2372000000000002e-07,
+      "loss": 1.1905,
+      "step": 51667
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.234514285714286e-07,
+      "loss": 1.0781,
+      "step": 51668
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2318285714285714e-07,
+      "loss": 0.8929,
+      "step": 51669
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2291428571428571e-07,
+      "loss": 1.067,
+      "step": 51670
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2264571428571431e-07,
+      "loss": 1.3887,
+      "step": 51671
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2237714285714286e-07,
+      "loss": 0.6603,
+      "step": 51672
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2210857142857143e-07,
+      "loss": 0.8871,
+      "step": 51673
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2183999999999998e-07,
+      "loss": 1.247,
+      "step": 51674
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.215714285714286e-07,
+      "loss": 1.3728,
+      "step": 51675
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2130285714285715e-07,
+      "loss": 1.1921,
+      "step": 51676
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2103428571428573e-07,
+      "loss": 0.772,
+      "step": 51677
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2076571428571427e-07,
+      "loss": 0.9249,
+      "step": 51678
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2049714285714287e-07,
+      "loss": 1.0297,
+      "step": 51679
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.2022857142857145e-07,
+      "loss": 1.2693,
+      "step": 51680
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1996e-07,
+      "loss": 1.0356,
+      "step": 51681
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.196914285714286e-07,
+      "loss": 1.1026,
+      "step": 51682
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1942285714285714e-07,
+      "loss": 1.1156,
+      "step": 51683
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.191542857142857e-07,
+      "loss": 0.8722,
+      "step": 51684
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1888571428571429e-07,
+      "loss": 1.472,
+      "step": 51685
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1861714285714286e-07,
+      "loss": 1.1694,
+      "step": 51686
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1834857142857143e-07,
+      "loss": 1.1205,
+      "step": 51687
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1808e-07,
+      "loss": 0.8182,
+      "step": 51688
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1781142857142858e-07,
+      "loss": 0.9943,
+      "step": 51689
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1754285714285715e-07,
+      "loss": 1.1676,
+      "step": 51690
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1727428571428572e-07,
+      "loss": 1.0933,
+      "step": 51691
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.170057142857143e-07,
+      "loss": 0.7514,
+      "step": 51692
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1673714285714284e-07,
+      "loss": 0.9477,
+      "step": 51693
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1646857142857144e-07,
+      "loss": 1.1399,
+      "step": 51694
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.162e-07,
+      "loss": 1.2658,
+      "step": 51695
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.159314285714286e-07,
+      "loss": 0.9415,
+      "step": 51696
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1566285714285714e-07,
+      "loss": 1.4179,
+      "step": 51697
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.153942857142857e-07,
+      "loss": 1.2945,
+      "step": 51698
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.151257142857143e-07,
+      "loss": 1.4151,
+      "step": 51699
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1485714285714286e-07,
+      "loss": 0.8445,
+      "step": 51700
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1458857142857146e-07,
+      "loss": 1.319,
+      "step": 51701
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1432e-07,
+      "loss": 0.9529,
+      "step": 51702
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.140514285714286e-07,
+      "loss": 1.0596,
+      "step": 51703
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1378285714285715e-07,
+      "loss": 0.8307,
+      "step": 51704
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1351428571428572e-07,
+      "loss": 1.1503,
+      "step": 51705
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.132457142857143e-07,
+      "loss": 1.061,
+      "step": 51706
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1297714285714287e-07,
+      "loss": 1.2847,
+      "step": 51707
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1270857142857144e-07,
+      "loss": 1.5268,
+      "step": 51708
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1244000000000001e-07,
+      "loss": 0.7426,
+      "step": 51709
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.121714285714286e-07,
+      "loss": 0.8494,
+      "step": 51710
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1190285714285716e-07,
+      "loss": 1.1215,
+      "step": 51711
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116342857142857e-07,
+      "loss": 0.8746,
+      "step": 51712
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.113657142857143e-07,
+      "loss": 1.1082,
+      "step": 51713
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1109714285714285e-07,
+      "loss": 0.7115,
+      "step": 51714
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1082857142857145e-07,
+      "loss": 0.9377,
+      "step": 51715
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1056e-07,
+      "loss": 1.0106,
+      "step": 51716
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.102914285714286e-07,
+      "loss": 1.2369,
+      "step": 51717
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1002285714285715e-07,
+      "loss": 1.0786,
+      "step": 51718
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0975428571428572e-07,
+      "loss": 0.9498,
+      "step": 51719
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.094857142857143e-07,
+      "loss": 1.0205,
+      "step": 51720
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0921714285714287e-07,
+      "loss": 1.0994,
+      "step": 51721
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0894857142857144e-07,
+      "loss": 1.2868,
+      "step": 51722
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0868e-07,
+      "loss": 1.2939,
+      "step": 51723
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0841142857142859e-07,
+      "loss": 1.2325,
+      "step": 51724
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0814285714285716e-07,
+      "loss": 1.1653,
+      "step": 51725
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.078742857142857e-07,
+      "loss": 0.8359,
+      "step": 51726
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.076057142857143e-07,
+      "loss": 1.2767,
+      "step": 51727
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0733714285714285e-07,
+      "loss": 1.3441,
+      "step": 51728
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0706857142857145e-07,
+      "loss": 1.0672,
+      "step": 51729
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.068e-07,
+      "loss": 1.2864,
+      "step": 51730
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.065314285714286e-07,
+      "loss": 1.3634,
+      "step": 51731
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0626285714285715e-07,
+      "loss": 1.1019,
+      "step": 51732
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0599428571428572e-07,
+      "loss": 0.9218,
+      "step": 51733
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.057257142857143e-07,
+      "loss": 1.1317,
+      "step": 51734
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0545714285714286e-07,
+      "loss": 1.3539,
+      "step": 51735
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0518857142857144e-07,
+      "loss": 1.3057,
+      "step": 51736
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0492e-07,
+      "loss": 1.3234,
+      "step": 51737
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0465142857142858e-07,
+      "loss": 1.2467,
+      "step": 51738
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0438285714285716e-07,
+      "loss": 1.1146,
+      "step": 51739
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.041142857142857e-07,
+      "loss": 1.0354,
+      "step": 51740
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.038457142857143e-07,
+      "loss": 1.3399,
+      "step": 51741
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0357714285714285e-07,
+      "loss": 0.8498,
+      "step": 51742
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0330857142857145e-07,
+      "loss": 1.2025,
+      "step": 51743
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0304e-07,
+      "loss": 0.6293,
+      "step": 51744
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.027714285714286e-07,
+      "loss": 1.1007,
+      "step": 51745
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0250285714285714e-07,
+      "loss": 1.1467,
+      "step": 51746
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0223428571428572e-07,
+      "loss": 1.2542,
+      "step": 51747
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.019657142857143e-07,
+      "loss": 0.642,
+      "step": 51748
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0169714285714286e-07,
+      "loss": 1.1064,
+      "step": 51749
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0142857142857144e-07,
+      "loss": 0.8354,
+      "step": 51750
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0116e-07,
+      "loss": 0.6807,
+      "step": 51751
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0089142857142858e-07,
+      "loss": 1.217,
+      "step": 51752
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0062285714285716e-07,
+      "loss": 1.2504,
+      "step": 51753
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.003542857142857e-07,
+      "loss": 0.9829,
+      "step": 51754
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.000857142857143e-07,
+      "loss": 1.2712,
+      "step": 51755
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9981714285714285e-07,
+      "loss": 1.157,
+      "step": 51756
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9954857142857145e-07,
+      "loss": 1.2557,
+      "step": 51757
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9928e-07,
+      "loss": 0.8126,
+      "step": 51758
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.990114285714286e-07,
+      "loss": 1.1628,
+      "step": 51759
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9874285714285714e-07,
+      "loss": 0.7849,
+      "step": 51760
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9847428571428571e-07,
+      "loss": 1.0551,
+      "step": 51761
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.982057142857143e-07,
+      "loss": 1.0149,
+      "step": 51762
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9793714285714286e-07,
+      "loss": 1.3371,
+      "step": 51763
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9766857142857143e-07,
+      "loss": 1.0921,
+      "step": 51764
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.974e-07,
+      "loss": 1.2506,
+      "step": 51765
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9713142857142858e-07,
+      "loss": 0.8781,
+      "step": 51766
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9686285714285715e-07,
+      "loss": 0.8499,
+      "step": 51767
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.965942857142857e-07,
+      "loss": 0.7777,
+      "step": 51768
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.963257142857143e-07,
+      "loss": 1.3537,
+      "step": 51769
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9605714285714285e-07,
+      "loss": 1.3388,
+      "step": 51770
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9578857142857145e-07,
+      "loss": 0.968,
+      "step": 51771
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9552e-07,
+      "loss": 1.1643,
+      "step": 51772
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.952514285714286e-07,
+      "loss": 0.9432,
+      "step": 51773
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9498285714285714e-07,
+      "loss": 1.106,
+      "step": 51774
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.947142857142857e-07,
+      "loss": 1.2236,
+      "step": 51775
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9444571428571429e-07,
+      "loss": 1.4562,
+      "step": 51776
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9417714285714286e-07,
+      "loss": 1.136,
+      "step": 51777
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9390857142857143e-07,
+      "loss": 1.1628,
+      "step": 51778
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9364e-07,
+      "loss": 1.1596,
+      "step": 51779
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9337142857142858e-07,
+      "loss": 1.2789,
+      "step": 51780
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9310285714285715e-07,
+      "loss": 0.9697,
+      "step": 51781
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.928342857142857e-07,
+      "loss": 1.274,
+      "step": 51782
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.925657142857143e-07,
+      "loss": 1.1729,
+      "step": 51783
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9229714285714284e-07,
+      "loss": 1.3887,
+      "step": 51784
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9202857142857144e-07,
+      "loss": 1.4449,
+      "step": 51785
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9176e-07,
+      "loss": 0.5644,
+      "step": 51786
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.914914285714286e-07,
+      "loss": 1.1313,
+      "step": 51787
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9122285714285714e-07,
+      "loss": 1.1499,
+      "step": 51788
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.909542857142857e-07,
+      "loss": 1.1124,
+      "step": 51789
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9068571428571428e-07,
+      "loss": 1.2637,
+      "step": 51790
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9041714285714286e-07,
+      "loss": 1.2097,
+      "step": 51791
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.9014857142857143e-07,
+      "loss": 1.1963,
+      "step": 51792
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8988e-07,
+      "loss": 1.1427,
+      "step": 51793
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8961142857142858e-07,
+      "loss": 1.2184,
+      "step": 51794
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8934285714285715e-07,
+      "loss": 1.1879,
+      "step": 51795
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.890742857142857e-07,
+      "loss": 1.1086,
+      "step": 51796
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.888057142857143e-07,
+      "loss": 1.4289,
+      "step": 51797
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8853714285714284e-07,
+      "loss": 1.3871,
+      "step": 51798
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8826857142857144e-07,
+      "loss": 0.999,
+      "step": 51799
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8800000000000002e-07,
+      "loss": 1.5751,
+      "step": 51800
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.877314285714286e-07,
+      "loss": 1.2921,
+      "step": 51801
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8746285714285716e-07,
+      "loss": 1.6706,
+      "step": 51802
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.871942857142857e-07,
+      "loss": 1.2067,
+      "step": 51803
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.869257142857143e-07,
+      "loss": 0.817,
+      "step": 51804
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8665714285714286e-07,
+      "loss": 0.9306,
+      "step": 51805
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8638857142857145e-07,
+      "loss": 1.3461,
+      "step": 51806
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8612e-07,
+      "loss": 1.0625,
+      "step": 51807
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.858514285714286e-07,
+      "loss": 0.7664,
+      "step": 51808
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8558285714285715e-07,
+      "loss": 0.9186,
+      "step": 51809
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8531428571428572e-07,
+      "loss": 1.4076,
+      "step": 51810
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.850457142857143e-07,
+      "loss": 1.1757,
+      "step": 51811
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8477714285714287e-07,
+      "loss": 1.1927,
+      "step": 51812
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8450857142857144e-07,
+      "loss": 1.4728,
+      "step": 51813
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8424000000000001e-07,
+      "loss": 0.6205,
+      "step": 51814
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.839714285714286e-07,
+      "loss": 1.0126,
+      "step": 51815
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8370285714285716e-07,
+      "loss": 1.2687,
+      "step": 51816
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.834342857142857e-07,
+      "loss": 1.0998,
+      "step": 51817
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.831657142857143e-07,
+      "loss": 1.1389,
+      "step": 51818
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8289714285714285e-07,
+      "loss": 0.9076,
+      "step": 51819
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8262857142857145e-07,
+      "loss": 1.1813,
+      "step": 51820
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8236e-07,
+      "loss": 1.0697,
+      "step": 51821
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.820914285714286e-07,
+      "loss": 1.0023,
+      "step": 51822
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8182285714285715e-07,
+      "loss": 1.6587,
+      "step": 51823
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8155428571428572e-07,
+      "loss": 1.3189,
+      "step": 51824
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.812857142857143e-07,
+      "loss": 0.8844,
+      "step": 51825
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8101714285714287e-07,
+      "loss": 1.1254,
+      "step": 51826
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8074857142857144e-07,
+      "loss": 1.2054,
+      "step": 51827
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8048e-07,
+      "loss": 1.0889,
+      "step": 51828
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.8021142857142859e-07,
+      "loss": 1.161,
+      "step": 51829
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7994285714285716e-07,
+      "loss": 1.4889,
+      "step": 51830
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.796742857142857e-07,
+      "loss": 1.0343,
+      "step": 51831
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.794057142857143e-07,
+      "loss": 0.9828,
+      "step": 51832
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7913714285714285e-07,
+      "loss": 1.1175,
+      "step": 51833
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7886857142857145e-07,
+      "loss": 1.3786,
+      "step": 51834
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.786e-07,
+      "loss": 1.4253,
+      "step": 51835
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.783314285714286e-07,
+      "loss": 0.9627,
+      "step": 51836
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7806285714285714e-07,
+      "loss": 0.9952,
+      "step": 51837
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7779428571428572e-07,
+      "loss": 0.8499,
+      "step": 51838
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.775257142857143e-07,
+      "loss": 0.9421,
+      "step": 51839
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7725714285714286e-07,
+      "loss": 0.8621,
+      "step": 51840
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7698857142857144e-07,
+      "loss": 0.9214,
+      "step": 51841
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7672e-07,
+      "loss": 1.0875,
+      "step": 51842
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7645142857142858e-07,
+      "loss": 1.0972,
+      "step": 51843
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7618285714285716e-07,
+      "loss": 0.6136,
+      "step": 51844
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.759142857142857e-07,
+      "loss": 0.9283,
+      "step": 51845
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.756457142857143e-07,
+      "loss": 1.1236,
+      "step": 51846
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7537714285714285e-07,
+      "loss": 1.459,
+      "step": 51847
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7510857142857145e-07,
+      "loss": 1.2454,
+      "step": 51848
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7484e-07,
+      "loss": 1.4196,
+      "step": 51849
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.745714285714286e-07,
+      "loss": 1.035,
+      "step": 51850
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7430285714285714e-07,
+      "loss": 1.1085,
+      "step": 51851
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7403428571428572e-07,
+      "loss": 0.9167,
+      "step": 51852
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.737657142857143e-07,
+      "loss": 1.0705,
+      "step": 51853
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7349714285714286e-07,
+      "loss": 0.9503,
+      "step": 51854
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7322857142857144e-07,
+      "loss": 0.9969,
+      "step": 51855
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7296e-07,
+      "loss": 1.4021,
+      "step": 51856
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7269142857142858e-07,
+      "loss": 1.4965,
+      "step": 51857
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7242285714285715e-07,
+      "loss": 1.2261,
+      "step": 51858
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.721542857142857e-07,
+      "loss": 1.3907,
+      "step": 51859
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.718857142857143e-07,
+      "loss": 1.4666,
+      "step": 51860
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7161714285714285e-07,
+      "loss": 0.7812,
+      "step": 51861
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7134857142857145e-07,
+      "loss": 1.2366,
+      "step": 51862
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7108e-07,
+      "loss": 0.7317,
+      "step": 51863
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.708114285714286e-07,
+      "loss": 1.3562,
+      "step": 51864
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7054285714285714e-07,
+      "loss": 1.3747,
+      "step": 51865
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.7027428571428571e-07,
+      "loss": 1.2633,
+      "step": 51866
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.700057142857143e-07,
+      "loss": 1.1217,
+      "step": 51867
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6973714285714286e-07,
+      "loss": 1.3998,
+      "step": 51868
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6946857142857143e-07,
+      "loss": 0.8821,
+      "step": 51869
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.692e-07,
+      "loss": 0.9058,
+      "step": 51870
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6893142857142858e-07,
+      "loss": 1.1951,
+      "step": 51871
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6866285714285715e-07,
+      "loss": 1.1431,
+      "step": 51872
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.683942857142857e-07,
+      "loss": 1.4028,
+      "step": 51873
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.681257142857143e-07,
+      "loss": 1.3023,
+      "step": 51874
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6785714285714285e-07,
+      "loss": 1.3651,
+      "step": 51875
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6758857142857145e-07,
+      "loss": 1.3433,
+      "step": 51876
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6732e-07,
+      "loss": 0.8228,
+      "step": 51877
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.670514285714286e-07,
+      "loss": 0.8783,
+      "step": 51878
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6678285714285714e-07,
+      "loss": 1.0632,
+      "step": 51879
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.665142857142857e-07,
+      "loss": 1.0958,
+      "step": 51880
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6624571428571428e-07,
+      "loss": 1.3172,
+      "step": 51881
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6597714285714286e-07,
+      "loss": 1.4067,
+      "step": 51882
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6570857142857143e-07,
+      "loss": 1.1604,
+      "step": 51883
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6544e-07,
+      "loss": 1.4922,
+      "step": 51884
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6517142857142858e-07,
+      "loss": 0.859,
+      "step": 51885
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6490285714285715e-07,
+      "loss": 1.1207,
+      "step": 51886
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.646342857142857e-07,
+      "loss": 1.3911,
+      "step": 51887
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.643657142857143e-07,
+      "loss": 0.7867,
+      "step": 51888
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6409714285714284e-07,
+      "loss": 0.9857,
+      "step": 51889
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6382857142857144e-07,
+      "loss": 1.0152,
+      "step": 51890
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6356e-07,
+      "loss": 0.9533,
+      "step": 51891
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.632914285714286e-07,
+      "loss": 1.2707,
+      "step": 51892
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6302285714285714e-07,
+      "loss": 1.2776,
+      "step": 51893
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.627542857142857e-07,
+      "loss": 1.323,
+      "step": 51894
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6248571428571428e-07,
+      "loss": 1.1852,
+      "step": 51895
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6221714285714286e-07,
+      "loss": 0.9654,
+      "step": 51896
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6194857142857143e-07,
+      "loss": 1.1491,
+      "step": 51897
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6168e-07,
+      "loss": 1.1941,
+      "step": 51898
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.614114285714286e-07,
+      "loss": 1.0521,
+      "step": 51899
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6114285714285715e-07,
+      "loss": 1.4086,
+      "step": 51900
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6087428571428572e-07,
+      "loss": 1.1193,
+      "step": 51901
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.606057142857143e-07,
+      "loss": 1.0203,
+      "step": 51902
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6033714285714287e-07,
+      "loss": 1.1884,
+      "step": 51903
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.6006857142857144e-07,
+      "loss": 1.4225,
+      "step": 51904
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5980000000000001e-07,
+      "loss": 1.1331,
+      "step": 51905
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.595314285714286e-07,
+      "loss": 1.3016,
+      "step": 51906
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5926285714285716e-07,
+      "loss": 1.4066,
+      "step": 51907
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.589942857142857e-07,
+      "loss": 1.144,
+      "step": 51908
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.587257142857143e-07,
+      "loss": 0.8641,
+      "step": 51909
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5845714285714285e-07,
+      "loss": 0.7982,
+      "step": 51910
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5818857142857145e-07,
+      "loss": 0.9011,
+      "step": 51911
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5792e-07,
+      "loss": 1.1188,
+      "step": 51912
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.576514285714286e-07,
+      "loss": 1.0894,
+      "step": 51913
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5738285714285715e-07,
+      "loss": 1.1856,
+      "step": 51914
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5711428571428572e-07,
+      "loss": 0.9548,
+      "step": 51915
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.568457142857143e-07,
+      "loss": 1.3934,
+      "step": 51916
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5657714285714287e-07,
+      "loss": 0.8189,
+      "step": 51917
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5630857142857144e-07,
+      "loss": 1.3098,
+      "step": 51918
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5604e-07,
+      "loss": 0.8663,
+      "step": 51919
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5577142857142859e-07,
+      "loss": 1.1773,
+      "step": 51920
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5550285714285716e-07,
+      "loss": 1.3318,
+      "step": 51921
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.552342857142857e-07,
+      "loss": 1.3279,
+      "step": 51922
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.549657142857143e-07,
+      "loss": 1.3925,
+      "step": 51923
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5469714285714285e-07,
+      "loss": 1.5304,
+      "step": 51924
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5442857142857145e-07,
+      "loss": 1.4139,
+      "step": 51925
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5416e-07,
+      "loss": 0.9835,
+      "step": 51926
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.538914285714286e-07,
+      "loss": 1.2068,
+      "step": 51927
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5362285714285715e-07,
+      "loss": 1.0093,
+      "step": 51928
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5335428571428572e-07,
+      "loss": 1.2499,
+      "step": 51929
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.530857142857143e-07,
+      "loss": 1.3632,
+      "step": 51930
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5281714285714286e-07,
+      "loss": 0.7551,
+      "step": 51931
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5254857142857144e-07,
+      "loss": 1.2133,
+      "step": 51932
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5228e-07,
+      "loss": 1.0545,
+      "step": 51933
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5201142857142858e-07,
+      "loss": 0.7414,
+      "step": 51934
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5174285714285716e-07,
+      "loss": 0.9442,
+      "step": 51935
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.514742857142857e-07,
+      "loss": 0.9908,
+      "step": 51936
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.512057142857143e-07,
+      "loss": 1.2754,
+      "step": 51937
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5093714285714285e-07,
+      "loss": 1.2295,
+      "step": 51938
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.5066857142857145e-07,
+      "loss": 0.9149,
+      "step": 51939
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.504e-07,
+      "loss": 1.3124,
+      "step": 51940
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.501314285714286e-07,
+      "loss": 1.4257,
+      "step": 51941
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4986285714285714e-07,
+      "loss": 0.9715,
+      "step": 51942
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4959428571428572e-07,
+      "loss": 1.3029,
+      "step": 51943
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.493257142857143e-07,
+      "loss": 0.6934,
+      "step": 51944
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4905714285714286e-07,
+      "loss": 1.1816,
+      "step": 51945
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4878857142857144e-07,
+      "loss": 1.195,
+      "step": 51946
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4852e-07,
+      "loss": 1.6611,
+      "step": 51947
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4825142857142858e-07,
+      "loss": 0.8316,
+      "step": 51948
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4798285714285716e-07,
+      "loss": 1.6044,
+      "step": 51949
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.477142857142857e-07,
+      "loss": 0.7858,
+      "step": 51950
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.474457142857143e-07,
+      "loss": 0.9939,
+      "step": 51951
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4717714285714285e-07,
+      "loss": 1.0887,
+      "step": 51952
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4690857142857145e-07,
+      "loss": 1.129,
+      "step": 51953
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4664e-07,
+      "loss": 1.4424,
+      "step": 51954
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.463714285714286e-07,
+      "loss": 1.2252,
+      "step": 51955
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4610285714285714e-07,
+      "loss": 1.1231,
+      "step": 51956
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4583428571428571e-07,
+      "loss": 1.0861,
+      "step": 51957
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.455657142857143e-07,
+      "loss": 0.7526,
+      "step": 51958
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4529714285714286e-07,
+      "loss": 1.0553,
+      "step": 51959
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4502857142857143e-07,
+      "loss": 0.9471,
+      "step": 51960
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4476e-07,
+      "loss": 1.2557,
+      "step": 51961
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4449142857142858e-07,
+      "loss": 1.0797,
+      "step": 51962
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4422285714285715e-07,
+      "loss": 1.7153,
+      "step": 51963
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.439542857142857e-07,
+      "loss": 1.2486,
+      "step": 51964
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.436857142857143e-07,
+      "loss": 0.9855,
+      "step": 51965
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4341714285714285e-07,
+      "loss": 1.0107,
+      "step": 51966
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4314857142857145e-07,
+      "loss": 1.3466,
+      "step": 51967
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4288e-07,
+      "loss": 0.9466,
+      "step": 51968
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.426114285714286e-07,
+      "loss": 0.9832,
+      "step": 51969
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4234285714285714e-07,
+      "loss": 0.8456,
+      "step": 51970
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.420742857142857e-07,
+      "loss": 1.0865,
+      "step": 51971
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4180571428571429e-07,
+      "loss": 1.0649,
+      "step": 51972
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4153714285714286e-07,
+      "loss": 1.3622,
+      "step": 51973
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4126857142857143e-07,
+      "loss": 1.1196,
+      "step": 51974
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.41e-07,
+      "loss": 1.0266,
+      "step": 51975
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4073142857142858e-07,
+      "loss": 1.1047,
+      "step": 51976
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4046285714285715e-07,
+      "loss": 0.8373,
+      "step": 51977
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.401942857142857e-07,
+      "loss": 1.0614,
+      "step": 51978
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.399257142857143e-07,
+      "loss": 0.9307,
+      "step": 51979
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3965714285714284e-07,
+      "loss": 1.616,
+      "step": 51980
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3938857142857144e-07,
+      "loss": 1.3298,
+      "step": 51981
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3912e-07,
+      "loss": 1.2106,
+      "step": 51982
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.388514285714286e-07,
+      "loss": 0.9269,
+      "step": 51983
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3858285714285714e-07,
+      "loss": 1.2087,
+      "step": 51984
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.383142857142857e-07,
+      "loss": 1.1574,
+      "step": 51985
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3804571428571428e-07,
+      "loss": 1.0263,
+      "step": 51986
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3777714285714286e-07,
+      "loss": 1.0491,
+      "step": 51987
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3750857142857143e-07,
+      "loss": 1.1784,
+      "step": 51988
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3724e-07,
+      "loss": 1.4875,
+      "step": 51989
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3697142857142858e-07,
+      "loss": 1.2785,
+      "step": 51990
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3670285714285715e-07,
+      "loss": 0.8126,
+      "step": 51991
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.364342857142857e-07,
+      "loss": 1.1155,
+      "step": 51992
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.361657142857143e-07,
+      "loss": 0.9421,
+      "step": 51993
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3589714285714284e-07,
+      "loss": 1.3614,
+      "step": 51994
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3562857142857144e-07,
+      "loss": 0.878,
+      "step": 51995
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3536e-07,
+      "loss": 1.3544,
+      "step": 51996
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.350914285714286e-07,
+      "loss": 1.3586,
+      "step": 51997
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3482285714285714e-07,
+      "loss": 1.0973,
+      "step": 51998
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.345542857142857e-07,
+      "loss": 1.5794,
+      "step": 51999
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.342857142857143e-07,
+      "loss": 1.2514,
+      "step": 52000
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3401714285714286e-07,
+      "loss": 0.9543,
+      "step": 52001
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3374857142857145e-07,
+      "loss": 1.0857,
+      "step": 52002
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3348e-07,
+      "loss": 1.2017,
+      "step": 52003
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.332114285714286e-07,
+      "loss": 0.9007,
+      "step": 52004
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3294285714285715e-07,
+      "loss": 1.1663,
+      "step": 52005
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3267428571428572e-07,
+      "loss": 0.9168,
+      "step": 52006
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.324057142857143e-07,
+      "loss": 1.2803,
+      "step": 52007
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3213714285714287e-07,
+      "loss": 1.04,
+      "step": 52008
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3186857142857144e-07,
+      "loss": 1.2557,
+      "step": 52009
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3160000000000001e-07,
+      "loss": 0.7519,
+      "step": 52010
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.313314285714286e-07,
+      "loss": 0.6677,
+      "step": 52011
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3106285714285716e-07,
+      "loss": 0.9709,
+      "step": 52012
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.307942857142857e-07,
+      "loss": 0.5171,
+      "step": 52013
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.305257142857143e-07,
+      "loss": 0.916,
+      "step": 52014
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.3025714285714285e-07,
+      "loss": 1.1127,
+      "step": 52015
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2998857142857145e-07,
+      "loss": 1.1131,
+      "step": 52016
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2972e-07,
+      "loss": 1.3674,
+      "step": 52017
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.294514285714286e-07,
+      "loss": 1.1142,
+      "step": 52018
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2918285714285715e-07,
+      "loss": 1.357,
+      "step": 52019
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2891428571428572e-07,
+      "loss": 1.0549,
+      "step": 52020
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.286457142857143e-07,
+      "loss": 0.8769,
+      "step": 52021
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2837714285714287e-07,
+      "loss": 1.0214,
+      "step": 52022
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2810857142857144e-07,
+      "loss": 1.0302,
+      "step": 52023
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2784e-07,
+      "loss": 0.9615,
+      "step": 52024
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2757142857142859e-07,
+      "loss": 1.399,
+      "step": 52025
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2730285714285716e-07,
+      "loss": 1.389,
+      "step": 52026
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.270342857142857e-07,
+      "loss": 1.4382,
+      "step": 52027
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.267657142857143e-07,
+      "loss": 1.3587,
+      "step": 52028
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2649714285714285e-07,
+      "loss": 1.0883,
+      "step": 52029
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2622857142857145e-07,
+      "loss": 1.1354,
+      "step": 52030
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2596e-07,
+      "loss": 1.3793,
+      "step": 52031
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.256914285714286e-07,
+      "loss": 1.1108,
+      "step": 52032
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2542285714285714e-07,
+      "loss": 1.1197,
+      "step": 52033
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2515428571428572e-07,
+      "loss": 1.0752,
+      "step": 52034
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.248857142857143e-07,
+      "loss": 0.9269,
+      "step": 52035
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2461714285714286e-07,
+      "loss": 0.7079,
+      "step": 52036
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2434857142857144e-07,
+      "loss": 1.0901,
+      "step": 52037
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2408e-07,
+      "loss": 1.5446,
+      "step": 52038
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2381142857142858e-07,
+      "loss": 1.2632,
+      "step": 52039
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2354285714285716e-07,
+      "loss": 1.3407,
+      "step": 52040
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.232742857142857e-07,
+      "loss": 1.2472,
+      "step": 52041
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.230057142857143e-07,
+      "loss": 1.199,
+      "step": 52042
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2273714285714285e-07,
+      "loss": 1.3462,
+      "step": 52043
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2246857142857145e-07,
+      "loss": 1.2138,
+      "step": 52044
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.222e-07,
+      "loss": 1.1145,
+      "step": 52045
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.219314285714286e-07,
+      "loss": 0.8529,
+      "step": 52046
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2166285714285714e-07,
+      "loss": 1.4765,
+      "step": 52047
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2139428571428572e-07,
+      "loss": 1.2483,
+      "step": 52048
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.211257142857143e-07,
+      "loss": 1.4927,
+      "step": 52049
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2085714285714286e-07,
+      "loss": 1.0783,
+      "step": 52050
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2058857142857143e-07,
+      "loss": 1.0424,
+      "step": 52051
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2032e-07,
+      "loss": 1.1277,
+      "step": 52052
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.2005142857142858e-07,
+      "loss": 1.0164,
+      "step": 52053
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.1978285714285715e-07,
+      "loss": 1.3274,
+      "step": 52054
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.195142857142857e-07,
+      "loss": 1.1585,
+      "step": 52055
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.192457142857143e-07,
+      "loss": 1.1211,
+      "step": 52056
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.1897714285714286e-07,
+      "loss": 1.3354,
+      "step": 52057
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.1870857142857143e-07,
+      "loss": 1.0938,
+      "step": 52058
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.1844e-07,
+      "loss": 1.3116,
+      "step": 52059
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.1817142857142858e-07,
+      "loss": 0.7945,
+      "step": 52060
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.1790285714285714e-07,
+      "loss": 1.2352,
+      "step": 52061
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.1763428571428573e-07,
+      "loss": 0.9148,
+      "step": 52062
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1736571428571429e-07,
+      "loss": 0.5954,
+      "step": 52063
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1709714285714286e-07,
+      "loss": 1.1745,
+      "step": 52064
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1682857142857142e-07,
+      "loss": 1.2023,
+      "step": 52065
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1656e-07,
+      "loss": 0.9944,
+      "step": 52066
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1629142857142857e-07,
+      "loss": 1.2722,
+      "step": 52067
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1602285714285715e-07,
+      "loss": 1.5742,
+      "step": 52068
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1575428571428571e-07,
+      "loss": 0.9057,
+      "step": 52069
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.154857142857143e-07,
+      "loss": 1.0277,
+      "step": 52070
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1521714285714286e-07,
+      "loss": 1.1131,
+      "step": 52071
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1494857142857143e-07,
+      "loss": 1.4247,
+      "step": 52072
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1467999999999999e-07,
+      "loss": 1.4166,
+      "step": 52073
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1441142857142858e-07,
+      "loss": 0.9056,
+      "step": 52074
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1414285714285714e-07,
+      "loss": 1.1512,
+      "step": 52075
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1387428571428572e-07,
+      "loss": 1.3927,
+      "step": 52076
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1360571428571428e-07,
+      "loss": 1.1035,
+      "step": 52077
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1333714285714286e-07,
+      "loss": 0.8422,
+      "step": 52078
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1306857142857142e-07,
+      "loss": 0.9065,
+      "step": 52079
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.128e-07,
+      "loss": 1.1488,
+      "step": 52080
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1253142857142856e-07,
+      "loss": 0.8922,
+      "step": 52081
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1226285714285715e-07,
+      "loss": 1.4363,
+      "step": 52082
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1199428571428571e-07,
+      "loss": 1.1877,
+      "step": 52083
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.117257142857143e-07,
+      "loss": 0.761,
+      "step": 52084
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1145714285714286e-07,
+      "loss": 1.3323,
+      "step": 52085
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1118857142857143e-07,
+      "loss": 1.3703,
+      "step": 52086
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1091999999999999e-07,
+      "loss": 1.126,
+      "step": 52087
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1065142857142858e-07,
+      "loss": 1.2842,
+      "step": 52088
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1038285714285714e-07,
+      "loss": 1.6387,
+      "step": 52089
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.1011428571428572e-07,
+      "loss": 1.1984,
+      "step": 52090
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.098457142857143e-07,
+      "loss": 1.3022,
+      "step": 52091
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0957714285714286e-07,
+      "loss": 0.7684,
+      "step": 52092
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0930857142857143e-07,
+      "loss": 1.1479,
+      "step": 52093
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0904e-07,
+      "loss": 0.9545,
+      "step": 52094
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0877142857142858e-07,
+      "loss": 1.1673,
+      "step": 52095
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0850285714285715e-07,
+      "loss": 0.9939,
+      "step": 52096
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0823428571428572e-07,
+      "loss": 1.5905,
+      "step": 52097
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.079657142857143e-07,
+      "loss": 1.0307,
+      "step": 52098
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0769714285714285e-07,
+      "loss": 0.6209,
+      "step": 52099
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0742857142857143e-07,
+      "loss": 1.1867,
+      "step": 52100
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0716e-07,
+      "loss": 1.2767,
+      "step": 52101
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0689142857142857e-07,
+      "loss": 1.0219,
+      "step": 52102
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0662285714285715e-07,
+      "loss": 1.1001,
+      "step": 52103
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0635428571428572e-07,
+      "loss": 0.891,
+      "step": 52104
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.060857142857143e-07,
+      "loss": 1.0464,
+      "step": 52105
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0581714285714285e-07,
+      "loss": 1.1884,
+      "step": 52106
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0554857142857143e-07,
+      "loss": 1.0451,
+      "step": 52107
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0528e-07,
+      "loss": 1.0732,
+      "step": 52108
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0501142857142857e-07,
+      "loss": 1.2786,
+      "step": 52109
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0474285714285715e-07,
+      "loss": 1.2847,
+      "step": 52110
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0447428571428572e-07,
+      "loss": 1.218,
+      "step": 52111
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0420571428571429e-07,
+      "loss": 0.7457,
+      "step": 52112
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0393714285714285e-07,
+      "loss": 1.3411,
+      "step": 52113
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0366857142857143e-07,
+      "loss": 0.7192,
+      "step": 52114
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.034e-07,
+      "loss": 0.9691,
+      "step": 52115
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0313142857142857e-07,
+      "loss": 1.0593,
+      "step": 52116
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0286285714285715e-07,
+      "loss": 1.3171,
+      "step": 52117
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0259428571428572e-07,
+      "loss": 1.1227,
+      "step": 52118
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0232571428571429e-07,
+      "loss": 0.666,
+      "step": 52119
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0205714285714285e-07,
+      "loss": 1.0447,
+      "step": 52120
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0178857142857143e-07,
+      "loss": 1.5576,
+      "step": 52121
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0152e-07,
+      "loss": 1.3469,
+      "step": 52122
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0125142857142857e-07,
+      "loss": 0.6946,
+      "step": 52123
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0098285714285714e-07,
+      "loss": 0.8334,
+      "step": 52124
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0071428571428572e-07,
+      "loss": 1.1164,
+      "step": 52125
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0044571428571429e-07,
+      "loss": 1.1711,
+      "step": 52126
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.0017714285714285e-07,
+      "loss": 1.2188,
+      "step": 52127
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.990857142857142e-08,
+      "loss": 1.4679,
+      "step": 52128
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.964e-08,
+      "loss": 1.3431,
+      "step": 52129
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.937142857142857e-08,
+      "loss": 1.0619,
+      "step": 52130
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.910285714285714e-08,
+      "loss": 1.2297,
+      "step": 52131
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.883428571428572e-08,
+      "loss": 1.5661,
+      "step": 52132
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.856571428571429e-08,
+      "loss": 1.1849,
+      "step": 52133
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.829714285714285e-08,
+      "loss": 0.8992,
+      "step": 52134
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.802857142857142e-08,
+      "loss": 1.3192,
+      "step": 52135
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.776e-08,
+      "loss": 1.2442,
+      "step": 52136
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.749142857142857e-08,
+      "loss": 1.3075,
+      "step": 52137
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.722285714285714e-08,
+      "loss": 1.4078,
+      "step": 52138
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.695428571428572e-08,
+      "loss": 1.1321,
+      "step": 52139
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.668571428571429e-08,
+      "loss": 1.1322,
+      "step": 52140
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.641714285714285e-08,
+      "loss": 0.8491,
+      "step": 52141
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.614857142857142e-08,
+      "loss": 1.3953,
+      "step": 52142
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.588e-08,
+      "loss": 1.1226,
+      "step": 52143
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.561142857142857e-08,
+      "loss": 0.8728,
+      "step": 52144
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.534285714285714e-08,
+      "loss": 1.1174,
+      "step": 52145
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.507428571428572e-08,
+      "loss": 1.0103,
+      "step": 52146
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.480571428571429e-08,
+      "loss": 1.0386,
+      "step": 52147
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.453714285714285e-08,
+      "loss": 1.2322,
+      "step": 52148
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.426857142857142e-08,
+      "loss": 0.9309,
+      "step": 52149
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.400000000000001e-08,
+      "loss": 1.2506,
+      "step": 52150
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.373142857142858e-08,
+      "loss": 0.9656,
+      "step": 52151
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.346285714285715e-08,
+      "loss": 1.3195,
+      "step": 52152
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.319428571428573e-08,
+      "loss": 1.1847,
+      "step": 52153
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.29257142857143e-08,
+      "loss": 1.2693,
+      "step": 52154
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.265714285714286e-08,
+      "loss": 1.3854,
+      "step": 52155
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.238857142857143e-08,
+      "loss": 1.0681,
+      "step": 52156
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.212000000000001e-08,
+      "loss": 1.0846,
+      "step": 52157
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.185142857142858e-08,
+      "loss": 1.4926,
+      "step": 52158
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.158285714285715e-08,
+      "loss": 1.1491,
+      "step": 52159
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.131428571428573e-08,
+      "loss": 0.8628,
+      "step": 52160
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.10457142857143e-08,
+      "loss": 0.889,
+      "step": 52161
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.077714285714286e-08,
+      "loss": 1.0188,
+      "step": 52162
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.050857142857143e-08,
+      "loss": 0.7439,
+      "step": 52163
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.024e-08,
+      "loss": 1.0583,
+      "step": 52164
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.997142857142858e-08,
+      "loss": 0.8365,
+      "step": 52165
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.970285714285715e-08,
+      "loss": 0.5351,
+      "step": 52166
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.943428571428573e-08,
+      "loss": 1.2207,
+      "step": 52167
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.91657142857143e-08,
+      "loss": 1.2365,
+      "step": 52168
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.889714285714286e-08,
+      "loss": 1.062,
+      "step": 52169
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.862857142857143e-08,
+      "loss": 0.861,
+      "step": 52170
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.836e-08,
+      "loss": 1.3075,
+      "step": 52171
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.809142857142858e-08,
+      "loss": 1.1609,
+      "step": 52172
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.782285714285715e-08,
+      "loss": 1.0566,
+      "step": 52173
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.755428571428572e-08,
+      "loss": 1.2969,
+      "step": 52174
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.72857142857143e-08,
+      "loss": 0.9114,
+      "step": 52175
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.701714285714286e-08,
+      "loss": 1.0157,
+      "step": 52176
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.674857142857143e-08,
+      "loss": 0.9451,
+      "step": 52177
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.648e-08,
+      "loss": 1.0278,
+      "step": 52178
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.621142857142858e-08,
+      "loss": 1.5033,
+      "step": 52179
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.594285714285715e-08,
+      "loss": 1.1565,
+      "step": 52180
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.567428571428572e-08,
+      "loss": 1.2969,
+      "step": 52181
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.54057142857143e-08,
+      "loss": 1.5452,
+      "step": 52182
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.513714285714286e-08,
+      "loss": 0.8987,
+      "step": 52183
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.486857142857143e-08,
+      "loss": 0.8476,
+      "step": 52184
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.46e-08,
+      "loss": 1.2141,
+      "step": 52185
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.433142857142858e-08,
+      "loss": 1.135,
+      "step": 52186
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.406285714285715e-08,
+      "loss": 1.051,
+      "step": 52187
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.379428571428572e-08,
+      "loss": 1.1785,
+      "step": 52188
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.35257142857143e-08,
+      "loss": 1.0047,
+      "step": 52189
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.325714285714286e-08,
+      "loss": 1.1851,
+      "step": 52190
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.298857142857143e-08,
+      "loss": 0.8744,
+      "step": 52191
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.272e-08,
+      "loss": 1.1235,
+      "step": 52192
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.245142857142858e-08,
+      "loss": 1.0576,
+      "step": 52193
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.218285714285715e-08,
+      "loss": 1.125,
+      "step": 52194
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.191428571428572e-08,
+      "loss": 1.2061,
+      "step": 52195
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.16457142857143e-08,
+      "loss": 1.4171,
+      "step": 52196
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.137714285714285e-08,
+      "loss": 1.138,
+      "step": 52197
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.110857142857143e-08,
+      "loss": 0.9285,
+      "step": 52198
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.084e-08,
+      "loss": 1.2177,
+      "step": 52199
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.057142857142857e-08,
+      "loss": 1.1731,
+      "step": 52200
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.030285714285715e-08,
+      "loss": 0.6556,
+      "step": 52201
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.003428571428572e-08,
+      "loss": 0.9912,
+      "step": 52202
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.97657142857143e-08,
+      "loss": 1.1725,
+      "step": 52203
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.949714285714285e-08,
+      "loss": 1.1378,
+      "step": 52204
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.922857142857143e-08,
+      "loss": 1.2602,
+      "step": 52205
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.896e-08,
+      "loss": 0.9435,
+      "step": 52206
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.869142857142857e-08,
+      "loss": 1.0746,
+      "step": 52207
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.842285714285715e-08,
+      "loss": 1.2571,
+      "step": 52208
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.815428571428572e-08,
+      "loss": 1.5281,
+      "step": 52209
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.788571428571429e-08,
+      "loss": 0.7756,
+      "step": 52210
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.761714285714285e-08,
+      "loss": 1.0612,
+      "step": 52211
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.734857142857143e-08,
+      "loss": 0.8167,
+      "step": 52212
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.708e-08,
+      "loss": 1.3683,
+      "step": 52213
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.681142857142857e-08,
+      "loss": 0.8321,
+      "step": 52214
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.654285714285715e-08,
+      "loss": 1.2667,
+      "step": 52215
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.627428571428572e-08,
+      "loss": 1.011,
+      "step": 52216
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.600571428571429e-08,
+      "loss": 1.1345,
+      "step": 52217
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.573714285714285e-08,
+      "loss": 1.1104,
+      "step": 52218
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.546857142857143e-08,
+      "loss": 1.3278,
+      "step": 52219
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.52e-08,
+      "loss": 1.2426,
+      "step": 52220
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.493142857142857e-08,
+      "loss": 0.7134,
+      "step": 52221
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.466285714285714e-08,
+      "loss": 1.3508,
+      "step": 52222
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.439428571428572e-08,
+      "loss": 1.2062,
+      "step": 52223
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.412571428571429e-08,
+      "loss": 0.7649,
+      "step": 52224
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.385714285714285e-08,
+      "loss": 0.71,
+      "step": 52225
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.358857142857142e-08,
+      "loss": 0.9541,
+      "step": 52226
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.332e-08,
+      "loss": 1.3067,
+      "step": 52227
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.305142857142857e-08,
+      "loss": 0.8106,
+      "step": 52228
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.278285714285714e-08,
+      "loss": 1.0071,
+      "step": 52229
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.251428571428572e-08,
+      "loss": 1.6682,
+      "step": 52230
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.224571428571429e-08,
+      "loss": 1.1546,
+      "step": 52231
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.197714285714285e-08,
+      "loss": 0.9133,
+      "step": 52232
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.170857142857142e-08,
+      "loss": 1.2602,
+      "step": 52233
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.144e-08,
+      "loss": 0.9101,
+      "step": 52234
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.117142857142857e-08,
+      "loss": 0.7478,
+      "step": 52235
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.090285714285714e-08,
+      "loss": 0.8074,
+      "step": 52236
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.063428571428572e-08,
+      "loss": 1.5266,
+      "step": 52237
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.036571428571429e-08,
+      "loss": 0.9749,
+      "step": 52238
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.009714285714285e-08,
+      "loss": 1.1277,
+      "step": 52239
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.982857142857142e-08,
+      "loss": 0.8014,
+      "step": 52240
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.956e-08,
+      "loss": 1.3175,
+      "step": 52241
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.929142857142857e-08,
+      "loss": 1.0899,
+      "step": 52242
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.902285714285714e-08,
+      "loss": 0.8027,
+      "step": 52243
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.875428571428572e-08,
+      "loss": 0.8931,
+      "step": 52244
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.848571428571429e-08,
+      "loss": 1.3826,
+      "step": 52245
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.821714285714285e-08,
+      "loss": 0.6581,
+      "step": 52246
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.794857142857142e-08,
+      "loss": 1.3961,
+      "step": 52247
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.768e-08,
+      "loss": 1.225,
+      "step": 52248
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.741142857142857e-08,
+      "loss": 1.1399,
+      "step": 52249
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.714285714285715e-08,
+      "loss": 1.0592,
+      "step": 52250
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.687428571428573e-08,
+      "loss": 1.3216,
+      "step": 52251
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.66057142857143e-08,
+      "loss": 0.8697,
+      "step": 52252
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.633714285714286e-08,
+      "loss": 0.8913,
+      "step": 52253
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.606857142857143e-08,
+      "loss": 1.2671,
+      "step": 52254
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.580000000000001e-08,
+      "loss": 1.277,
+      "step": 52255
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.553142857142858e-08,
+      "loss": 1.3252,
+      "step": 52256
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.526285714285715e-08,
+      "loss": 1.2122,
+      "step": 52257
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.499428571428573e-08,
+      "loss": 0.9535,
+      "step": 52258
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.47257142857143e-08,
+      "loss": 1.077,
+      "step": 52259
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.445714285714286e-08,
+      "loss": 1.6015,
+      "step": 52260
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.418857142857143e-08,
+      "loss": 1.134,
+      "step": 52261
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.392e-08,
+      "loss": 1.5104,
+      "step": 52262
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.365142857142858e-08,
+      "loss": 1.4905,
+      "step": 52263
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.338285714285715e-08,
+      "loss": 1.2832,
+      "step": 52264
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.311428571428573e-08,
+      "loss": 1.0836,
+      "step": 52265
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.28457142857143e-08,
+      "loss": 1.1738,
+      "step": 52266
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.257714285714286e-08,
+      "loss": 1.2634,
+      "step": 52267
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.230857142857143e-08,
+      "loss": 1.6029,
+      "step": 52268
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.204e-08,
+      "loss": 1.0609,
+      "step": 52269
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.177142857142858e-08,
+      "loss": 1.3378,
+      "step": 52270
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.150285714285715e-08,
+      "loss": 1.2616,
+      "step": 52271
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.123428571428572e-08,
+      "loss": 1.0564,
+      "step": 52272
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.09657142857143e-08,
+      "loss": 1.1634,
+      "step": 52273
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.069714285714286e-08,
+      "loss": 1.2951,
+      "step": 52274
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.042857142857143e-08,
+      "loss": 1.5231,
+      "step": 52275
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.016e-08,
+      "loss": 0.6965,
+      "step": 52276
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.989142857142858e-08,
+      "loss": 1.1947,
+      "step": 52277
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.962285714285715e-08,
+      "loss": 0.8376,
+      "step": 52278
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.935428571428572e-08,
+      "loss": 1.1304,
+      "step": 52279
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.908571428571429e-08,
+      "loss": 0.9312,
+      "step": 52280
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.8817142857142863e-08,
+      "loss": 1.2192,
+      "step": 52281
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.854857142857143e-08,
+      "loss": 0.9411,
+      "step": 52282
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.828e-08,
+      "loss": 0.7408,
+      "step": 52283
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.8011428571428576e-08,
+      "loss": 1.2473,
+      "step": 52284
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.774285714285715e-08,
+      "loss": 1.08,
+      "step": 52285
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.7474285714285716e-08,
+      "loss": 0.7927,
+      "step": 52286
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.720571428571429e-08,
+      "loss": 1.2525,
+      "step": 52287
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.693714285714286e-08,
+      "loss": 0.6871,
+      "step": 52288
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.666857142857143e-08,
+      "loss": 1.574,
+      "step": 52289
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.64e-08,
+      "loss": 1.1604,
+      "step": 52290
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.6131428571428575e-08,
+      "loss": 1.424,
+      "step": 52291
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.586285714285715e-08,
+      "loss": 0.9813,
+      "step": 52292
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.5594285714285715e-08,
+      "loss": 1.3557,
+      "step": 52293
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.532571428571429e-08,
+      "loss": 0.7341,
+      "step": 52294
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.505714285714286e-08,
+      "loss": 0.9177,
+      "step": 52295
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.478857142857143e-08,
+      "loss": 1.0999,
+      "step": 52296
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.452e-08,
+      "loss": 0.7243,
+      "step": 52297
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.4251428571428574e-08,
+      "loss": 1.2712,
+      "step": 52298
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.398285714285715e-08,
+      "loss": 1.0559,
+      "step": 52299
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.3714285714285714e-08,
+      "loss": 1.0048,
+      "step": 52300
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.344571428571429e-08,
+      "loss": 0.7955,
+      "step": 52301
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.317714285714286e-08,
+      "loss": 1.0483,
+      "step": 52302
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.290857142857143e-08,
+      "loss": 1.3484,
+      "step": 52303
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.264e-08,
+      "loss": 1.1285,
+      "step": 52304
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.2371428571428573e-08,
+      "loss": 1.0045,
+      "step": 52305
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.2102857142857147e-08,
+      "loss": 1.1712,
+      "step": 52306
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.183428571428571e-08,
+      "loss": 0.7025,
+      "step": 52307
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.1565714285714286e-08,
+      "loss": 0.9425,
+      "step": 52308
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.129714285714286e-08,
+      "loss": 1.0889,
+      "step": 52309
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.1028571428571426e-08,
+      "loss": 1.3793,
+      "step": 52310
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.076e-08,
+      "loss": 0.8148,
+      "step": 52311
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.049142857142857e-08,
+      "loss": 1.2687,
+      "step": 52312
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.0222857142857146e-08,
+      "loss": 1.6758,
+      "step": 52313
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.995428571428571e-08,
+      "loss": 1.022,
+      "step": 52314
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.9685714285714285e-08,
+      "loss": 1.0484,
+      "step": 52315
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.941714285714286e-08,
+      "loss": 1.0121,
+      "step": 52316
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.9148571428571425e-08,
+      "loss": 1.1785,
+      "step": 52317
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.888e-08,
+      "loss": 1.0976,
+      "step": 52318
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.861142857142857e-08,
+      "loss": 1.3914,
+      "step": 52319
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.8342857142857145e-08,
+      "loss": 1.3219,
+      "step": 52320
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.807428571428571e-08,
+      "loss": 1.3184,
+      "step": 52321
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.7805714285714284e-08,
+      "loss": 1.1411,
+      "step": 52322
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.753714285714286e-08,
+      "loss": 0.987,
+      "step": 52323
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.7268571428571424e-08,
+      "loss": 1.7952,
+      "step": 52324
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.7000000000000004e-08,
+      "loss": 0.8158,
+      "step": 52325
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.673142857142858e-08,
+      "loss": 1.1351,
+      "step": 52326
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.646285714285715e-08,
+      "loss": 1.0304,
+      "step": 52327
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.619428571428572e-08,
+      "loss": 1.4957,
+      "step": 52328
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.592571428571429e-08,
+      "loss": 0.943,
+      "step": 52329
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.565714285714286e-08,
+      "loss": 1.27,
+      "step": 52330
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.538857142857143e-08,
+      "loss": 1.1377,
+      "step": 52331
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.512e-08,
+      "loss": 1.5718,
+      "step": 52332
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.4851428571428576e-08,
+      "loss": 1.2657,
+      "step": 52333
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.458285714285715e-08,
+      "loss": 0.9229,
+      "step": 52334
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.4314285714285716e-08,
+      "loss": 1.1193,
+      "step": 52335
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.404571428571429e-08,
+      "loss": 1.0519,
+      "step": 52336
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.377714285714286e-08,
+      "loss": 1.195,
+      "step": 52337
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.350857142857143e-08,
+      "loss": 0.8882,
+      "step": 52338
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.324e-08,
+      "loss": 1.251,
+      "step": 52339
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2971428571428575e-08,
+      "loss": 1.2299,
+      "step": 52340
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.270285714285715e-08,
+      "loss": 1.0086,
+      "step": 52341
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2434285714285715e-08,
+      "loss": 1.034,
+      "step": 52342
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.216571428571429e-08,
+      "loss": 1.5285,
+      "step": 52343
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.189714285714286e-08,
+      "loss": 1.2067,
+      "step": 52344
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.162857142857143e-08,
+      "loss": 1.3179,
+      "step": 52345
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.136e-08,
+      "loss": 1.1034,
+      "step": 52346
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.1091428571428574e-08,
+      "loss": 0.7574,
+      "step": 52347
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.082285714285715e-08,
+      "loss": 1.2586,
+      "step": 52348
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.0554285714285714e-08,
+      "loss": 1.0076,
+      "step": 52349
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.028571428571429e-08,
+      "loss": 1.0162,
+      "step": 52350
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.001714285714286e-08,
+      "loss": 1.1756,
+      "step": 52351
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.974857142857143e-08,
+      "loss": 0.8984,
+      "step": 52352
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.948e-08,
+      "loss": 1.0142,
+      "step": 52353
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.9211428571428573e-08,
+      "loss": 0.9051,
+      "step": 52354
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.8942857142857147e-08,
+      "loss": 1.3901,
+      "step": 52355
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.867428571428571e-08,
+      "loss": 1.001,
+      "step": 52356
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.8405714285714286e-08,
+      "loss": 0.9258,
+      "step": 52357
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.813714285714286e-08,
+      "loss": 1.2592,
+      "step": 52358
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.7868571428571426e-08,
+      "loss": 0.8324,
+      "step": 52359
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.76e-08,
+      "loss": 0.7572,
+      "step": 52360
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.733142857142857e-08,
+      "loss": 0.946,
+      "step": 52361
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.7062857142857146e-08,
+      "loss": 1.2066,
+      "step": 52362
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.679428571428571e-08,
+      "loss": 1.6975,
+      "step": 52363
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.6525714285714285e-08,
+      "loss": 0.8313,
+      "step": 52364
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.625714285714286e-08,
+      "loss": 1.2995,
+      "step": 52365
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.5988571428571425e-08,
+      "loss": 1.2835,
+      "step": 52366
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.572e-08,
+      "loss": 1.4257,
+      "step": 52367
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.545142857142857e-08,
+      "loss": 1.4026,
+      "step": 52368
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.5182857142857145e-08,
+      "loss": 0.8374,
+      "step": 52369
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.491428571428571e-08,
+      "loss": 1.1144,
+      "step": 52370
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.4645714285714284e-08,
+      "loss": 1.1972,
+      "step": 52371
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.437714285714286e-08,
+      "loss": 1.1517,
+      "step": 52372
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.4108571428571424e-08,
+      "loss": 1.1846,
+      "step": 52373
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.384e-08,
+      "loss": 1.5695,
+      "step": 52374
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.357142857142858e-08,
+      "loss": 1.0801,
+      "step": 52375
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.330285714285715e-08,
+      "loss": 0.9468,
+      "step": 52376
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.303428571428572e-08,
+      "loss": 1.246,
+      "step": 52377
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.276571428571429e-08,
+      "loss": 1.3333,
+      "step": 52378
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.249714285714286e-08,
+      "loss": 1.536,
+      "step": 52379
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.222857142857143e-08,
+      "loss": 1.204,
+      "step": 52380
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.196e-08,
+      "loss": 0.903,
+      "step": 52381
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.1691428571428576e-08,
+      "loss": 1.5717,
+      "step": 52382
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.142285714285715e-08,
+      "loss": 0.844,
+      "step": 52383
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.1154285714285716e-08,
+      "loss": 0.7447,
+      "step": 52384
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.088571428571429e-08,
+      "loss": 1.0381,
+      "step": 52385
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.061714285714286e-08,
+      "loss": 1.1512,
+      "step": 52386
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.034857142857143e-08,
+      "loss": 1.0817,
+      "step": 52387
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.008e-08,
+      "loss": 1.1255,
+      "step": 52388
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.9811428571428575e-08,
+      "loss": 1.3102,
+      "step": 52389
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.9542857142857145e-08,
+      "loss": 1.1614,
+      "step": 52390
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.9274285714285715e-08,
+      "loss": 1.2057,
+      "step": 52391
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.9005714285714288e-08,
+      "loss": 1.0566,
+      "step": 52392
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.8737142857142858e-08,
+      "loss": 1.4086,
+      "step": 52393
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.846857142857143e-08,
+      "loss": 1.2092,
+      "step": 52394
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.82e-08,
+      "loss": 0.9427,
+      "step": 52395
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.7931428571428574e-08,
+      "loss": 1.075,
+      "step": 52396
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.7662857142857144e-08,
+      "loss": 0.8663,
+      "step": 52397
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.7394285714285714e-08,
+      "loss": 0.6832,
+      "step": 52398
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.7125714285714287e-08,
+      "loss": 1.1982,
+      "step": 52399
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.6857142857142857e-08,
+      "loss": 1.0376,
+      "step": 52400
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.658857142857143e-08,
+      "loss": 1.3312,
+      "step": 52401
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.632e-08,
+      "loss": 1.0754,
+      "step": 52402
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.6051428571428573e-08,
+      "loss": 0.9593,
+      "step": 52403
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.5782857142857143e-08,
+      "loss": 1.0529,
+      "step": 52404
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.5514285714285713e-08,
+      "loss": 1.3507,
+      "step": 52405
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.5245714285714286e-08,
+      "loss": 1.116,
+      "step": 52406
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.4977142857142856e-08,
+      "loss": 1.2243,
+      "step": 52407
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.470857142857143e-08,
+      "loss": 1.17,
+      "step": 52408
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.444e-08,
+      "loss": 1.247,
+      "step": 52409
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.4171428571428572e-08,
+      "loss": 1.0772,
+      "step": 52410
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.3902857142857142e-08,
+      "loss": 0.9309,
+      "step": 52411
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.3634285714285712e-08,
+      "loss": 1.3779,
+      "step": 52412
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.336571428571429e-08,
+      "loss": 1.3017,
+      "step": 52413
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.309714285714286e-08,
+      "loss": 1.1081,
+      "step": 52414
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.282857142857143e-08,
+      "loss": 1.4158,
+      "step": 52415
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.256e-08,
+      "loss": 1.1348,
+      "step": 52416
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.2291428571428575e-08,
+      "loss": 1.4476,
+      "step": 52417
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.2022857142857145e-08,
+      "loss": 1.5406,
+      "step": 52418
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1754285714285714e-08,
+      "loss": 1.4575,
+      "step": 52419
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1485714285714288e-08,
+      "loss": 1.3102,
+      "step": 52420
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1217142857142857e-08,
+      "loss": 1.2784,
+      "step": 52421
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.094857142857143e-08,
+      "loss": 1.0815,
+      "step": 52422
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.068e-08,
+      "loss": 1.1394,
+      "step": 52423
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.0411428571428574e-08,
+      "loss": 1.0648,
+      "step": 52424
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.0142857142857144e-08,
+      "loss": 1.3986,
+      "step": 52425
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.9874285714285713e-08,
+      "loss": 0.9637,
+      "step": 52426
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.9605714285714287e-08,
+      "loss": 0.8671,
+      "step": 52427
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.9337142857142857e-08,
+      "loss": 1.2536,
+      "step": 52428
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.906857142857143e-08,
+      "loss": 1.1428,
+      "step": 52429
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.88e-08,
+      "loss": 1.2869,
+      "step": 52430
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.8531428571428573e-08,
+      "loss": 1.2773,
+      "step": 52431
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.8262857142857143e-08,
+      "loss": 1.4216,
+      "step": 52432
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.7994285714285713e-08,
+      "loss": 0.7962,
+      "step": 52433
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.7725714285714286e-08,
+      "loss": 0.8905,
+      "step": 52434
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.7457142857142856e-08,
+      "loss": 1.1153,
+      "step": 52435
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.718857142857143e-08,
+      "loss": 0.8943,
+      "step": 52436
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.692e-08,
+      "loss": 1.262,
+      "step": 52437
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.6651428571428575e-08,
+      "loss": 1.215,
+      "step": 52438
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.6382857142857145e-08,
+      "loss": 1.3228,
+      "step": 52439
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.6114285714285715e-08,
+      "loss": 1.3533,
+      "step": 52440
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.5845714285714288e-08,
+      "loss": 1.4094,
+      "step": 52441
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.5577142857142858e-08,
+      "loss": 1.4444,
+      "step": 52442
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.530857142857143e-08,
+      "loss": 1.2047,
+      "step": 52443
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.504e-08,
+      "loss": 0.7092,
+      "step": 52444
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.4771428571428573e-08,
+      "loss": 1.3908,
+      "step": 52445
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.4502857142857144e-08,
+      "loss": 0.585,
+      "step": 52446
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.4234285714285716e-08,
+      "loss": 1.1535,
+      "step": 52447
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.3965714285714287e-08,
+      "loss": 1.5997,
+      "step": 52448
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.3697142857142857e-08,
+      "loss": 1.1935,
+      "step": 52449
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.3428571428571429e-08,
+      "loss": 0.5365,
+      "step": 52450
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.316e-08,
+      "loss": 0.8667,
+      "step": 52451
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.2891428571428572e-08,
+      "loss": 1.1835,
+      "step": 52452
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.2622857142857143e-08,
+      "loss": 1.2893,
+      "step": 52453
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.2354285714285715e-08,
+      "loss": 1.6043,
+      "step": 52454
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.2085714285714286e-08,
+      "loss": 1.4081,
+      "step": 52455
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.1817142857142856e-08,
+      "loss": 0.8373,
+      "step": 52456
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.154857142857143e-08,
+      "loss": 1.0508,
+      "step": 52457
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.128e-08,
+      "loss": 1.2877,
+      "step": 52458
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.1011428571428572e-08,
+      "loss": 1.4601,
+      "step": 52459
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.0742857142857144e-08,
+      "loss": 1.4806,
+      "step": 52460
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.0474285714285715e-08,
+      "loss": 1.076,
+      "step": 52461
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.0205714285714287e-08,
+      "loss": 1.1308,
+      "step": 52462
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 9.937142857142857e-09,
+      "loss": 1.0298,
+      "step": 52463
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 9.668571428571428e-09,
+      "loss": 0.8237,
+      "step": 52464
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 9.4e-09,
+      "loss": 1.2788,
+      "step": 52465
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 9.131428571428571e-09,
+      "loss": 1.0907,
+      "step": 52466
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.862857142857143e-09,
+      "loss": 0.756,
+      "step": 52467
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.594285714285714e-09,
+      "loss": 0.986,
+      "step": 52468
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.325714285714288e-09,
+      "loss": 0.9607,
+      "step": 52469
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.057142857142857e-09,
+      "loss": 1.2852,
+      "step": 52470
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.788571428571429e-09,
+      "loss": 1.2518,
+      "step": 52471
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.52e-09,
+      "loss": 0.9463,
+      "step": 52472
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.251428571428572e-09,
+      "loss": 0.7156,
+      "step": 52473
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.9828571428571436e-09,
+      "loss": 1.1035,
+      "step": 52474
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.714285714285714e-09,
+      "loss": 1.3405,
+      "step": 52475
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.445714285714286e-09,
+      "loss": 1.471,
+      "step": 52476
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.177142857142857e-09,
+      "loss": 0.9897,
+      "step": 52477
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.908571428571428e-09,
+      "loss": 1.1065,
+      "step": 52478
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.64e-09,
+      "loss": 1.3486,
+      "step": 52479
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.371428571428572e-09,
+      "loss": 1.2495,
+      "step": 52480
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.1028571428571434e-09,
+      "loss": 1.3736,
+      "step": 52481
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.834285714285714e-09,
+      "loss": 1.2326,
+      "step": 52482
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.565714285714286e-09,
+      "loss": 1.1533,
+      "step": 52483
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.297142857142857e-09,
+      "loss": 1.0382,
+      "step": 52484
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.028571428571429e-09,
+      "loss": 1.0888,
+      "step": 52485
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.76e-09,
+      "loss": 0.9944,
+      "step": 52486
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.4914285714285718e-09,
+      "loss": 1.3519,
+      "step": 52487
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.222857142857143e-09,
+      "loss": 1.748,
+      "step": 52488
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.954285714285714e-09,
+      "loss": 1.0751,
+      "step": 52489
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.685714285714286e-09,
+      "loss": 1.2507,
+      "step": 52490
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.417142857142857e-09,
+      "loss": 0.9434,
+      "step": 52491
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1485714285714286e-09,
+      "loss": 1.079,
+      "step": 52492
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.88e-09,
+      "loss": 1.0413,
+      "step": 52493
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.6114285714285714e-09,
+      "loss": 0.8492,
+      "step": 52494
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.342857142857143e-09,
+      "loss": 1.3799,
+      "step": 52495
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.0742857142857143e-09,
+      "loss": 1.2282,
+      "step": 52496
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.057142857142857e-10,
+      "loss": 1.5034,
+      "step": 52497
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 5.371428571428571e-10,
+      "loss": 1.5059,
+      "step": 52498
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.685714285714286e-10,
+      "loss": 1.2937,
+      "step": 52499
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0,
+      "loss": 0.9877,
+      "step": 52500
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 52500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "total_flos": 2.455750530994176e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}